<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Lior Swissa on Medium]]></title>
        <description><![CDATA[Stories by Lior Swissa on Medium]]></description>
        <link>https://medium.com/@liorswissa?source=rss-1ae9eb713cd8------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*xK5hG2YFikyOwY0SFzE47g.jpeg</url>
            <title>Stories by Lior Swissa on Medium</title>
            <link>https://medium.com/@liorswissa?source=rss-1ae9eb713cd8------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 01:41:46 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@liorswissa/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[From Pistons to Pixels: A Journey through the Changing Landscape of Vehicle Parts]]></title>
            <link>https://medium.com/@liorswissa/from-pistons-to-pixels-a-journey-through-the-changing-landscape-of-vehicle-parts-e92a3c5385ba?source=rss-1ae9eb713cd8------2</link>
            <guid isPermaLink="false">https://medium.com/p/e92a3c5385ba</guid>
            <category><![CDATA[innovation]]></category>
            <category><![CDATA[electric-vehicles]]></category>
            <category><![CDATA[transportation]]></category>
            <category><![CDATA[autonomous-cars]]></category>
            <dc:creator><![CDATA[Lior Swissa]]></dc:creator>
            <pubDate>Wed, 14 Feb 2024 07:43:25 GMT</pubDate>
            <atom:updated>2024-02-14T07:43:25.820Z</atom:updated>
            <content:encoded><![CDATA[<p>My 4-year-old son and I were watching Storybots when a question came up: How many parts make up a car? Care to take a guess?</p><p>30,000. That’s the magic number.</p><p>My aim here is to delve into the realm of innovation and its impact on the intricate web of components within our vehicles.</p><p>Picture this: instead of lifting the hood of a traditional car to be greeted by a cacophony of greasy pistons and gears, envision a sleek, silent electric motor and a dashboard adorned with sensors and screens. This isn’t a scene from a sci-fi flick; it’s the swiftly unfolding reality of transportation. As we transition from gas-guzzling behemoths to clean, autonomous pods, the very fabric of vehicle construction is undergoing a profound metamorphosis. Join me as we unravel this intriguing puzzle, exploring what stays, what goes, and the reasons behind it all.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jruHIa25XpQ6Y4aFyTGnWQ.jpeg" /></figure><p><strong>The ICE Age: Where Parts Run Rampant</strong></p><p>An Internal Combustion Engine (ICE) vehicle typically consists of around 30,000 components.Yes, you read that right: Thirty thousand intricate pieces work in harmony, like a finely tuned orchestra, each playing a vital role to propel you forward.</p><p>- The combustion engine: the pulsating heart of the machine, orchestrating a complex ballet of pistons, spark plugs, and valves to transform gasoline into raw power.<br>- The transmission: a symphony of gears seamlessly shifting to optimize performance and efficiency.<br>- The fuel system: a network of pumps, injectors, and filters delivering the lifeblood of the engine.<br>- The exhaust system: a labyrinth of pipes directing spent fumes away from passengers and the environment.</p><p>While this elaborate ballet ensures mobility, it’s not without its drawbacks:</p><p>- Complexity: Maintaining and repairing this legion of parts can be a costly and time-consuming affair. Picture the frustration of a single faulty spark plug throwing the entire orchestra out of harmony!<br>- Emissions: The combustion process churns out pollutants like carbon dioxide and nitrogen oxides, exacerbating climate change and air quality concerns.<br>- Noise: The roar of the engine can be both a nuisance and a health hazard, particularly in densely populated urban settings.</p><p><strong>The Electric Evolution: Streamlining and Simplification</strong></p><p>Enter the realm of electric vehicles (EVs), where we witness a remarkable simplification process. The electric motor, with its elegantly straightforward design, sheds approximately 50% of the parts found in traditional ICE vehicles. Gone are the convoluted fuel injection systems, labyrinthine exhaust manifolds, and intricate gearboxes. The result? A cleaner, quieter, and more efficient ride:</p><p>- Reduced Complexity: Fewer parts equate to fewer headaches and lower maintenance costs. Think of it as downsizing the orchestra — an intimate ensemble that’s easier to manage and keep in tune.<br>- Zero Emissions: Bid farewell to tailpipe emissions! EVs run on electricity, leaving the air cleaner and fresher for all.<br>- Smooth, Silent Rides: Electric motors hum softly, offering a serene and tranquil driving experience. Picture gliding down the road without the incessant drone of an engine.</p><p>But EVs aren’t just about shedding parts; they also introduce some essential new elements:</p><p>- The Battery Pack: The new power source, comprising lithium-ion cells that store the electricity fueling the motor.<br>- The Electric Motor: The silent conductor, converting electrical energy into mechanical power to propel the vehicle.<br>- The Charging System: An efficient and user-friendly interface designed to recharge the battery, ensuring seamless continuation of the symphony.</p><p><strong>The AV Ascent: Adding Intelligence, Redefining Mobility</strong></p><p>Taking the simplicity of EVs to the next level, autonomous vehicles (AVs) revolutionize the driving experience. While retaining many streamlined components of their electric counterparts, AVs introduce a whole new layer of technology:</p><p>- Sensors: Lidar, radar, and cameras act as the vehicle’s eyes and ears, providing real-time data to its AI brain. Imagine the orchestra with omniscient vision, aware of its surroundings at all times.<br>- Computing Power: High-performance processors analyze this data, making split-second driving decisions with unparalleled accuracy. Picture a virtuoso conductor leading the ensemble, ensuring every note is executed flawlessly.<br>- Connectivity: Communication systems enable AVs to interact with each other and the surrounding infrastructure, orchestrating a seamless dance of movement. Envision vehicles gliding in perfect harmony, sidestepping collisions and optimizing traffic flow.</p><p>While this infusion of technology poses its share of challenges, from ongoing development to integration with existing infrastructure, the potential rewards are staggering:</p><p>- Safety: AVs hold the promise of significantly reducing accidents caused by human error. Imagine a world where drunk driving, distracted driving, and speeding are relics of the past.<br>- Efficiency: Streamlined traffic flow and optimized route planning could spell the end of rush hour gridlock. Envision smoother commutes and reduced congestion becoming the new norm.<br>- Accessibility: Self-driving technology opens doors to mobility for those unable or unwilling to drive themselves. Picture a future where safe and convenient transportation is within reach for everyone.</p><p>In Conclusion…</p><p>As we navigate the shifting landscape of transportation, from pistons to pixels, one thing remains abundantly clear: the future of vehicles is poised for an exhilarating transformation. With each innovation, we inch closer to a cleaner, safer, and more accessible mobility ecosystem. So, buckle up and enjoy the ride — because the best is yet to come.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e92a3c5385ba" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Visionary Part of Automotive]]></title>
            <link>https://becominghuman.ai/the-visionary-part-of-automotive-a9b7726d3345?source=rss-1ae9eb713cd8------2</link>
            <guid isPermaLink="false">https://medium.com/p/a9b7726d3345</guid>
            <category><![CDATA[computer-vision]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[video-teleconferencing]]></category>
            <category><![CDATA[accident]]></category>
            <dc:creator><![CDATA[Lior Swissa]]></dc:creator>
            <pubDate>Tue, 10 Aug 2021 08:07:03 GMT</pubDate>
            <atom:updated>2021-08-15T11:55:56.291Z</atom:updated>
            <content:encoded><![CDATA[<p>When we talk about AV, we usually think about sensor fusion. However, today I want to talk about only one sensor, the camera, and the fantastic things we can benefit from (I’m a PM in the Video Telematics domain). Hey — Elon Musk / Telsa, up to very recently thought cameras made most other sensors (including Lidar) redundant. Let’s dive in and see why.</p><p>Cameras are the alternative to our eyes. The camera sees ahead, which helps us sense and understand the environment around us aided by Computer Vision and Artificial Intelligence. Moreover, what the camera allows on top of human capabilities is feedback on the human, as a driver. This means that a camera can track our behavior and alert us when we are in dangerous situations, and by doing so — prevent accidents.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FtADvFcj7Cm4%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DtADvFcj7Cm4&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FtADvFcj7Cm4%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/0e89a277959941e67e51b28ed53c4ba1/href">https://medium.com/media/0e89a277959941e67e51b28ed53c4ba1/href</a></iframe><p>There are two main types of cameras: A Dash camera and an outfacing camera (several cameras can obtain a 360-degree view). In this way, we can ‘see’ what is happening both inside and outside of the vehicle, and thus cover many cases. Not only ‘seeing’ but also ‘aware of’ — With AI, the camera can detect and alert about many types of risks. Some of those tools assist individual drivers, while others can assist managers of larger fleets — in both cases, these capabilities can save lives — either by immediate feedback, or by changing behaviors that are more likely to cause accidents.</p><figure><a href="https://aijobsboard.com/?fbclid=IwAR0GcsLs8A6CFM-fynrZMM3sTgo_Zpbto2CjglJm0Dmi6otC6YWc7CDFpQk"><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dAlwXfVrtYspdCzmC8iADw.png" /></a><figcaption>Big Data Jobs</figcaption></figure><p>In this blog, I want to give you a few examples to see how wide the use cases are.</p><p><strong>1 — Distracted driver:</strong> Anything that takes the drivers’ attention away from driving can be a distraction — sending a text message, talking on a cell phone, using a navigation system, and even eating. Any of these distractions can endanger you, your passengers, and others on the road.</p><p>There are three main types of distractions:</p><ul><li>Visual: taking your eyes off the road</li><li>Manual: taking your hands off the wheel</li><li>Cognitive: taking your mind off driving</li></ul><p>Distracted driving is dangerous, claiming 3,142 lives in 2019.(*)</p><p>Cabin cameras face the driver, and can alert when any of those behaviors happen.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*KfJbDTGlXswaOQHj.jpg" /></figure><p><strong>2 — Drowsy Driving:</strong> the dangerous combination of driving and sleepiness or fatigue. This usually happens when a driver has not slept enough, but it can also occur due to untreated sleep disorders, medications, drinking alcohol, and shift work.</p><p>Drowsy driving is a significant problem in the United States. The risk, danger, and sometimes tragic results of drowsy driving are alarming. Alerting about this risk is essential for drivers and fleet managers so that they can prevent collisions: NHTSA estimates that in 2017, 91,000 police-reported crashes involved drowsy drivers. These crashes led to an estimated 50,000 people being injured and nearly 800 (!) deaths.(**)</p><h3>Trending AI Articles:</h3><blockquote><a href="https://becominghuman.ai/why-corporate-ai-projects-fail-part-1-4-3b820041ab00">1. Why Corporate AI projects fail?</a></blockquote><blockquote><a href="https://becominghuman.ai/how-ai-will-power-the-next-wave-of-healthcare-innovation-695a2196aae8">2. How AI Will Power the Next Wave of Healthcare Innovation?</a></blockquote><blockquote><a href="https://becominghuman.ai/machine-learning-by-using-regression-model-f0c7993a66c8">3. Machine Learning by Using Regression Model</a></blockquote><blockquote><a href="https://becominghuman.ai/top-data-science-platforms-in-2021-other-than-kaggle-a1b4609c06b0">4. Top Data Science Platforms in 2021 Other than Kaggle</a></blockquote><p><strong>3 — Smoking</strong>: identify if a driver is smoking for safety purposes or regulations. Drivers of commercial vehicles (at least) are prohibited from smoking or carrying lighted smoking materials at any time while on or within 25 feet of such a vehicle.(***)</p><p><strong>4 — Traffic Signs Recognition: </strong>For ADAS (driver assistance) purposes, the camera can identify stop signs and red traffic lights and alert the driver if he is about to violate those rules (or just fail to notice a sign).</p><p>Running a red light is more likely to cause significant injury than other types of accidents. When you run a red light, you typically move through the intersection at a higher-than-average rate of speed. Unfortunately, so are drivers in the crossing traffic that has a green light. Additionally, red-light running accidents usually result in a “t-bone” collision in which one car hits one side of the other. These “angular” collisions often cause severe injuries and fatalities to the occupants of the car hit on its side. That’s because the fronts and rears of most cars have significant protections that absorb the force of an accident. There are no such “crumple zones” on the sides of cars. Mere inches separate the driver and passengers from the source of a side impact, increasing the likelihood they’ll suffer terrible harm.</p><p>Red-light running accidents cause heavy fatalities. In 2015, more than 200,000 injuries and 800 deaths per year were caused by this — two fatal accidents from running a red light per day across America. (****)</p><p><strong>5 — Identify pedestrians:</strong> Pedestrians in the vehicle path are in danger of being hit, thus causing severe injury to pedestrians and vehicle occupants. The camera can identify and alert in real-time about a potential collision — and prevent it.</p><p>In 2017, 5,977 pedestrians were killed in traffic crashes in the United States. That’s about one death every 88 minutes. Additionally, an estimated 137,000 pedestrians were treated in emergency departments for nonfatal crash-related injuries in 2017. (*****)</p><p><strong>6 — Face forward-collision:</strong> cameras can detect when a vehicle comes within seconds of a forward collision, keeping drivers and others on the road safe from potential accidents. In fact, in many <a href="https://www.youtube.com/watch?v=KHvLFujWBeg">videos</a>, we can see that the Tesla camera ‘sensed’ something wrong much before we as humans could — and saves the driver’s life.</p><p><strong>7 — Lane detection:</strong> Many people die each year in roadway departure crashes caused by driver inattention. Lane detection systems provide a safe pavement to avoid these accidents as safety is the main purpose of these systems. The camera detects the lane marks and will warn the driver if the vehicle tends to depart from the lane.</p><p><strong>8 — Insurance purposes and wrong violations:</strong> Have you ever been involved in an accident? Perhaps it was a hit-and-run in which you couldn’t find the vehicle or be caught by the police when you did nothing wrong? Now that the camera recording is available, you can get a video of the accident and provide solid proof to the police and insurance company. At my job, I see time and again that fleet managers consider this an essential capability, as their drivers are often innocent, yet the assumption is that the trucker is guilty.</p><p><strong>9 — Child safety:</strong> As a parent, I’m always afraid I’ll forget my son in the car on a hot summer day (or every day)! When the engine is turned off, the camera in the car can detect if there are still people inside the vehicle by using face recognition and alert the parent immediately.</p><p>According to the Beterem association, from 2008 until May 2016, 381 children left alone in a vehicle were reported in the media. Nineteen cases ended with the death of the child. Of the total number of children left in a vehicle, 10% were left in commercial passenger vehicles.(******)</p><p><strong>10 — Inspect the car conditions from the outside:</strong> Automatically flag exterior issues such as dents, scratches, broken, misaligned, or replaced panels in vehicles to onsite inspectors or remote body shops. An Israeli company that provides this solution and more is called <a href="https://www.ravin.ai/">Ravin.AI</a>.</p><p>The ability to track the condition of vehicles is essential to AV fleet managers since everything is already automated, and a scalable solution is needed.</p><p>All those capabilities come in handy for future mobility. Think of all the accidents we can prevent and how many lives we can save even when we only use the camera’s capabilities.</p><p>In my very first blog post, I wrote that as a PM, I first and foremost strive to save lives — and only then come fleet manager costs.</p><p>And how does this post relate to AV ? Well, the debate is still ongoing: are cameras enough or do we absolutely need a Lidar for AVs ?</p><p>Thank you for reading!</p><p>If you like this post, please clap for about 5 seconds :) so it can reach even more people.</p><p>I’d love to hear your thoughts and comments — please post below.</p><p>(*) <a href="https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/813060">https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/813060</a></p><p>(**) <a href="https://www.nhtsa.gov/risky-driving/drowsy-driving">https://www.nhtsa.gov/risky-driving/drowsy-driving</a>; <a href="https://www.cdc.gov/sleep/about_sleep/drowsy_driving.html">https://www.cdc.gov/sleep/about_sleep/drowsy_driving.html</a></p><p>(***) <a href="https://www.fmcsa.dot.gov/regulations/title49/section/397.13">https://www.fmcsa.dot.gov/regulations/title49/section/397.13</a></p><p>(****) <a href="https://www.grossandschuster.com/the-dangers-of-running-red-lights/">https://www.grossandschuster.com/the-dangers-of-running-red-lights/</a></p><p>(*****)<a href="https://www.cdc.gov/transportationsafety/pedestrian_safety/index.html">https://www.cdc.gov/transportationsafety/pedestrian_safety/index.html</a></p><p>(******)<a href="https://www.health.gov.il/English/Topics/KidsAndMatures/safety/car/Pages/car.aspx">https://www.health.gov.il/English/Topics/KidsAndMatures/safety/car/Pages/car.aspx</a></p><h3>Don’t forget to give us your 👏 !</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/780/0*2lvCls4yjxVMfZSR" /></figure><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fupscri.be%2F8f5f8b%3Fas_embed%3Dtrue&amp;dntp=1&amp;url=https%3A%2F%2Fupscri.be%2F8f5f8b&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=upscri" width="800" height="400" frameborder="0" scrolling="no"><a href="https://medium.com/media/c43026df6fee7cdb1aab8aaf916125ea/href">https://medium.com/media/c43026df6fee7cdb1aab8aaf916125ea/href</a></iframe><figure><a href="https://becominghuman.ai/artificial-intelligence-communities-c305f28e674c"><img alt="" src="https://cdn-images-1.medium.com/max/255/1*2f7OqE2AJK1KSrhkmD9ZMw.png" /></a></figure><figure><a href="https://upscri.be/8f5f8b"><img alt="" src="https://cdn-images-1.medium.com/max/255/1*v-PpfkSWHbvlWWamSVHHWg.png" /></a></figure><figure><a href="https://becominghuman.ai/write-for-us-48270209de63"><img alt="" src="https://cdn-images-1.medium.com/max/255/1*Wt2auqISiEAOZxJ-I7brDQ.png" /></a></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a9b7726d3345" width="1" height="1" alt=""><hr><p><a href="https://becominghuman.ai/the-visionary-part-of-automotive-a9b7726d3345">The Visionary Part of Automotive</a> was originally published in <a href="https://becominghuman.ai/">Becoming Human: Artificial Intelligence Magazine</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Flit towards AVs — how fleet management technology can help?]]></title>
            <link>https://becominghuman.ai/flit-towards-avs-how-fleet-management-technology-can-help-930c7ca4f480?source=rss-1ae9eb713cd8------2</link>
            <guid isPermaLink="false">https://medium.com/p/930c7ca4f480</guid>
            <category><![CDATA[self-driving-cars]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[autonomous-cars]]></category>
            <category><![CDATA[fleet-management]]></category>
            <dc:creator><![CDATA[Lior Swissa]]></dc:creator>
            <pubDate>Tue, 13 Oct 2020 06:04:43 GMT</pubDate>
            <atom:updated>2020-10-23T14:16:37.531Z</atom:updated>
            <content:encoded><![CDATA[<h3>Flit towards AVs — how fleet management technology can help?</h3><p>The autonomous vehicles’ race has been hot for a while. However, there is one field, less heard of, whose companies collect data that brings tremendous value and paves the way for Autonomous Vehicles — the fleet management technology industry.</p><p>I will focus on three key components already covered within fleet management, and explain how they contribute to the future of AVs.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/804/1*Ee_ftZOzIdkgggs-2j4YqA.gif" /></figure><p><strong>1 — Maintenance</strong></p><p>For Autonomous vehicles, remote maintenance tools are crucial to enhance performance and safety on the road when there’s no driver to watch over the car.</p><ul><li><strong>Refueling/Recharging </strong>(In the AV case — battery usage). For AVs, it’s essential to know how to plan the upcoming trips appropriately. The system knows, for example, that from Monday to Friday, you commute to the workplace, the same as human fleet managers know how to load balance their fleet based on different days of the week. Therefore, it can schedule battery charging according to actual and planned usage. Moreover, the system is equipped with GPS analytics and can identify map locations in which charging would occur at the right timing. As in current fleet management, this is all about maximizing the car’s utility to the owner, not necessarily maximizing road time.</li></ul><figure><a href="https://aijobsboard.com/?fbclid=IwAR0GcsLs8A6CFM-fynrZMM3sTgo_Zpbto2CjglJm0Dmi6otC6YWc7CDFpQk"><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dAlwXfVrtYspdCzmC8iADw.png" /></a><figcaption>Big Data Jobs</figcaption></figure><ul><li><strong>Route tracking: </strong>Any vehicle needs to plan which route to take based on various factors — some, as mentioned before, are related to maintenance, while others are related to road conditions. Fleet management tools can find vehicles and route history on a map with real-time GPS tracking. Point-by-point location information includes the date, time, altitude, speed, and any moments of risky driving behavior. This data matters because it helps us to forecast the road conditions and predict the next steps. For example, if most vehicles break instantly at a specific location, we can understand that there might be a hazard on that spot. This information can shed light on road structure/conditions and help optimize our fleet — either managing driving behavior in this specific road or trying to avoid this road overall.</li><li><strong>Kilometers remaining until scheduled service </strong>Whether it’s a personal AV or part of a fleet such as Uber, this aspect is important. The AV is a new kind of more complex vehicle that consumes much power to process and run algorithms. We should know how to maintain safety, and vehicle function at least as the way do now — which poses a challenge, having no driver who ‘feels’ the car is ‘a bit off.’ We can use the data collected from EV car fleets today to predict when is the next time the vehicle should get to the shop and ensure this does not have urgent business implications.</li></ul><h3>Trending AI Articles:</h3><blockquote><a href="https://becominghuman.ai/fundamentals-of-ai-ml-for-product-managers-64279e9d3681">1. Fundamentals of AI, ML and Deep Learning for Product Managers</a></blockquote><blockquote><a href="https://becominghuman.ai/the-unfortunate-power-of-deep-learning-3c06aef76e63">2. The Unfortunate Power of Deep Learning</a></blockquote><blockquote><a href="https://becominghuman.ai/graph-neural-network-for-3d-object-detection-in-a-point-cloud-309ba40897b2">3. Graph Neural Network for 3D Object Detection in a Point Cloud</a></blockquote><blockquote><a href="https://becominghuman.ai/know-the-biggest-notable-difference-between-ai-vs-machine-learning-cc1d7d18976a">4. Know the biggest Notable difference between AI vs. Machine Learning</a></blockquote><p><strong>2 — ADAS/Behavior</strong></p><p>ADAS (Advanced Driver Assistance System) smart solutions take us one step closer to fully autonomous driving, not only by advancing safety but also by capturing data that could be used for training. Many fleets are equipped with camera vision (even 360 degrees), allowing to capture footage of the road ahead and the vehicle’s cabin to detect risk wherever it occurs.</p><p>For this feature, I think it’s pretty obvious to have both a camera in front of the vehicle, one behind it, and one within the cabin itself to monitor driver behavior.</p><p>These cameras’ data is perfect for AV as it captures precisely human behavior and reaction to road conditions as we humans see it.</p><p>The computer vision data can be used to “teach” AV models how we humans drive. Moreover, I think this type of data is needed for AVs to imitate the exact human behavior: people don’t have lidars or thermal cameras, just our eyes, and we get pretty good results — there’s much to learn from both how we drive, and how we REACT to road conditions even before we made a driving decision.</p><p><strong>3 — AV fleet management</strong></p><p>When you manage a fleet of AVs, you have much more flexibility as human limitations do not apply. However, you also lose the independent judgment of human drivers. Thus you have to issue very precise instructions.</p><p>Today’s tools allow you to get instant access to live video from the road, manage risky driving, and review recordings to investigate incidents at any time. By doing so, this feature allows AV fleet management to scale their capabilities and monitor the vehicles to perform better, and eventually make an optimized decision for the fleet.</p><p>For example, we can optimize scheduling for real-time driving requests like Uber or food shipment or fleet management for trucks. We already have all statistics: their ride, their schedule, their performance etc — now, we can actively manage the AV fleet using this data to optimize the entire fleet’s results.</p><p>Conclusion –</p><p>Data has become the next currency and is valued more than ever (ask Google). Moreover, data is a big part of what we need to accelerate AV R&amp;D. Fleet management technology exists today and produces a ton of data — a perfect combo.</p><p>Thank you for reading!</p><p>If you like this post, please clap for about 5 seconds :) so it can reach even more people.</p><p>I’d love to hear your thoughts and comments — please post below.</p><h3>Don’t forget to give us your 👏 !</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/780/0*2lvCls4yjxVMfZSR" /></figure><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fupscri.be%2F8f5f8b%3Fas_embed%3Dtrue&amp;dntp=1&amp;url=https%3A%2F%2Fupscri.be%2F8f5f8b&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=upscri" width="800" height="400" frameborder="0" scrolling="no"><a href="https://medium.com/media/c43026df6fee7cdb1aab8aaf916125ea/href">https://medium.com/media/c43026df6fee7cdb1aab8aaf916125ea/href</a></iframe><figure><a href="https://becominghuman.ai/artificial-intelligence-communities-c305f28e674c"><img alt="" src="https://cdn-images-1.medium.com/max/255/1*2f7OqE2AJK1KSrhkmD9ZMw.png" /></a></figure><figure><a href="https://upscri.be/8f5f8b"><img alt="" src="https://cdn-images-1.medium.com/max/255/1*v-PpfkSWHbvlWWamSVHHWg.png" /></a></figure><figure><a href="https://becominghuman.ai/write-for-us-48270209de63"><img alt="" src="https://cdn-images-1.medium.com/max/255/1*Wt2auqISiEAOZxJ-I7brDQ.png" /></a></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=930c7ca4f480" width="1" height="1" alt=""><hr><p><a href="https://becominghuman.ai/flit-towards-avs-how-fleet-management-technology-can-help-930c7ca4f480">Flit towards AVs — how fleet management technology can help?</a> was originally published in <a href="https://becominghuman.ai/">Becoming Human: Artificial Intelligence Magazine</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Self-driving cars: Simulate or be too late]]></title>
            <link>https://medium.com/@liorswissa/self-driving-cars-simulate-or-be-too-late-81ab545bab97?source=rss-1ae9eb713cd8------2</link>
            <guid isPermaLink="false">https://medium.com/p/81ab545bab97</guid>
            <category><![CDATA[simulation]]></category>
            <category><![CDATA[google]]></category>
            <category><![CDATA[self-driving-cars]]></category>
            <category><![CDATA[nvidia]]></category>
            <category><![CDATA[autonomous-cars]]></category>
            <dc:creator><![CDATA[Lior Swissa]]></dc:creator>
            <pubDate>Tue, 27 Mar 2018 15:00:20 GMT</pubDate>
            <atom:updated>2018-03-27T15:02:36.155Z</atom:updated>
            <content:encoded><![CDATA[<p>Waymo, the pioneers of self-driving cars, reached five million self-driving miles last month, which is a remarkable achievement but not quite enough. For a self-driving car to drive the way a human drives (or better) we must accumulate an enormous number of miles, approximately 11 billion miles. Most importantly, those miles/data should have as many scenarios as possible to have high accuracy and allow good decision making by the car.</p><p>Surely 25 cities in the US (driven by Waymo) would not be enough to cover even just the roads in the US, let alone train a fully ‘anywhere’ self-driving car. Moreover, for every different scenario, we can have endless combinations to consider such as different landscapes (uphill or downhill),</p><p>type of road (Asphalt/dirt), is the road with lanes or without, weather conditions, driving culture (Minnesota nice or China crazy), daylight or nighttime, heavy traffic vs. few cars on the roads, dense urban areas vs. highways and the list goes on and on. The self-driving car should be ready for anything that may come — just as we would expect a human driver.</p><p>As we can understand, the same road segment is very different for the car to perceive, even for one minor change in the environment (such as sun behind you vs. sun in front of you).</p><p>Moreover, there are special cases we call ‘edge cases’ that are impossible to test in real life, for example, children jumping on the road or a car on the way that cuts you off — we can’t risk lives while training the car, thus we need another solution.</p><p>A niche in a market that emerges to solve some of those needs is an autonomous car simulation.</p><p>The advantages of a simulator are:</p><p>Time to market- the order of magnitude of miles Waymo drove so far would take us in the current pace more than a century to reach the goal. A simulator can run millions of simulations in parallel, consequently we can speed up the training process dramatically.</p><p>Cost- Every physically logged mile cost in the range of tens to hundreds of dollars. Simulations have no need for hardware, cars &amp; people (driver time, car cost, sensor’s high cost, fuel &amp; maintenance and so on) — thus are cheaper.</p><p>Geographic location — because we cannot be everywhere all the time a simulator can learn and drive on any given road in the world. In addition, different cities require different measurements such as SF foggy weather is different from the complicated roads in India.</p><p>Safe — training the AV system in a virtual environment is safer, and you will be able to try scenarios that would be prohibited or dangerous in real life. This means you can simulate accidents, traffic rules violations of other cars on the road, children jumping and many other cases that would be impossible to test otherwise.</p><p>A / B testing — in a simulator you can test the same conditions with a slight change to isolate a problem, or re-check the exact scenario you failed in the past to see if a bug was fixed.</p><p>Let’s go back to Google that has figured the above a long time ago and developed their own simulator system called “CarCraft”. So far, Waymo has been driving five million miles, but once they made a change to the system it’s not practical to drive all those miles again. That’s where CarCraft comes in: it replays all those miles again with the updated AV software to check if the outcomes improved. In this way, Waymo can drive ‘Way Mo’ miles (OK, I had to add that).</p><p>Waymo also has their own (physical) facility, an old air force base, that helps them to ‘feed’ the simulator and improve it with new and interesting scenarios. In this way, they can physically create a hazard that they won’t be able to safely create on a real road, such as a log on the road, and then input this into the simulator, which will then permute this into other scenarios. For example, roads in different locations (put a log on a specific problematic road).</p><p>Moreover, Waymo uses the actual driven miles as the base and then augment them into new situations by changing the parameters and test them against the AV system (for example, adding more cars, changing the speed of the car, etc.). By doing so, Waymo can scale and “drove” 2.5 Billion miles with their 25,000 virtual car fleet running 24/7 in their datacenter.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/974/1*s1RkxDPxeJ96iYagYZ_Nng.png" /></figure><p>(you can read more about it here: <a href="https://www.theatlantic.com/technology/archive/2017/08/inside-waymos-secret-testing-and-simulation-facilities/537648/">https://www.theatlantic.com/technology/archive/2017/08/inside-waymos-secret-testing-and-simulation-facilities/537648/</a> )</p><p>The second company that I want to mention is Nvidia, which has the natural capability to build such a simulator as they have 20+ years of experience in the gaming industry (on top of AVs being a core of their strategy).</p><p>In my opinion, the major difference between the two companies’ approach is that Waymo is using only their sensors (such as a Waymo specific Lidar) as appose to Nvidia that actually works with multiple partners and OEMs in the industry, and thus needs to be able to integrate much more hardware (i.e., 2 OEMs may use different Lidars).</p><p>I was completely blown away when I recently watched Nvidia’s CES keynote about their ‘AutoSim’ simulator. Mr. Mark Daly has presented their simulator and showed the audience in real time how the system works and how easy it is to create an environment for the virtual car or manipulate elements of such environment. More importantly, Nvidia not only changes the parameter of the surroundings but also the type and position of sensors on the vehicle.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/974/1*K5_rXWL7oS9rsq6MrPufzw.png" /></figure><p>You can watch a video about AutoSim here:</p><p><a href="https://www.youtube.com/watch?v=booEg6iGNyo">https://www.youtube.com/watch?v=booEg6iGNyo</a></p><p>Other than Waymo and Nvidia, there are naturally other companies in this sphere, some of them are specifically oriented to develop testing tools for the mobility world, such as IPG Automotive that exist for 30 years, and developed a simulator called “CarMaker”. Simulators are becoming a hot topic with more and more talent getting involved.</p><p>Forward-looking, it is clear that simulation will play a crucial role on the road to self-driving cars. There are multiple challenges in developing such systems. However, progress has been rapid and promising. Personally, I’m very happy to see the efforts companies put into those simulation tools as these will make AVs much safer in the future.</p><p>Thank you for reading!</p><p>If you like this post please clap for about 5 seconds :) so it can reach even more people.</p><p>I’d love to hear your thoughts and comments — please post below.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=81ab545bab97" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The self-driving car is on truck]]></title>
            <link>https://medium.com/@liorswissa/the-self-driving-car-is-on-truck-7ee8e2a3008a?source=rss-1ae9eb713cd8------2</link>
            <guid isPermaLink="false">https://medium.com/p/7ee8e2a3008a</guid>
            <category><![CDATA[automotive]]></category>
            <category><![CDATA[self-driving-cars]]></category>
            <category><![CDATA[trucking]]></category>
            <category><![CDATA[usa]]></category>
            <category><![CDATA[autonomous-cars]]></category>
            <dc:creator><![CDATA[Lior Swissa]]></dc:creator>
            <pubDate>Thu, 23 Nov 2017 16:28:21 GMT</pubDate>
            <atom:updated>2017-11-24T13:05:35.288Z</atom:updated>
            <content:encoded><![CDATA[<p>The truck industry correlates with our everyday life in every aspect, whether it’s the food we buy or the house we live in or just our shopping craziness to buy (everything) from Amazon. Trucks are always involved in the process. Therefore, the revenue of this delivery industry is about US$719 Billion!</p><p>Let’s dive deeper into the facts and the problems that we are having within this industry:</p><p>Human resources:<br>There are approximately 3.2 million truck drivers in the US. To be a truck driver, you’ll need a special license, which requires going through special training. The more I think about the truck driver, the more I realize that this occupation is very hard and demanding: The drivers are all by themselves in the car driving a fair amount of miles every single day, being far from family and friends. A complete isolation with no stimulation for the driver! Thus, to summarize, to be a truck driver is, to say the least not easy.</p><p>Time:<br>Time is of an essence — as time is money, especially in this industry where you can make more deliveries at this time. There are so many rules and regulations to protect the driver on the road, and thus limit the time he/she is driving to ensure he is refreshed and alert.<br>The hours-of-service regulations focus on when and how long truck drivers are allowed to drive by placing specific limits on the amount of time they drive the truck and how many total hours they can work before they are no longer permitted to drive a commercial motor vehicle. The drivers must follow three maximum duty limits at all times. They are the 14-hour “driving window” limit, 11-hour driving limit, and 60-hour/7-day and 70-hour/8-day duty limits:</p><p>• 14 hours window limit - You are allowed a period of 14 consecutive hours in which to drive up to 11 hours after being off duty for 10 or more consecutive hours. <br>Example: You have had ten continuous hours off, and you come to work at 6:00 a.m. you must not drive your truck after 8:00 p.m. that evening, which is 14 hours later. You may do other work after 8:00 p.m., but you cannot do any more driving until you have taken another ten consecutive hours off, or the equivalent of at least ten consecutive hours off duty.</p><p>• 11 hours limit - Once the driver has driven a total of 11 hours, he has to reach the driving limit and must be off duty for another ten consecutive hours (or equivalent) before driving the truck again. <br>Example: You have had ten consecutive hours off. You come to work at 6:00 a.m. and drive from 7:00 a.m. until 2:00 p.m. (7 hours driving). You take a 30-minute break as required, and then can drive for another 4 hours until 6:30 p.m. You must not drive again until you have at least ten consecutive hours off duty. You may do other work after 6:30 p.m., but you cannot do any more driving of a commercial motor vehicle on a public road.</p><p>• 60/70 hours limit - An addition to the limits mentioned above is the 60/70-hour limit. This limit is based on a 7 or 8-day period, starting at the time specified by the motor carrier for the start of a 24-hour period. This limit is sometimes thought of as a “weekly” limit. However, this limit is not based on a “set” week, such as Sunday through Saturday. The limit is based on a “rolling” or “floating” 7-day or 8-day period. The oldest day’s hours drop off at the end of each day when you calculate the total on-duty time for the past 7 or 8 days. For example, if you operate on a 70-hour/8-day schedule, the current day would be the newest day of your 8-day period and the hours you worked nine days ago would drop out of the calculation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/380/1*0njxVgIOhmZA6w1lZKZgvA.png" /></figure><p><a href="https://www.fmcsa.dot.gov/sites/fmcsa.dot.gov/files/docs/Drivers%20Guide%20to%20HOS%202015_508.pdf">https://www.fmcsa.dot.gov/sites/fmcsa.dot.gov/files/docs/Drivers%20Guide%20to%20HOS%202015_508.pdf</a></p><p>The best way to explain the insufficiency of the driver’s time is by thinking that most of the day the truck isn’t in use because of the driver limitation hours and most likely can’t be replaced by another driver as he is on the road in the middle of somewhere. Naturally, you can think of other ways to solve this, such as having two drivers — but those will be expensive and complex.</p><p>Accidents:<br>Most of the accidents involving trucks are caused by human error. Unfortunately, it takes the lives of a lot of people as we can see from the table below the number of casualties is growing.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ArDle8SfGGgWywN5Fvjelg.png" /></figure><p>A total of 3,852 people died in large truck crashes in 2015 in the US. Sixteen percent of these deaths were truck occupants, 69 percent were occupants of cars and other passenger vehicles, and 15 percent were pedestrians, bicyclists or motorcyclists. The number of people who died in large truck crashes was 22 percent higher in 2015 than in 2009, when it was lower than at any year since the collection of fatal crash data began in 1975. The number of truck occupants who died was 34 percent higher than in 2009. Since 1979, when deaths in large truck crashes were at an all-time high, there has been a greater percentage decline among occupants of large trucks (56 percent) than among occupants of passenger vehicles (37 percent).<br>However, Eleven percent of all motor vehicle crash deaths in 2015 occurred in large truck crashes. Thus this is still a very important area to improve for general road safety.</p><p>Where are these accidents happening? Fifty-three percent of deaths in large truck crashes in 2015 occurred on major roads other than interstates and freeways, 30 percent occurred on interstates and freeways, and 14 percent occurred on minor roads. This means that most accidents happen in areas that are easier for computers to control (as compared to small city roads with a lot of unrespectable things where humans are better), and thus are probably more likely to be caused by human error.</p><p>Source: <a href="http://www.iihs.org/iihs/topics/t/large-trucks/fatalityfacts/large-trucks">http://www.iihs.org/iihs/topics/t/large-trucks/fatalityfacts/large-trucks</a></p><p>Moreover, the cost of all of those accidents is very high and sums up to billions of dollars, on top of the thousands of lives taken every year.</p><p>By now we can read between the lines, and conclude that all of those factors are problematic to this significant industry.</p><p>We can look at it from two aspects:<br>1)Saving lives: More and more people are killed on the road because of drivers!</p><p>2)Money<br>*Pay salary for the truck drivers, who have a very demanding job.<br>*Insufficient use of the truck because of rules that apply to the drivers’ driving time.<br>*Cost of insurance and repairing of trucks (because of accidents).<br>*Logistic to manage the fleet of trucks and drivers.</p><p>It’s clear to see how a self-driving truck would solve all of those problems. Although trucks are not easy and creating a self-driving truck is still challenging, the good news is most of the driving trucks do are on large roads that are much more predictable and manageable for a computer (since most interstates, for example, ‘look and feel the same’ — vs. a street in SF, and a street in NYC that may be VERY different). Thus, self-driving trucks that solve the above problems might not only solve a super important need but also might be closer to market than the average self-driving car (as the urban landscape is much more important for a typical car ride than for trucks). These trucks might still need a human touch in the last mile, but this accounts for very little of the driving time of the truck.</p><p>A few companies foresee this opportunity and are working on self-driving trucks. <br>One of them is a Chinese company named “TuSimple”.<br>TuSimple is very much alike to a self-driving car with some changes to the sensor’s sensitivity and range because of the relatively slow reaction of trucks. AV truck sensors need to be more accurate and precise when the truck reacts to road changes, as it takes a truck (due to its size) much more time to react and break to a full stop, or maneuver to avoid an obstacle.</p><p>“Self-driving trucks could cut logistics costs by 40 percent in the U.S. and 25 percent in China as they can run longer than human-piloted rigs without rest and save at least 10 percent on fuel. They could also improve safety, especially in China, where trucks kill about 25,000 people a year, according to the Ministry of Public Security.” Said Chen Mo, co-founder and CEO of TuSimple to Bloomberg. I could not agree more with this critical need.</p><p>I can’t finish this article without mentioning “Tesla” which disrupted everything we know about the automotive industry. Now, with the Tesla electric Semi-truck, we can expect dramatic changes to electrified trucks in the future. This will save fuel, and enable better adoption of AVs. I think what Elon Musk is doing is beyond amazing and keep surprising us all. I can’t wait to see their semi-truck on the road.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/1*oJNmrrWnPjKpZHlolbhtAg.gif" /></figure><p>Thank you for reading!</p><p>If you like this post please clap for about 5 seconds :) so it can reach even more people.</p><p>I’d love to hear your thoughts and comments — please post below.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7ee8e2a3008a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Self-Driving cars are electrifying!!!]]></title>
            <link>https://medium.com/@liorswissa/self-driving-cars-are-electrifying-2f85a6ecc98e?source=rss-1ae9eb713cd8------2</link>
            <guid isPermaLink="false">https://medium.com/p/2f85a6ecc98e</guid>
            <category><![CDATA[mobility]]></category>
            <category><![CDATA[self-driving-cars]]></category>
            <category><![CDATA[automotive]]></category>
            <category><![CDATA[battery]]></category>
            <category><![CDATA[electric-car]]></category>
            <dc:creator><![CDATA[Lior Swissa]]></dc:creator>
            <pubDate>Mon, 25 Sep 2017 15:55:34 GMT</pubDate>
            <atom:updated>2020-08-06T07:51:13.336Z</atom:updated>
            <content:encoded><![CDATA[<iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Ftenor.com%2Fembed%2F3455033&amp;display_name=Tenor&amp;url=https%3A%2F%2Ftenor.com%2Fview%2Fgrease-john-travolta-sexy-funny-its-electrifying-gif-3455033&amp;image=https%3A%2F%2Fmedia.tenor.com%2Fimages%2Ff2ab5f7684f3bebbd49ac5b51ea11717%2Fraw&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=tenor" width="600" height="400" frameborder="0" scrolling="no"><a href="https://medium.com/media/5c29aa3506c641114e800d104f3ce884/href">https://medium.com/media/5c29aa3506c641114e800d104f3ce884/href</a></iframe><p>Recently I read a lot of media reports about electric vehicles: Most Automakers announced they would electrify their cars by ~2020–2025. Even OEMs that do not currently have an EV or a hybrid car seem to follow this trend. Another big direction I noticed is: most companies seem to partner with Chinese companies, or find other ways to break into the Chinese car market. China has already announced it will ban fossil fuel cars in the future, which will make China not only the largest car market in the world, but most likely also the largest EV market in the world.</p><p>In this post, I’d like to share a bit of the news about EVs, how those are related to self-driving cars, and also some cool technological developments that happened recently, all of which make this period of time super exciting for both EVs and AVs.</p><p>First, let me share a short recap of the announcements and the automakers plan to go electric:</p><ul><li><a href="https://medium.com/kylevogt/how-we-built-the-first-real-self-driving-car-really-bd17b0dbda55"><strong>GM and Cruise</strong></a> are ready to begin mass production of electric autonomous vehicles. What a bold (or is it Bolt?) move!</li><li><a href="http://money.cnn.com/2017/08/22/investing/ford-china-electric-cars/index.html"><strong>Ford</strong></a> announced their plan to partner with a Chinese company for electric vehicles, Anhui Zotye Automobile Co., to create a new brand for their electric vehicle to break into the largest market — China</li><li><a href="https://electrek.co/2017/01/03/ford-new-electric-cars"><strong>Ford</strong></a><strong> </strong>announced that by 2020 they would have fully EV models following their announcement back in 2015 to invest $4.5B in electrification.</li><li><a href="https://electrek.co/2017/09/11/vw-massive-billion-investment-in-electric-cars-and-batteries/"><strong>Volkswagen</strong></a> finalized a joint venture with Anhui Jianghuai Automobile (JAC Motor) to produce electric vehicles. Volkswagen also announced this month a massive investment of $84B in EV and batteries to bring 300 EV models by 2030 and $60B would be to battery production (This is a significant move by the company that had a major issue with their diesel scandal).</li><li><a href="https://www.wired.com/story/jaguar-land-rover-electric-e-type/"><strong>Jaguar Land Rover</strong></a> announced that by 2020, it would offer an electric or hybrid variant of every model it makes.</li><li>Last year <a href="http://www.businessinsider.com/bmw-launching-electric-cars-by-2020-2016-10"><strong>BMW</strong></a> announced they will produce at least 2 EV models by 2020 and plan to go head to head with their competitors for the luxurious models with a longer range. By 2025, <a href="https://www.cnbc.com/2017/09/07/bmw-readies-mass-production-of-electric-cars-12-models-by-2025.html">BMW</a> will offer 25 electrified vehicles of which 12 will be fully-electric.</li><li><a href="https://www.theverge.com/2017/9/22/16350506/daimler-mercedes-benz-eq-1-billion-expansion"><strong>Daimler</strong></a> plans to electrify all of its cars by 2022 and will spend $1B to build EVs in their US factory.</li><li><a href="http://www.businessinsider.com/honda-to-launch-2-electric-cars-by-2018-2017-8"><strong>Honda</strong></a><strong> </strong>is also entering the game and launching two electric cars by 2018. At first, it would be specifically designed for the Chinese market and later on for the European market. Honda is aiming for two-thirds of its vehicles to be electrified by 2030.</li><li>A <a href="https://electrek.co/2017/09/15/renault-nissan-mitsubishi-alliance-12-new-all-electric-vehicles/"><strong>Renault, Nissan &amp; Mitsubishi</strong></a> alliance will accelerate collaboration on common platforms, powertrains and next-generation electric, autonomous and connected technologies.</li><li><a href="https://www.nytimes.com/2017/07/05/business/energy-environment/volvo-hybrid-electric-car.html?mcubz=3"><strong>Volvo</strong></a> would introduce five models from 2019 to 2021 that would run solely on electric power. Volvo twitted that they are committed to electrification.</li></ul><p>Those announcements were not only about electric vehicles but also had a lot to do with autonomous ones. So I thought “why those things seem to go hand in hand?”. Here are some explanations I could think of:</p><p><strong>Computers run on electric power:</strong></p><p>It’s much more natural for computers to work with batteries (Electric vehicles) than with combustion engines and there is no need to design something new for the AV if it runs on pure electricity. However, autonomous vehicles will need more electric power to manage all the processes like mapping, computer vision, communication and the volume of data processing in real time. To deal better with the computation challenges of Autonomous vehicle we need to have a very strong power source-batteries!</p><p><strong>More miles:</strong></p><p>Nowadays the usage of a typical car is ~5%, and it lasts, let’s say ~10 years. Autonomous vehicle will change the whole mobility concept and use the car much more, thus if a car uses a combustion engine, it might need to be replaced on an annual basis or even more. Electric cars last longer based on current research. This is since electric motors are just more simple and have less ways to “go bad” — no oil leaks, less mechanical system issues etc. Therefore, electric vehicles are the perfect solution for self-driving cars: while you still may need to replace the car every 5–10 years, now you would get an order of magnitude more miles on that car.</p><p><strong>Electric charge:</strong></p><p>It’s easier to charge a battery than to develop ways to fuel a self-driving car.</p><p>Moreover, because self-driving cars would be on the road most of the time, they might also charge themselves with solar energy — getting even more miles per day.</p><p><strong>Future legislation and regulations:</strong></p><p>China, the biggest market for automakers, will ban fossil fuel in the near future to push forward electric vehicles on the road. The air pollution issue in China is very severe, and I think this is a major step toward cleaner air and better quality of life for people in China. Once this shift takes place, other countries will follow as we see France and the UK are also aiming at Electrification and banning gasoline and diesel.</p><p><a href="https://www.bloomberg.com/news/articles/2017-09-10/china-s-fossil-fuel-deadline-shifts-focus-to-electric-car-race-j7fktx9z">https://www.bloomberg.com/news/articles/2017-09-10/china-s-fossil-fuel-deadline-shifts-focus-to-electric-car-race-j7fktx9z</a></p><p>Hence seems like the conclusion is that self-driving cars, as the future of the car industry, will have to be electric.</p><p>As we see, the shift toward electric vehicles is happening at lightning speed, this is a huge potential market for battery companies!</p><p>Battery technology hasn’t changed for a while and still works on lithium-ion components (similar to your laptop or mobile phone). So…what’s happening in the EV scene recently and who is working to reinvent the wheel (Or shall I say the batteries)?</p><p>There are multiple companies that are currently working on technologies that might be a breakthrough for EV batteries:</p><p>The first one is <a href="https://www.store-dot.com/business-units"><strong><em>StoreDot</em></strong></a><strong><em>,</em></strong> an Israeli startup that has recently raised $60M in funding from Daimler. StoreDot is developing a new type of electric-car battery consisting of proprietary organic compounds. The EV FlashBattery enables full charge in 5 minutes (!), providing up to 300 miles (480 km) of driving distance, depending on the model of the EV. As we said, with AVs as the car needs to be on the road more, reducing charging time is crucial. Moreover, EV FlashBattery architecture is built with a highly stable electrode structure, and contains materials that are far less flammable and more stable at high temperatures than traditional Li-ion technology (No Samsung incidents). In addition, FlashBattery is environmentally safer than a Li-ion battery, utilizing organic compounds and a friendlier aquatic based manufacturing process.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/970/1*ZmiAj-GrtETxS2eLSfhN5A.png" /></figure><p>The second company is <a href="https://www.qualcomm.com/solutions/automotive/wevc"><strong>Qualcomm’s Halo system</strong></a>, a wireless charging system for electric cars. Their vision is to create a dynamic electric vehicle charging (DEVC) that allows to charge while driving (again, going back to AVs being on the road most of the time). This technology is at a very early stage, and there are 25 partner companies across Europe involved in the tests, which should conclude by the end of the year.</p><p>This technology is a perfect solution for self-driving cars also because it’s seamless and does not require developing ways for the AV to ‘know’ how to recharge itself.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/1*Hyk9TUKiU4_lJAVelJUVtg.gif" /></figure><p>As I already mentioned above, the automakers invest a lot of money in electric vehicles and batteries, but they are not the only ones to understand the potential of this market. The top three South Korean battery makers: <a href="http://m.yna.co.kr/mob2/en/contents_en.jsp?cid=AEN20170908006300320&amp;site=0500000000&amp;mobile">LG Chem Ltd., Samsung SDI Co., and SK Innovation Co.</a> plan to invest US$2.3 billion by 2020 to upgrade battery technology for electric vehicles and expand their local production facilities.</p><p>The implications of EV on quality of life would be not only on the economical aspect, but also on the environmental side. As I currently live in Shanghai, I think the air pollution will decrease significantly, and I could finally breath fresh air and be able to see blue skies.</p><p>So for me (and for EVs and AVs), the future seems brighter than ever :)</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2f85a6ecc98e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How will Israeli innovation contribute to the future of mobility?]]></title>
            <link>https://medium.com/@liorswissa/how-will-israeli-innovation-contribute-to-the-future-of-mobility-ec3f6f8762db?source=rss-1ae9eb713cd8------2</link>
            <guid isPermaLink="false">https://medium.com/p/ec3f6f8762db</guid>
            <category><![CDATA[self-driving-cars]]></category>
            <category><![CDATA[venture-capital]]></category>
            <category><![CDATA[automotive]]></category>
            <category><![CDATA[israel]]></category>
            <category><![CDATA[autonomous-cars]]></category>
            <dc:creator><![CDATA[Lior Swissa]]></dc:creator>
            <pubDate>Wed, 30 Aug 2017 13:48:35 GMT</pubDate>
            <atom:updated>2017-08-30T13:48:35.491Z</atom:updated>
            <content:encoded><![CDATA[<p>While Israel is already known as the “start-up nation”, was all that high-tech translated into the world of mobility? Is Israel also a “self-driving cars” nation?</p><p>To dig deeper into that question, I interviewed the founder of the sole Israeli VC focusing exclusively on Mobility — “<a href="http://www.maniv.com/">Maniv Mobility</a>” (Maniv in Hebrew means to bear fruit/produce): Michael Granoff, and his partner, Olaf Sakkers. (Interviews conducted during August 2017)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/768/0*2mybZ8-f2fXvJNW2." /></figure><p>First, let’s review the evolution of autonomous Vehicles in Israel.</p><p>Autonomous vehicles and mobility in general give rise to multiple critical categories:</p><ol><li><strong>Localization and hail transportation/ride</strong>: find the location of the car, passenger and the route it needs to drive within a map of the near environment.</li><li><strong>Communication/connectivity and analytics</strong>: transmit and process the data that has been collected by the sensors to make a driving decision within a fraction of a second, while notifying other vehicles of your location.</li><li><strong>Sensors and Computer vision:</strong> The eyes on the road. One of the critical components of the car is to capture the surrounding and make a decision based on all the data gathered through the sensors (Camera, Lidar, radar, ultrasonic, etc).</li><li><strong>Cyber security</strong>: Connectivity makes the car vulnerable to cyber security attacks. To protect the car from hackers, the car needs safer protocols and sophisticated software so that no one will have control of the car but the car itself.</li></ol><p>(Cyber is the most important topic in my opinion, and you can read more about it on one of my previous posts: <a href="https://medium.com/@liorswissa/hacking-cars-cyber-security-for-autonomous-vehicles-are-we-there-yet-d43ac350de41">https://medium.com/@liorswissa/hacking-cars-cyber-security-for-autonomous-vehicles-are-we-there-yet-d43ac350de41</a> )</p><p>All the technologies mentioned above might be new to the automotive industry to build self-driving cars, but NOT AT ALL to Israel. Israel has proven itself as a major player in all of the areas above. Elbit and Rafael are two veteran companies in the areas of sensors and computer vision (mainly developing weapon systems). Checkpoint is one of the world’s leading cyber security companies. Mellanox is a leading communications company. They are all leading in their fields and are all based in Israel, employing (and giving top-notch training to) thousands of engineers, some of whom are now thinking about how to apply their knowledge to self-driving cars. In my view, those companies’ expertise are important building blocks of the new autonomous vehicle innovations we see (and will see) coming out of Israel.</p><p>Big car companies such as General Motors, Daimler Group (BMW), Volvo, Honda and China’s SAIC Motors, already recognized the talent that Israel has and opened local R&amp;D centers.</p><p>Moreover, recently (August 2017) Renault-Nissan announced that it would open a smart-car incubator in Israel. Even Intel joined the race for AV and acquired Mobileye, an Israeli technology company that develops vision-based advanced driver-assistance systems providing warnings for collision prevention and mitigation. Nvidia, one of the leading companies in self-driving cars also opened an R&amp;D center in Israel and for the first time will have their GTC conference in Israel, this coming October.</p><p>It’s overwhelming to see how many innovations and companies in the area of self-driving cars rise from Israel in the last few years (see picture below for examples of startups in the field).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/930/0*Qoo_9o08KtAkKAGO." /></figure><p>Now that we have a better understanding of Israel’s self-driving cars sphere, I’d like to share the insights I learned from Michael and Olaf as experts in this field.</p><p><strong>Q: Why do you think Israel is a hub for self-driving cars and why did you choose to invest there? What are Israel’s competitive advantages?</strong></p><p>Michael: “Disciplines that have become essential to the future of automotive and mobility more broadly are some of the disciplines that Israel had in the past like machine vision and machine learning, a lot of sensors, software of cyber security and so forth. In our fund a lot of companies are not coming from first-time founders but from seasoned entrepreneurs (with) their second, third or fourth startup, with exits in other areas, some of which overlap to what they are doing.”</p><p>Olaf: “Israel is a small country, and you can deploy AVs at very high numbers which creates some network effects to allow for second-order benefits for self-driving cars. At the same time, Israel is a small and innovative country with a really strong tech ecosystem. Outside of the US, Israel is probably the densest tech ecosystem, so those two things can combine with the deployment of autonomous vehicles and a local kind of development community.”</p><p><strong>Q: China is the biggest car market in the world. Are there partnerships or interest from Chinese investors regarding Israeli technology?</strong></p><p>Michael: “There has been a lot of competition in the last major automotive technology around electrification between China and the US, and Israeli technology is in the middle of that. Now I think we’re seeing the same kind of dynamic and I think that China is going to play a major role in the future of automotive technology and Israel-China axis is going to be very common on that. I think Israel has a lot to offer to China.”</p><p><strong>Q: Culture, driving behaviors, rules and conventions in different countries such as Israel, US and China are very different. How do you think Israeli companies approach that and do you think Israel has a disadvantage being far from the market? For example in your portfolio you have “</strong><a href="http://www.cognata.com/"><strong>Congnata</strong></a><strong>” doing simulation. How those differences impact a company like that?</strong></p><p>Olaf: “I think it’s quite obvious to Israelis that there are different styles of driving and that self-driving cars will have to deal with a variety of driving conditions to reach scale. Cognata is a good example because the system allows that scaling. It’s designed to scale many different cities and test them in simulation and then take a lot of different data sets of driving behaviors and use all of that to allow for self-driving car developers to scale their systems globally. Right now all self-driving cars are being tested in a very narrow set of applications. Even Waymo which has done the most significant range of testing, are only testing in four different US cities. That’s still a very limited range of driving profiles and behaviors.</p><p><strong>Q: Israel is already a cyber-hub. Will it also be a car cyber hub?</strong></p><p>Michael: “Yes, absolutely. We have looked at 15 automotive cyber companies in Israel in the last few years. We invested in one called ‘<a href="https://www.upstreamsecurity.com/">Upstream security</a>’ that does security on the cloud.</p><p>In this field, there’s not going to be a silver bullet. There are going to be a bunch of different solutions, but we’re being very conservative in our approach to this, because it’s going to be very dynamic and hard to figure out. Cyber is definitely a critical topic: even a simple thing such as putting masking tape on stop signs in various configurations can confuse the sensors. There are many ways of malicious behaviors and you need to protect around that: in the code or the infrastructure or something else in the system. It is going to be a big business to keep these things safe.</p><p><strong>Q: Why did you decide to invest in ‘Upstream Security’?</strong></p><p>Michael: “(They have a) very specific and dedicated approach to helping fleets protect themselves using the cloud and communications between the fleet vehicles and the cloud that coordinates them. The team is terrific: has done analogous situations in enterprise security. Also, co-investors are “Glilot”, who are one of the leading investors in Israel focused on cyber security.”</p><p><strong>Q: What do you think about cyber security regarding to rules and regulations of different countries?</strong></p><p>Michael: “It’s very early. There’s a bill in the US Congress which speaks about the need for cyber. I think it is likely that the global regulation will be reactive and it’s hard to predict what it would be reacting to. But I think even more important than that, the public is not going to feel comfortable about automated vehicles unless they have a sense of security and a sense of comfort that they are protected from bad actors.</p><p>It’s a complex topic. There are a lot of responsibilities on the manufacturing level and at the service provider level and a lot of room for smart, creative innovators to get involved in filling all those needs. I’m confident that it’s not going to prevent this revolution from happening, but still (regulation and cybersecurity) is going to be a very big topic.”</p><p><strong>Q: How do you think regulation will influence where R&amp;D will be done?</strong></p><p>Michael: “We’re seeing a bit of competition within the US on this topic. A couple of states such as Nevada and Florida are in the forefront of creating regulatory regimes to encourage development to come to their states. Globally, the UK has been aggressive as has Singapore. I think it’s a big opportunity for Israel and this is something that the government starts to pay attention and recognize. The US bill I referred to would create some harmonization of regulation across the US so you would not have a situation that a car could not cross across the states line because of different regulations.”</p><p><strong>Q: Intel CEO, Brian Krzanich, points out in a recent article that each autonomous vehicle will be generating ~4TB of data a day. Do you think the current computing HW, storage and software are ready for this load? If not, where will the most significant improvements be in the future: in the computing HW (semiconductor industry)? In cloud storage and communication speed? In getting smarter algorithms that do not require collecting that much data and can pinpoint more on the relevant data?</strong></p><p>Michael: “The simple answer is all the above. There is no end to the volume of the data that will be collected with all the sensors and all of the software. We are seeing a beginning of this already, with companies such as “<a href="https://www.getnexar.com/">Nexar</a>”, from our portfolio, that is able to figure out what of the data is significant and what is unusual and should be preserved, and transmitted to the cloud — just the relevant data and not the whole bulk of it.</p><p>“<a href="http://otonomo.io/">Otonomo</a>” in our portfolio is another company that monetizes this data in the cloud that is another element to it. All of these things are coming to play, and this is the area that might be the most interesting right now in terms of opportunities around this business because this data is beginning to come in huge volumes”.</p><p><strong>Q: Daimler’s head of Group Research &amp; Mercedes-Benz Cars Development, Ola Kaellenius, said he thinks self-driving taxis are the only way forward, because privately owned self-driving cars would be too expensive, at least in the beginning. What do you think on this topic?</strong></p><p>Olaf: “The cost of shared mobility services is relatively lower than the cost of owning a vehicle. Autonomous might be more expensive initially and shared autonomous vehicles will be significantly cheaper. Consumers are cost conscious, I think, and when given a choice between an Uber or a slightly cheaper Lyft they will choose a slightly cheaper Lyft. It is hard to compare the decision of buying a car and using a car on an on-demand basis, but if the on-demand basis is significantly cheaper a lot of people will shift.</p><p>Finally, I think there will be other benefits for autonomous vehicles that will push people towards them, such as you don’t need to park with autonomous vehicle nor have to spend any energy driving the vehicle, so you’ll be able to do other things. The downside, compared to owning your own car, is sharing but I think overall, the various benefits will shift a lot of people toward the autonomous vehicle solution once the technology is ready.”</p><p><strong>Q: What do you think is the role of startups in self-driving cars vs the role of the big companies? Is there things you think big companies can only do, and other things only startups will be doing?</strong></p><p>Michael: “The question is whether the car manufacturers are well positioned to survive or to thrive in the new environment, whether it’s the technology companies such as Google or Baidu or whether it is startups or Uber and other ride hailing companies. I think they all have their advantages and disadvantages.</p><p>It’s proven that it’s not easy to scale on manufacturing and therefore even if we have fewer vehicles that have higher utilization rates, the factories and the OEMs are not going to go away. There might be consolidation, and there is still an open question as to whether they will continue to be brands. In a world in which you have mobility as a service on demand (there’s still a question if) people pay more or less attention to the type of vehicle and whether the features that matter have to do with traditional OEM manufacturing or have to do with the digital layer added on by a technology company. Those are all open questions and the automotive industry is recognizing that this disruption is happening and tries to respond to it.</p><p>One of the reasons we invest in early stage startups is because we find it easier to figure out what technologies are going to be relevant to the future, than figure out who is going to be the integrator and who is going to be the supplier to the end user. We are going to help our companies try to continue developing technologies and get them into the selection.”</p><p><strong>Q: By now AV is a sort of a hype. Do you think Israeli startups have come to their peak or it’s just the beginning?</strong></p><p>Olaf: “It’s always hard to tell when you are in a bubble mentality or when companies are legitimately interesting and get lots of investment because of real value. The fundamentals here are extremely good: there are a billion cars in the world, and the utilization of these assets is on average about 5%. Imagine you’ve got a factory that’s generating a billion different parts and has a trillion dollars’ worth of capital investment, but it’s only operating at about 5%. Anybody with sense would want to try and find ways to improve efficiency and utilization. That’s on top of the fact that more and more people die every year in car accidents, and billions of hours are lost in productivity just through waiting in traffic jams.</p><p>The danger of the hype is around timing. People believe this technology is about to be deployed on a mass scale and that’s just not the case. The technology is very difficult technically: there are still a lot of scaling issues related to fundamental technology. There could be a correction to the hype because people are overestimating the readiness of this technology, but the fundamentals are very strong, and it’s a great time for us to invest, especially in the seed stage, which is less crowded and we as a fund can add a lot of value.</p><p><strong>Q: When do you think we will see the first self-driving car?</strong></p><p>Olaf: “The problem state of the entire world is too broad and complex for computer systems to deal with. All the exceptions create a very long tail of edge cases, so you have to narrow the problem. One way in which you can narrow the problem space is to geo fence the area to which autonomous vehicles can be deployed, and the other way is by limiting the functionality of autonomous systems: that’s why we have <a href="https://www.sae.org/misc/pdfs/automated_driving.pdf">L3 (level 3) and L4 (level4)</a>. The first case is “everything somewhere”, so the autonomous vehicle can act autonomously but in a limited geographic area and the second case is “something everywhere”, so the vehicle can do something autonomously everywhere but there’s no full autonomy. Those things are already happening and will continue to happen increasingly and incrementally.</p><p>If I want to put my money on a number, I would say that in 2025 I would expect (fully) autonomous vehicle services to be commercially available and maybe sooner.”</p><p>Michal: “We already see Waymo operating in Arizona also GM in SF. The question is when will it become mainstream. I think the key measure as to when this revolution will have really occurred is when the majority of miles traveled are done in vehicles that are computer driven and I think that will happen in 2030, that’s my guess. How fast this will happen is not clear, but it’s really only 12–13 years, which is anyway pretty soon.”</p><p>Thank you for reading! If you like this post please clap for about 5 seconds :) so it can reach even more people.</p><p>I’d Love to hear your thoughts and comments — please post below</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ec3f6f8762db" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Nvidia and the GPU: contribution to the AI world of self-driving cars]]></title>
            <link>https://becominghuman.ai/nvidia-and-the-gpu-contribution-to-the-ai-world-of-self-driving-cars-1f00e3212508?source=rss-1ae9eb713cd8------2</link>
            <guid isPermaLink="false">https://medium.com/p/1f00e3212508</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[autonomous-cars]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[self-driving-cars]]></category>
            <dc:creator><![CDATA[Lior Swissa]]></dc:creator>
            <pubDate>Tue, 18 Jul 2017 13:31:53 GMT</pubDate>
            <atom:updated>2017-07-24T15:08:14.947Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/720/0*7dReqQXElneHBWUr.jpg" /><figcaption>Source</figcaption></figure><p>The first time I watched the DARPA challenge for self-driving cars I thought this was a breakthrough I want to be involved in. But to my surprise, the challenge was in 2005 — so long ago and yet nothing came to the market for a long time. Moreover, Artificial Neural Networks — the algorithms to process all this data was developed in the 1960’s (!!).</p><p>So what happened all of a sudden and what has changed to make self-driving cars (and AI) go forward? The answer is Nvidia happened, aka GPU chip designs (Graphics Processing Units) — even if at the time the big breakthrough happened, Nvidia was not aware they made this huge contribution.<a href="http://www.zipcar.com/ziptopia/future-city/self-driving-cars-are-really-happening">http://www.zipcar.com/ziptopia/future-city/self-driving-cars-are-really-happening</a></p><p><a href="https://www.eventbrite.com/e/the-chatbot-conference-tickets-34868758395?aff=CBL">The Chatbot Conference</a></p><p>A little background on Nvidia GPU, before I further explain. The term GPU was popularized by Nvidia in 1999, who marketed the GeForce 256 as “the world’s first GPU”, or Graphics Processing Unit. It was presented as a “single-chip processor with integrated transform, lighting, triangle setup/clipping, and rendering engines. Nvidia was first to produce a chip capable of programmable shading, the GeForce 3. Each pixel could now be processed by a short “program” that could include additional image textures as inputs, and each geometric vertex could likewise be processed by a short program before it was projected onto the screen (<em>Source: Wikipedia</em>). But the GPU was mainly meant for gaming, and other graphic implementations.</p><p>The major breakthrough from a CPU to GPU happened in 2013 when Andrew Ng from Stanford realized his team could run NN in parallel on a GPU (rather than sequential on a CPU).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/863/0*3XosKGvx5F3t--f4." /></figure><p>So why is GPU good for AI and DL (deep learning)?</p><p>To understand this better I will explain a little bit about how a NN (neural network) works.</p><p>NN is a collection of simple trainable mathematical units that collectively learn complex functions. Given sufficient data, the NN can approximate very complex functions mapping raw data to output decisions.</p><p>Neural networks are structured as an algebraic unit — an artificial neuron. It has three layers: The input layer, the data we pass to the Network. The hidden layer where all the computations are made and the output layer which is where the prediction classification or the output decision is made. We Feedforward data into the NN and compare the prediction with the ground truth (label we have). If the prediction is incorrect we need to refine the model so that the error rate would be smaller and thus the performance would be better. Hence, we train the NN with a lot of data so that the average performance across all the data will be satisfactory to deploy this model with a lower error rate and a high accuracy.</p><p>Neural networks’ operations are: Convolution, Pooling, Activation function and Backpropagation — exactly what a GPU is good at!</p><p>I read an analogy (from multiple sources online) that explains this beautifully :</p><p>Assume you want to move a house and are required to transfer your belongings from your current house to the new house. You have two choices: a Ferrari or a truck. While a Ferrari would be faster per trip it can only contain a small amount of your things and requires a lot of power. More than that, you have to go back and forth multiple times with the Ferrari, while if you take a truck you can take all of you belongings in a fair cost utilizing one trip, and have access to the things faster as they are already there (on the truck).</p><p>In other words, GPU delivers better prediction accuracy, faster results, smaller footprint, lower power and lower costs. On a single computation, a CPU might be better, but for the mass repetitive calculations AI needs, a GPU is WAY better, and faster.</p><p>As the creator of the GPU, Nvidia found themselves with a goldmine at hand — a component that stands at the core of AI. How did Nvidia further enhance their GPUs to work better for NN and their solution?</p><p>What is fascinating about Nvidia is that it has a full stack solution architecture for DL applications, making it easier and faster for data scientist engineers to deploy their programs.</p><p>It’s amazing that Nvidia (remember, originally only a GPU manufacturer) came up with this brilliant ecosystem! They tailored the software side to run better on their GPU to enhance the performances of the GPU for NN.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*WROgaOU35HN_r16v." /></figure><p>In more depth, the system, Nvidia’s Cuda, is a parallel computing platform and programming model. It enables dramatic increases in computing performance by harnessing the power of the graphics processing unit (GPU). Using high-level languages, GPU-accelerated applications run the sequential part of their workload on the CPU — which is optimized for single-threaded performance — while accelerating parallel processing on the GPU. This is called “GPU computing.”</p><p>How did Nvidia implement all this end-to-end solution in self-driving cars?</p><p>As we will now see all these layers mentioned above, adds up to the “End-to-End Learning for Self-Driving Cars” system Nvidia has built for autonomous vehicles.</p><ol><li>Nvidia NN system</li></ol><p>As part of a complete software stack for autonomous driving, NVIDIA created a neural-network-based system, known as PilotNet, which outputs steering angles given images of the road ahead. PilotNet is trained using road images paired with the steering angles generated by a human driving a data-collection car. It derives the necessary domain knowledge by observing human drivers. This eliminates the need for human engineers to anticipate what is important in an image and foresee all the necessary rules for safe driving. Road tests demonstrated that PilotNet can successfully perform lane keeping in a wide variety of driving conditions, regardless of whether lane markings are present or not.</p><p>In addition to learning the obvious features such as lane markings, edges of roads, and other cars, PilotNet learns more subtle features that would be hard to anticipate and program by engineers, for example, bushes lining the edge of the road and atypical vehicle classes (<em>Source :Cornell university CS department</em>).</p><p>2. Nvidia DRIVE PX</p><p>This part is after we already trained the NN. DrivePx is the self-driving car computer within the car, which processes all the data (images) it gets from the environment (that the sensors provide) of the car and delivers the output — driving decisions.</p><p>DRIVE PX can understand in real-time what’s happening around the vehicle, precisely locate itself on an HD map, and plan a safe path forward. It’s the world’s most advanced self-driving car platform — combining deep learning, sensor fusion (multiple types of sensors), and surround vision to change the driving experience.</p><p>3. Nvidia DriveWorks</p><p>NVIDIA DriveWorks is a Software Development Kit (SDK) that contains reference applications, tools, and library modules as part of the structure layers mentioned above. It also includes a run-time pipeline framework that integrates every aspect of the driving pipeline, from detection to mapping and localization to path planning to visualization.</p><p>4. Nvidia HD Mapping</p><p>MapWorks is built in compliance with the architecture of the Nvidia DriveWorks for autonomous driving. Both, MapWorks and DriveWorks are designed for carmakers, map companies and startups to accelerate development. When just using cameras, the full system incorporates deep learning algorithms to detect lanes, signs and other landmarks, and can be used to both create maps and determine when the environment has changed.</p><p>For localization, structure-from-motion algorithms enable data from multiple cameras to be converted into detailed 3D mapping information. Combining data from various inertial sensors in the car, along with GPS data and cameras, enables precise positioning of key landmarks. LIDAR information can be utilized to create even richer maps with greater detail. A combination of artificial intelligence and VSLAM (Visual Simultaneous Localization and Mapping) handle all stages of map creation.</p><p>5. Nvidia AI co-pilot</p><p>This system is to monitor the driver of the car to make a better and smart decision on the road and help the driver avoids danger. The system helps you have a better driving experience as it detects your movement in the car while you’re driving. It uses sensor data from the microphone and cameras, both inside and outside the car, to track the environment around the driver. It can warn you and make a decision if the driver is in danger, for example break when there’s a collusion risk. Co-Pilot is always engaged. Using deep learning networks, it can track head movement and gaze, and have a conversation with you using advanced speech recognition, lip reading, and natural language understanding.</p><p>I believe this technology will be implemented in the next Audi A8 model.</p><p>To summarize I think what Nvidia did with their End to End solution for AI in General and Self driving cars in particular, is more than incredible. Nvidia has already partnered with many leading carmakers to work on self-driving cars. No wonder they got to be the #1 in the smartest companies<a href="https://www.technologyreview.com/lists/companies/2017/"> list by MIT tech review 2017</a></p><p>I can’t wait to get my hands on one of their autonomous cars, and off the wheel.</p><p>I added a video below of their self-driving car demo from CES 2017, so you can see for yourself the hard work Nvidia put into this project.</p><p>Enjoy :)</p><p><a href="https://www.youtube.com/watch?v=fmVWLr0X1Sk">https://www.youtube.com/watch?v=fmVWLr0X1Sk</a></p><ul><li>Source of many parts of this post is the Nvidia website.</li></ul><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fupscri.be%2F8f5f8b%3Fas_embed%3Dtrue&amp;dntp=1&amp;url=https%3A%2F%2Fupscri.be%2F8f5f8b&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=upscri" width="800" height="400" frameborder="0" scrolling="no"><a href="https://medium.com/media/c43026df6fee7cdb1aab8aaf916125ea/href">https://medium.com/media/c43026df6fee7cdb1aab8aaf916125ea/href</a></iframe><figure><a href="https://becominghuman.ai/artificial-intelligence-communities-c305f28e674c"><img alt="" src="https://cdn-images-1.medium.com/max/255/1*2f7OqE2AJK1KSrhkmD9ZMw.png" /></a></figure><figure><a href="https://upscri.be/8f5f8b"><img alt="" src="https://cdn-images-1.medium.com/max/255/1*v-PpfkSWHbvlWWamSVHHWg.png" /></a></figure><figure><a href="https://becominghuman.ai/write-for-us-48270209de63"><img alt="" src="https://cdn-images-1.medium.com/max/255/1*Wt2auqISiEAOZxJ-I7brDQ.png" /></a></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1f00e3212508" width="1" height="1" alt=""><hr><p><a href="https://becominghuman.ai/nvidia-and-the-gpu-contribution-to-the-ai-world-of-self-driving-cars-1f00e3212508">Nvidia and the GPU: contribution to the AI world of self-driving cars</a> was originally published in <a href="https://becominghuman.ai/">Becoming Human: Artificial Intelligence Magazine</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How does Baidu’s plan to attain self-driving car leadership?]]></title>
            <link>https://medium.com/@liorswissa/how-does-baidus-plan-to-attain-self-driving-car-leadership-1680c52d05d2?source=rss-1ae9eb713cd8------2</link>
            <guid isPermaLink="false">https://medium.com/p/1680c52d05d2</guid>
            <category><![CDATA[baidu]]></category>
            <category><![CDATA[automotive]]></category>
            <category><![CDATA[autonomous-cars]]></category>
            <category><![CDATA[self-driving-cars]]></category>
            <category><![CDATA[open-source]]></category>
            <dc:creator><![CDATA[Lior Swissa]]></dc:creator>
            <pubDate>Wed, 21 Jun 2017 13:49:17 GMT</pubDate>
            <atom:updated>2017-06-21T13:49:17.749Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gA_R0NHrTBEMVzAagyFolg.jpeg" /></figure><p>I went to the coolest show in town last week — CES ASIA</p><p>The first thing I did was to go to the automotive section to enjoy and learn more about automakers’ plans for the future. It assured me that all automakers are progressing towards self-driving cars. My first stop was at Baidu’s self-driving car.</p><p>Baidu started to work on autonomous vehicles back in 2014. They have a great group of talents both in Silicon Valley and Beijing. Baidu is already a tech giant, using AI for their search engine (China’s leading search engine) and also have enhanced maps platform which basically means that they have A LOT of data — which plays a big role generally in AI and particularly in self-driving cars. Moreover, until last April the “King of AI”, Andrew Ng, was leading Baidu’s AI technology group — showing how seriously Baidu takes this initiative.</p><p>I was interested in Baidu’s self-driving car because this is the only company that opened their platform — “Apollo” — to the public so that they could enhance the technology and bring it to market the fastest.</p><p>Apollo platform includes a vehicle platform, a hardware platform, a software platform and cloud data services, will be opened to help others in the industry, particularly car manufacturers, to develop self-driving cars.</p><p>But wait — aren’t tech companies supposed to compete and save the data to themselves so that their product will be the best in the market?<br>I personally think this is a smart move from Baidu that decided to open source their code/platform starting July, and thus can partner with a lot of other companies that not necessarily have the full capabilities to develop self-driving cars, but still have some capabilities that can complement Baidu’s strengths, and help Baidu get to a leading position.</p><p>Recently, The COO of Baidu said: “An open, innovative industry ecosystem initiated by Baidu will accelerate the development of autonomous driving in the US and other developed automotive markets.”</p><p>Baidu not only invests their resources in software, but has also taken active positions in self-driving car hardware. For example, Baidu invested $75M USD in “Velodyne LiDAR” company, based in the US. Lidar sensor is a crucial component for driverless autonomous vehicles, constantly scanning the vehicle’s surroundings and comparing the images with previously created 3-dimensional maps. Baidu uses this technology not only for their cars, but also to create its 3D street view version of Baidu Maps, using a fleet of small vehicles with a roof-based sensor, looking very much like the Google Streetview cars which use the same technology.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/837/1*I201ysSYgtgVKt048UeAKA.jpeg" /></figure><p>Baidu’s efforts are getting attention from some of AI’s leading companies, possibly because of their deep roots in the Chinese market — a key future market for self-driving cars.</p><p>In September 2016, Baidu and Nvidia announced a partnership that combines Nvidia’s self-driving computing platform with Baidu’s cloud and mapping technology to develop an algorithm-based operating system capable of powering complex navigation systems in autonomous vehicles.</p><p>“In order to have high quality autonomous driving, you need to have an AI super computer in the car, AI algorithms that are doing perception reasoning and driving, rich amount of software and it has to be connected to a HD cloud that gives you information around your surrounding” said Nvidia CEO Jen-Hsun Huang in 2016 GTC China conference. He continued “This partnership between those two companies of enormous scale allows us to solve the problems of autonomous driving from in car computer all the way to the cloud computer. From the algorithms in the car all the way to the algorithms in the cloud and for it to be connected all over China. What’s unique about this partnership is that whatever is developed in this platform will be made available to car OEMs who would like to participate — it’s an open platform”.</p><p>China is already the biggest market for automakers but it is also one of the most complex in terms of driving conditions. “Fully autonomous driving under mixed road conditions is universally challenging, with complexity further heightened by Beijing’s road conditions and unpredictable driver behavior,” said Wang Jing, SVP of Baidu and General Manager of Baidu’s Autonomous Driving Business Unit.</p><p>I think this is a huge challenge to test drive here in China but Baidu is on the right path — to paraphrase Frank Sinatra’s famous song — for self-driving cars — “if you can make it in China, you can make it anywhere”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1/0*sWKj_f-m2fPmVOBL." /></figure><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FKVLKfxLlp5U%3Ffeature%3Doembed&amp;url=http%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DKVLKfxLlp5U&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FKVLKfxLlp5U%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="640" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/dfe6b0db7340680fe9a3592937d82ec2/href">https://medium.com/media/dfe6b0db7340680fe9a3592937d82ec2/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/1/0*5UPF2OLDu-urvMoa." /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1680c52d05d2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to physically build a Self-Driving car ?]]></title>
            <link>https://medium.com/@liorswissa/how-to-physically-build-a-self-driving-car-8a4525a740a4?source=rss-1ae9eb713cd8------2</link>
            <guid isPermaLink="false">https://medium.com/p/8a4525a740a4</guid>
            <category><![CDATA[automotive]]></category>
            <category><![CDATA[autonomous-cars]]></category>
            <category><![CDATA[sensors]]></category>
            <category><![CDATA[hardware]]></category>
            <category><![CDATA[self-driving-cars]]></category>
            <dc:creator><![CDATA[Lior Swissa]]></dc:creator>
            <pubDate>Mon, 12 Jun 2017 13:17:43 GMT</pubDate>
            <atom:updated>2017-06-12T13:20:03.136Z</atom:updated>
            <content:encoded><![CDATA[<p>A great online lecture from MIT about “Deep Learning for Self-Driving Cars” by Lex Fridman, describes among the rest the hardware required for self-driving cars. I’d like to share with you a summary of what I have learned, especially on physical sensors that are added to cars to allow them to ‘drive themselves’.</p><p>First, what are the main tasks for a car to drive itself?<br>a. Localization and Mapping — “where am I?”<br>b. Scene Understanding — “Where is everyone else?”<br>c. Movement Planning — “How do we get from A to B?”</p><p>For a human these tasks sound simple, but when you think about it, this requires a lot of input (even before considering software and calculation). Let’s dive in into the hardware required for those tasks:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/728/1*1agYEfWr7ejyAV1u_JDnNQ.jpeg" /></figure><p><strong>GPS/IMU</strong> — GPS is an essential function for autonomous vehicles, but it’s not sufficient by itself. The GPS signal is blocked by canyons, tunnels, radio interference, and many other factors, and these outages can last for many minutes and longer. To supplement the GPS, the autonomous vehicle uses inertial guidance which requires no external signal of any type. The inertial measurement unit (IMU) consists of a platform fixed to the vehicle, and this platform has three gyroscopes and three accelerometers, one pair oriented each for of the orthogonal X, Y, and Z axes. These sensors provide data on the rotational and linear motion of the platform, which then is used to calculate the motion and position of the vehicle regardless of speed or any sort of signal obstruction. Note that an IMU cannot tell you where you are, only the motion, so the initial location of the vehicle must be determined by GPS or entered manually.</p><p><strong>Stereo vision camera</strong> — This is a type of camera with two or more lenses with a separate image sensor or film frame for each lens. This allows the camera to simulate human binocular vision, and therefore gives it the ability to capture three-dimensional images, a process known as stereo photography (similar to the human process — having 2 ‘pictures’ for depth assessment).<br>The stereo input also helps to undistort and rectify the images. We have 2 images and from that need to compute the depth map. For every single pixel we need to compute the best estimate of the depth of that pixel, the 3-dimensional position relative to the camera, to compute the disparity map to get the distance. It detects the lane’s width and the proximity of an object on the road. In other words, this sensor allows the car to follow the lanes and the cars on the road.</p><p><strong>Radar </strong>— This is an object-detection system that uses radio waves to determine the range, angle, or velocity of objects. It sends electromagnetic waves once they hit an object the waves are reflected revealing how far the object is and how fast it’s approaching. In other words, it measures the distance between other objects and gets the speed of the objects toward us. <br>*This technology doesn’t measure height, only scans horizontally.</p><p><strong>Lidar (Light Detection and Ranging)</strong> — This is a remote sensing method that uses light in the form of a pulsed laser to measure ranges (variable distances) to Earth. A Lidar creates a 3D image of the car’s environment with the help of the data from the camera sensors classifies the object in the scene.<br>Lidars are the best sensors to give the least 3D noisy information of the external environment.<br>*This technology is sensitive and is not good in harsh weather conditions such as rain, snow etc.</p><p><strong>Infrared camera</strong> — This is a device that forms an image using infrared radiation. We use it for night drive when the vision is very limited, to produce a thermal image and perform temperature calculations, find objects at night and present it on the dashboard (especially for semi-self-driving cars).</p><p><strong>CAN (Controller Area Network)</strong> — This is not a sensor, but rather the “CPU” of the car. The modern automobile may have as many as 70 electronic control units (ECU) for various subsystems. Typically the biggest processor is the engine control unit. Others are used for transmission, airbags, antilock braking/ABS, cruise control, electric power steering, audio systems, power windows, doors, mirror adjustment, battery and recharging systems for hybrid/electric cars, etc. Some of these form independent subsystems, but communications among others are essential. A subsystem may need to control actuators or receive feedback from sensors. The CAN controls the commands of the car.</p><p>All of the hardware mentioned above just accumulates a collection of data, simulating our eyes and ears, however, software plays an important role to simulate our brain and understand the input and process the right output and decisions such as the direction of the car, where to go, in what speed etc. This is naturally a big topic for another blog post.</p><p>A current issue with available sensors is their (high) cost, especially for Lidars. <br>The hardware can cost more than tens of thousands of dollars (Not even taking into account the cost of the expensive talented engineers in Deep learning required for design). Companies in automotive are trying to reduce that cost by implementing better software with less expensive hardware.</p><p>Let’s take Tesla as an example. This is the only current company (that I know of) that is not using a Lidar for their autopilot system and are relying solely on cameras and a radar. Hence the system is limited to certain driving areas, for example not yet capable for driving in city zones.<br>But let’s remember it’s only a semi-automated car, as opposed to other self-driving cars such as Waymo (Google self-driving car) that rely on Lidar heavily. In fact, this technology is what made “them” (Stanford team’s Stanley) win the Darpa challenge back in 2005 (On a side note — I’m Still amazed it’s been 12 years since this technology really emerge to the world and not yet out there on the roads). Moreover, the current importance of the Lidar is shown by the big lawsuit of Waymo against Uber regarding the designs of custom lidar sensors (I don’t have facts on that matter, just according to what I have seen on the news).</p><p>The best examples for reducing costs of hardware by improving software example so far to build a self-driving car come from startup companies that give a hardware kit with their deep learning solutions. An exceptional solution seems to come from the startup company Drive.ai. Their technology is a sort of a kit that you might buy when you purchase a vehicle — with software that can be integrated with potentially any car.</p><p>For us driving seems easy but when it comes down to numbers there are a lot of accidents and lives taken on the road. I believe that with those sophisticated sensors we can get a better analysis of the environment and react much better (and quicker) to dangerous situations on the road, resulting in saving many lives. The Tesla video below is a great example. Moreover, as technology gets more sophisticated, future generations would probably be shocked we have ever driven without sensors and computer aided safety measures.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FAPnN2mClkmk%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DAPnN2mClkmk&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FAPnN2mClkmk%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/8afdc0dec502aa619f712987dc94785a/href">https://medium.com/media/8afdc0dec502aa619f712987dc94785a/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8a4525a740a4" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>