<?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[nopilot.ai — Autonomous software for self-learning robots. - Medium]]></title>
        <description><![CDATA[We are building an experience-based system for autonomous robots. Our system is based on recent advances in deep learning. It could navigate anywhere by using prior experience and its own acquired knowledge. - Medium]]></description>
        <link>https://medium.com/nopilot?source=rss----115006a4f4d---4</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>nopilot.ai — Autonomous software for self-learning robots. - Medium</title>
            <link>https://medium.com/nopilot?source=rss----115006a4f4d---4</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 20:23:15 GMT</lastBuildDate>
        <atom:link href="https://medium.com/feed/nopilot" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Job of the future is a Courier. Not a Data Scientist.]]></title>
            <link>https://medium.com/nopilot/job-of-the-future-is-a-courier-not-a-data-scientist-521dd23dd43f?source=rss----115006a4f4d---4</link>
            <guid isPermaLink="false">https://medium.com/p/521dd23dd43f</guid>
            <category><![CDATA[food-delivery]]></category>
            <category><![CDATA[egrocery]]></category>
            <category><![CDATA[self-driving]]></category>
            <category><![CDATA[robotics]]></category>
            <category><![CDATA[delivery-robot]]></category>
            <dc:creator><![CDATA[Mike Melanin]]></dc:creator>
            <pubDate>Wed, 03 Aug 2022 12:25:55 GMT</pubDate>
            <atom:updated>2022-08-03T12:25:55.056Z</atom:updated>
            <content:encoded><![CDATA[<p>This is not a usual TEDTalk you’d like to hear. The predictions of futurologists regarding future professions are very beautiful and inspiring, but let’s look at our future from a more skeptical angle.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TYl0ulgz5Otp0VetbPnY5w.png" /></figure><p>One thing you can be sure about is that mass consumption habits and everything around it have already changed. Instant delivery services will stay with us for a long time. The situation with Covid-19 only spurred mass adoption of such services. The US is lagging behind other developed countries, but by 2025 it is predicted that up to a quarter of the grocery retail orders will be fulfilled online.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/770/1*hz1FLz8y7wKR0cewF8WdKA.png" /></figure><p>At the same time, for some reason, everyone ignores the “elephant in the room”, or rather the “gray rhino”, — <a href="https://techcrunch.com/2022/06/03/the-delivery-market-is-coming-down-from-its-pandemic-highs/?guccounter=1&amp;guce_referrer=aHR0cHM6Ly93d3cuZ29vZ2xlLmNvbS8&amp;guce_referrer_sig=AQAAAHXeGd7nwOtjRurmYMv1_UZvugXWb_UjY5JYGw4-R2zQFuW0po4EnMw8_nsP2smr4PwDvcL1Hhg4AGO5CYCGR9Y6citTaaXctdGh8DywzaD92jAqTlaSx2Icz4mSENs4s2UfpvEBOFajDm7v5uTfHOKwURb0XA2LQ42flGtbV5HM">the economy of new instant delivery services does not work</a>. I’m not even saying that the idea of <a href="https://www.ted.com/talks/ali_kashani_a_friendly_autonomous_robot_that_delivers_your_food?language=en">hauling burritos in a two-ton car is pure madness</a>.</p><p>Even relying on bike couriers for local grocery deliveries is an inefficient use of human labor.</p><p>The personal courier who goes shopping and brings me food is a shared servant from the 19th century, but not the future I really want. High tech, low life.<em> </em>Hello Cyberpunk, please, stay in video games.</p><p>Did we dream of such a society, was that the promise of futurologists? Technology was supposed to free up people’s time, and not become a new means of oppressing some over others. Technology should free us from the 4D work of Dull, Dirty, Dangerous, and Dear,<strong> </strong>instead of giving a rise to<strong> </strong>to more professions of this kind.</p><p>Now let’s evaluate the level of complexity of implementing robotics and AI among the most popular professions. Again,<a href="https://www.careerprofiles.info/careers-largest-employment.html"> we will use the statistics for the USA, since most of the data comes from the Bureau of Labo</a>r.</p><p>A retail salesperson, cashier, taxi driver, nurse, waiter, bartender. Couriers have slipped out of the statistics because they don’t sign a regular job contract. <a href="https://therideshareguy.com/doordash-statistics/">But the reporting of delivery services reveals fairly accurate numbers.</a></p><p>We can immediately discard those professions that are associated with social intelligence (for example, a sales assistant), since the level of AI is not yet at the level to imitate human communication. We can also discard professions requiring phenomenal technological complexity, as it turns out.</p><p>Let’s take household chores. I can’t imagine a robot that even makes my bed, makes dishes, and cleans up dog poop without smearing it all over the floor. The cost of such a robot will skyrocket, and in some cases it will be impossible to manufacture.</p><p>Try to teach a manipulator of this sort to put dirty dishes, who knows of what shape, into the dishwasher without breaking anything. For this kind of manipulation alone there have been advanced competitions held and new control methods have been explored for several years.</p><p>In some cases, we are really making progress, although it is much slowlier than the claimed level. Robot-vacuums really do help with dust if you’re in the habit of not throwing anything on the floor. Self-driving cars are already rolling around our streets. Yes, their cost is still high, unpredictable traffic situations drive them into a dead end, and overpromises from top leaders are disappointing. Our icon, Elon, has been promising that “<a href="https://youtu.be/o7oZ-AQszEI">full autonomy will be ready this year</a>” for 5 years in a row with the same wording.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EgOtumUbkRiO7AUbUxfiHg.jpeg" /></figure><p>Here come delivery robots. The explosion of instant delivery has already skyrocketed the demand for couriers, and it’s only going to get worse (or better, depending on which side you’re on). What is the job of a courier? Pick up the delivery, cruising around town, call client and go up to the floor, if necessary. And I did not find any of the critical technological problems, why this cannot be automated now. Unless stairs, non-automatic doors and elevators can be a problem at this stage of integration with the infrastructure. Many insiders (delivery startups both robotics and conventional) told me that even now they can technically perform up to 80% of all retail deliveries with a robot!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LSIeDgff0VJ07P_6zQollg.jpeg" /></figure><p>Those who have reached this part may object that the autonomous robot is still expensive (sometimes it even includes lidar)! Here I have two arguments. Firstly, lidars are getting cheaper so we need to consider their cost in 2–4 years. Secondly, for limited autonomy, the problem can be solved using stereo cameras, computer vision and sometimes deep-learning algorithms (aka end-to-end approach), which significantly reduces the cost of the stack. Yes, this is not a trivial task and the only company that has done it so far is Starship.</p><p>My prediction is that new instant delivery services will not be able to provide the same level of service. Either they will focus on something more modest like interval delivery (see Amazon), or they will revise their valuation to investors (see Instacart). Only those who invested/collaborated/piloted robotics solutions already will be able to provide a new level of service and remain viable from a business point of view. <a href="https://www.wired.com/story/the-speedy-downfall-of-rapid-delivery/">Mass reduction of human-labored instant deliveries only proves my assumptions.</a></p><p>What can you do now and how does it affect you? For roboticists, I encourage you to be realistic about your next project. For municipalities — the speedy development of legislation and infrastructure for courier robots. After all, it is your task to provide a framework for implementing technologies in life, and not to let it unfold on its own, as was the case with electric scooters. For investors — look for your investment champion. For instant delivery companies, test your business model on a scale, without the crazy assumption that one courier will make up to 5 deliveries per hour.</p><p>Yes, the title is a little bit of clickbait. But if you are still looking seriously for an answer, then a lot of the professions will involve robotics. From mechanical and electronics specialist to ML and back-end developer with experience in robotics.</p><p>The favourite argument of those pessimistic about the future about tmass unemployment is not valid. Technologies are being adopted rather slowly and the demand for personnel outstrips supply in many ways. By the time robots will be (are) ubiquitous, relatively unskilled workers will already be in demand in other areas. Not to mention what new professions we will have: robot fleet manager, teleoperator, maintenance mechanic, and so on.</p><p>I wrote this article before the infamous conflict that rocked the global economy. And although this led to a correction of robotization plans for a couple of years, I am convinced that the mainstream of development will go in this direction. Delivery will also remain in demand and the first candidate for robotization.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=521dd23dd43f" width="1" height="1" alt=""><hr><p><a href="https://medium.com/nopilot/job-of-the-future-is-a-courier-not-a-data-scientist-521dd23dd43f">Job of the future is a Courier. Not a Data Scientist.</a> was originally published in <a href="https://medium.com/nopilot">nopilot.ai — Autonomous software for self-learning robots.</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Let robots learn like a child]]></title>
            <link>https://medium.com/nopilot/let-robots-learn-like-a-child-4becd13862de?source=rss----115006a4f4d---4</link>
            <guid isPermaLink="false">https://medium.com/p/4becd13862de</guid>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[ecommerce]]></category>
            <category><![CDATA[robotics]]></category>
            <category><![CDATA[delivery]]></category>
            <category><![CDATA[autonomous-vehicles]]></category>
            <dc:creator><![CDATA[Mike Melanin]]></dc:creator>
            <pubDate>Thu, 18 Apr 2019 16:40:28 GMT</pubDate>
            <atom:updated>2019-04-18T17:53:51.378Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9Mi43B0UQ3_g3CCFH9gOng.jpeg" /><figcaption>Mike Melanin (CEO at nopilot.ai) training robot for the follow-me use case.</figcaption></figure><p>Let’s talk e-commerce first. We are accustomed to thinking of e-commerce as something that appeared at the dawn of the Internet, then experienced rapid growth during the advent of Web 2.0 and became firmly established in our lives. All big players like Amazon, Alibaba, and JD have eaten e-comm market and innovations are rare or redundant. What surprised me is that e-commerce is still only a small fraction of retail spending. It takes only 10% worldwide with projected YoY growth 2% over the next decade. We will see how many untouched areas of retail will be drastically transformed by new technologies.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GdVzB4jk19yz5vT-3QnJKw.png" /><figcaption><a href="https://www.statista.com/statistics/534123/e-commerce-share-of-retail-sales-worldwide/">E-commerce share of total global retail sales from 2015 to 2021</a></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PsH3SbgOVfnFXjfbBwJn6Q.png" /><figcaption><a href="https://www.statista.com/statistics/255083/online-sales-as-share-of-total-retail-sales-in-selected-countries/">E-commerce sales as a percentage of total retail sales</a></figcaption></figure><p>Surely enough, there is always a catch. As Benedict Evans <a href="https://youtu.be/RF5VIwDYIJk">emphasizes</a> in his talk at <a href="https://a16z.com/tag/summit-2018/">Andreessen Horowitz Summit</a> last November the era of simple tools for low touch goods has passed. Explicit examples include airplane ticket aggregators, restaurant review sites. Now we face the need for full stack solutions where the information is used as a system rather than arbitrage. Those kinds of solutions require a lot more investments compared to services developed over the last 20 years. Another good example taken from Benedict’s talk is Yelp vs. DoorDash. We used to do restaurant reviews and now you can get a hot meal delivered to your door in less than 30-minutes. The same goes for apartment listings. Now with Opendoor, you can buy a home.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Oe-0tRLjoA2l7QRZ1zyUUA.png" /><figcaption>Benedict Evans, <a href="https://www.ben-evans.com/benedictevans/2018/11/16/the-end-of-the-beginning">Presentation: The End of the Beginning</a></figcaption></figure><p>The rapid <a href="https://www.statista.com/outlook/243/100/ecommerce/worldwide#market-revenue">growth in e-commerce</a> caused a surge in demand for delivery, particularly <a href="https://www.businessinsider.com/last-mile-delivery-shipping-explained">its most expensive part</a> — the point at which the package arrives at the buyer’s door, or so-called last-mile delivery. The Internet made us accustomed to instant services. We are keen on two-day delivery, same-day delivery, and in some cases — instant delivery. However, we are not ready to pay a premium for FMCG and restaurants delivery, <a href="https://www.mckinsey.com/industries/travel-transport-and-logistics/our-insights/technology-delivered-implications-for-cost-customers-and-competition-in-the-last-mile-ecosystem">McKinsey study said</a>.</p><p>Many startups and incumbents are aware of this opportunity and are trying to respond to this market needs using autonomous ground robots. A tsunami of great promise and huge investments in automotive self-driving was followed by a wave of interest in smaller <a href="https://www.nanalyze.com/2018/04/8-delivery-robot-startups-last-mile-delivery/">unmanned ground vehicles</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pswjv1IbpZDq_EgI71HUCw.jpeg" /></figure><p>If you’ll follow the AGV / UGV startup life cycle, you can see a pattern. They are mainly focused on hardware, mechanics and sensors. Hence, delivery is not a luxury industry where you can charge a premium for expensive sensors. The attempt to reduce the cost of sensors only makes autonomous navigation task much harder with a classical approach.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KrznASFZCYTDu_Dvb5iKvg.jpeg" /><figcaption>Autonomous delivery carts look the same.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/926/1*iq6AIDrZGI0JmOODZXNC-w.png" /><figcaption>Delivery is not a luxury industry where you can charge a premium for expensive sensors.</figcaption></figure><p>For driving policy AGV always use classical computer vision based approach that inherited from the automotive self-driving. In fact, most of them poorly utilize machine learning and only for <a href="https://medium.com/starshiptechnologies/how-neural-networks-power-robots-at-starship-3262cd317ec0">perception part</a>. The classical self-driving approach to autonomous driving produced numerous <a href="https://www.cnbc.com/2018/07/31/alphabet-waymo-pilot-for-public-transportation-in-phoenix-arizona.html">limited</a> <a href="https://venturebeat.com/2019/03/25/starship-technologies-rolls-out-delivery-robots-to-northern-arizona-universitys-flagstaff-campus/">pilots</a> in geo-fenced areas. But after 15 years of development with $5bn spent annually, there are still no fully autonomous commercial vehicles.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lz9mHX1S4hL00nbhNz15OQ.png" /><figcaption>Complex hand-coded systems for classical self-driving approach.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7jkwzw55-fucPpRMGvMXNA.png" /><figcaption>It’s not dead. It’s just resting.</figcaption></figure><p>It is enough to compare the complexity of rule-based systems for ADAS (advanced driver-assistance systems), to understand that intelligent behavior cannot be programmed in an explicit manner.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_PkCHNcyqMB8s5gTU7InKg.png" /><figcaption>Lines of software code in modern vehicles.</figcaption></figure><blockquote><em>It gets even worse in pedestrian zones devoid of structure. As the founders themselves admit that unlike a road, a sidewalk full of people and obstacles. </em><a href="https://www.wired.com/story/delivery-bots-have-awkward-sidewalk-interactions-too/"><em>It’s chaos</em></a><em>.</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/1*QX0YbHBHiaKLCSZar9Cr5A.gif" /><figcaption>Shibuya Crossing in Tokio</figcaption></figure><blockquote><em>But wait! Why do we need more sensors and rules instead of better brains? A person copes with the task of driving a car with only a stereo vision and vestibular apparatus.</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6iX7-ZR47JJ3ZSSVSqHh7w.jpeg" /><figcaption>Carnegie Mellon Tartan Racing vs Ryan Gosling in Drive</figcaption></figure><p>Try to remember where you got driving or riding skills for the first time? Do you remember how your father or mother taught you how to handle a bike? Your father obviously did not give you a notebook with analytical formulas for balance and trip planning. You certainly didn’t draw bounding boxes around objects to find your way through the crowd.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*j9XARmQz8BgnAm8Qz3JrXQ.png" /></figure><p>You simply looked around and learned by the means of trials and errors. I think everyone remembers hurt knees and bruises. At least I remember a little.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Lf2bP7hFopRZQAzMEGlJ5g.jpeg" /><figcaption>Mike learns to ride a bike.</figcaption></figure><p>We created our system in a similar way by allowing the robot to learn as a child. We placed a complex deep neural network at its core and used only low-res cameras as cheap sensors. This approach uses a huge amount of data and cameras provide it well enough.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MNNAJzBg_xPajNb_taX_bA.jpeg" /></figure><p>A sequence of images and control signals were fed as input to the neural network. With the help of a couple of modern machine learning techniques, this approach alone can achieve good results. It was reflected in the work of <a href="https://devblogs.nvidia.com/deep-learning-self-driving-cars/">Nvidia</a> and <a href="https://arxiv.org/abs/1710.02410">Intel Labs</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bgxCqAFQQv8KzPqIJf78jA.png" /><figcaption>Supervised pre-training</figcaption></figure><p>In the second stage, we use reinforcement learning. We train a robot without direct supervision. It learns from trials and errors. For training, we used both the real environment and the <a href="http://carla.org/">CARLA simulator</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*s_f0zkFONy31Huhz55nMqg.png" /><figcaption>Control model learning</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*sWPncdCNMVqxEKsVaO2xTQ.gif" /><figcaption>Trial and error in CARLA simulator</figcaption></figure><p>Combining two approaches allowed us to achieve impressive results on the low-res cameras alone.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FolFDBKJY_o8%3Ffeature%3Doembed&amp;url=http%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DolFDBKJY_o8&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FolFDBKJY_o8%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/318fbae23db892b4149385b946a05ba8/href">https://medium.com/media/318fbae23db892b4149385b946a05ba8/href</a></iframe><p>This proof of the concept shows the applicability of our approach for autonomous vehicles navigation in unmapped areas.</p><p>One of the amusing findings was that the robot does not only learn like a human but also fails in the same manner. On the GIF you can see how robot bumps into the glass door during a test. In our office, there is a poor fellow mathematician who meets his forehead with this glass every week.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*MdSx3Uzj1aV9qPDrNwyFdA.gif" /><figcaption>Learning by trial and error.</figcaption></figure><h3>What’s next for nopilot?</h3><p>We designed the new chassis with additional sensors and selected just-released Nvidia Jetson Nano as the GPU. The total cost of the setup is less than $1k. We called our new platform MULE (Multipurpose Unmanned Learning-based Electric vehicle). I would also like to emphasize that we do not concentrate on hardware, but our business model based on software IP Licensing. In our case, the hardware is only a tool needed for running technology and software validation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4T-kV8PMruRyBLTkw822Vg.jpeg" /><figcaption>New plarform — MULE (Multipurpose Unmanned Learning-based Electric vehicle)</figcaption></figure><h3>Honorable Mentions</h3><p>We strongly believe that the only way to make robots smart enough for real-life tasks is to let them learn. In our desire to teach, not program robots, we are not alone. <a href="https://wayve.ai/">Wayve</a> is the only company in SD which apply sophisticated models to autonomous cars. It is also worth mentioning the recent success of Open.ai in <a href="https://openai.com/blog/learning-dexterity/">learning robot hand</a> and <a href="https://venturebeat.com/2019/01/23/ai-teaches-4-legged-robots-how-to-recover-when-they-fall/">ETH Zurich</a> in “Recovery Quadrupedal Robot using Deep Reinforcement Learning”.</p><p>We will continue our experiments with various models in simulation and in a real environment. If you want to watch our robot “grow up” and getting more sophisticated, <strong>do not forget to join our journey!</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*J7PyxwVHqVeVOHC0cmAcIA.jpeg" /><figcaption>nopilot’s MULE (Multipurpose Unmanned Learning-based Electric vehicle) and Donkey Car.</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4becd13862de" width="1" height="1" alt=""><hr><p><a href="https://medium.com/nopilot/let-robots-learn-like-a-child-4becd13862de">Let robots learn like a child</a> was originally published in <a href="https://medium.com/nopilot">nopilot.ai — Autonomous software for self-learning robots.</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>