<?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 Junko Yoshida on Medium]]></title>
        <description><![CDATA[Stories by Junko Yoshida on Medium]]></description>
        <link>https://medium.com/@jyoshidaparis?source=rss-9453c74879da------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*zZ1PwC-MXH1aWUjkJt52-w.jpeg</url>
            <title>Stories by Junko Yoshida on Medium</title>
            <link>https://medium.com/@jyoshidaparis?source=rss-9453c74879da------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 16:15:53 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@jyoshidaparis/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[Will Aurora ‘Do the Right Thing’?]]></title>
            <link>https://medium.com/@jyoshidaparis/will-aurora-do-the-right-thing-9178c137958f?source=rss-9453c74879da------2</link>
            <guid isPermaLink="false">https://medium.com/p/9178c137958f</guid>
            <category><![CDATA[autonomous-vehicles]]></category>
            <category><![CDATA[standards]]></category>
            <category><![CDATA[autonomous-cars]]></category>
            <category><![CDATA[sec]]></category>
            <category><![CDATA[safety]]></category>
            <dc:creator><![CDATA[Junko Yoshida]]></dc:creator>
            <pubDate>Mon, 23 Aug 2021 21:18:56 GMT</pubDate>
            <atom:updated>2021-08-24T11:36:49.492Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UvcvTG7lekpHL8nGZ0bnKg.jpeg" /></figure><p>Debates over the safety of autonomous vehicles boil down to one simple question: Is a machine driver safer than a human being?</p><p>Beliefs and convictions abound among both experts and lay people on the topic of robocars. Yet, today, no regulators, no carmakers, no AV startups, and no industry organizations appear ready to answer the $64 billion question with certainty.</p><p>It’s easier to skirt the issue.</p><p>Assume that you believe in highly automated technologies, and you claim that robocars, when commercially deployed at scale, will save people’s lives. The question, then, becomes how can you prove, with a high degree of certainty, that robocars are safer than “hand-driven” vehicles? What’s the scientific evidence that supports your claim?</p><p>On the other hand, let’s say you don’t believe in highly automated technologies backed by machine learning. You think the technologies remain immature. You argue that human drivers are safer than machines. Here again, asked to prove your contention, what’s the evidence?</p><p>William Widen, professor at the University of Miami School of Law, calls this conundrum the “AV problem.” In a recent conversation, he bluntly pointed out to me that “It seems whoever has the burden [of proof], loses [the argument].</p><p>Widen is not a safety expert. He teaches law. However, his years of experience as a lawyer on Wall Street give him an illuminating lens on the corporate finance world. Today, he’s applying this expertise to the process of AV companies trying to balance their technology quandary — “how safe is safe enough for AV technology to be deployed at scale” — with their financial goals and the milestones they must pass to keep investors happy.</p><p>Widen looked specifically into the self-driving car startup Aurora. He recently published a paper called <a href="https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3902217">Autonomous Vehicles, Moral Hazards &amp; the “AV Problem.”</a></p><p>Last month (July 15, 2021), Aurora announced it would go public by merging with a blank-check company Reinvent Technology Partners Y (RTPY). Aurora values the deal at $13 billion.</p><p>As part of that business combination procedure, <a href="https://www.sec.gov/Archives/edgar/data/1828108/000119312521216134/d184562ds4.htm">RTPY filed Form S-4</a> with the US Securities and Exchange Commission, which Widen has examined.</p><p>Risk factors admitted in Aurora’s S-4 crystallize the AV conundrum. In the S-4, here’s Aurora wrote:</p><p><em>Solving self-driving is one of the most difficult engineering challenges of our generation. The industry can be characterized by a significant number of technical and commercial challenges, including </em><strong><em>an expectation for better-than-a-human driving performance</em></strong><em>, large funding requirements, long vehicle development lead times, specialized skills and expertise requirements of personnel, inconsistent and evolving regulatory frameworks</em><strong><em>, a need to build public trust</em></strong><em> and brand image, and real world operation of an entirely new technology. If we are not able to overcome these challenges, our business, prospects, financial condition, and results of operations will be negatively impacted and our ability to create a viable business may not materialize at all.</em></p><p>(Emphasis is mine)</p><p>As for safety standards, Aurora explained in the S-4, “There has been relatively little mandatory government regulation of the self-driving industry to-date. Currently, there are no Federal Motor Vehicle Safety Standards (FMVSS) that relate to the performance of self-driving technology. Further, there are currently no widely accepted uniform standards to certify self-driving technology and its commercial use on public roads.”</p><p>In other words, Aurora acknowledges that it has little idea of the safety standards to which it might have to conform in the future. It says, however, that it’s committed to “doing the right thing.” Aurora claims to have built “organizational, operational, and safety processes to ensure that the performance of our technology meets rigorous standards.”</p><p>Meanwhile, the form S-4 describes Aurora’s plan to commercialize the “Aurora Driver” [Aurora’s software-based machine driver] in late 2023 for trucking and late 2024 for ride hailing. If Aurora can’t hit these deadlines, it will miss its financial goals. Given that time is money, it appears Aurora have opted for an easier burden [of proof for safety] and will deploy the AV Driver by 2023, Widen observed. “At least, it wants to keep the option open to satisfy a lesser deployment standard,” he added.</p><p><strong>Has Aurora given us enough disclosure?</strong></p><p>So, I asked Widen: Is there anything in the S-4 — such as a lack of disclosure — that might pose legal trouble for Aurora?</p><p>Widen observed, “I believe that the S-4 was done very carefully, by people who know a great deal about securities law liability. They’re very careful in the disclosure of the risk factors. And they’re very careful in the way that they word them.” Widen sees Aurora as having kept its option open to following “a very loose standard” to deploy AVs in time “without too much securities law liability.”</p><p>He explained, “One of the things that Aurora is very careful about — even though they have had an opportunity to say otherwise — is they’re very careful to not say, what Aurora’s standard is.” On one hand, “they do tell you that the industry expects the machine driver performs better than a human driver.” On the other, they never articulate how the Aurora Driver performs when compared to a carbon-based life form.</p><p>As Widen explains this paradox, the AV startup believes that one of its challenges in commercializing its technology, is “to meet public expectation and get public acceptance.” And yet, Widen noted, “they don’t take the next step and tell you what their [own] standard is to meet the public’s expectation at the time of deployment.”</p><p>If Aurora truly wants to build trust, “they would lay out the standard. They certainly would describe all the benefits that might come from their [AV] technology. But they would also say what their standard is. That standard is material, because the timing of deployment of the technology is critical to the success of the company. One can meet a looser standard faster than a strict one.”</p><p>Aurora’s filed S-4 offers a lot of catchy marketing goals and aspirational corporate values. They include: “Do the right thing” (Our task is to build and earn trust with everything that we do), “No jerks” (We assume our people’s best intentions and we are all working toward a common goal), and “Set outrageous goals” (We impose ambitious objectives and move urgently to deliver our best work).</p><p><strong>Conclusion</strong></p><p>With all the “oratory” (that’s how Widen describes it) that goes into its S-4, the startup offers little insight into a substantive plan to make its Aurora Driver safer than a human behind the wheel. As a reporter, I find it difficult to take Aurora’s word for it when it says it “will do the right thing.” First tell me what your deployment standard is. Then show me the proof.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9178c137958f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Can ADAS Aid Us on a Dark and Stormy Night?]]></title>
            <link>https://medium.com/@jyoshidaparis/can-adas-aid-us-on-a-dark-and-stormy-night-e55daca9e5d7?source=rss-9453c74879da------2</link>
            <guid isPermaLink="false">https://medium.com/p/e55daca9e5d7</guid>
            <category><![CDATA[bad-weather]]></category>
            <category><![CDATA[advanced-driver-assist]]></category>
            <category><![CDATA[sensors]]></category>
            <category><![CDATA[autonomous-cars]]></category>
            <dc:creator><![CDATA[Junko Yoshida]]></dc:creator>
            <pubDate>Thu, 08 Jul 2021 16:30:21 GMT</pubDate>
            <atom:updated>2021-07-08T16:30:21.917Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fe4RjShZcDBxnURbnR68eA.jpeg" /></figure><p>It’s one thing to ridicule Elon Musk for realizing, belatedly, that real-world AI is easier preached than done. Last week, in a tweet, Musk wrote: “Generalized self-driving is a hard problem, as it requires solving a large part of real-world AI. Didn’t expect it to be so hard, but the difficulty is obvious in retrospect. Nothing has more degrees of freedom than reality.”</p><p>Shading the truth about highly automated vehicles’ capabilities isn’t Tesla’s hallmark alone. Most carmakers have yet to admit that, despite the many sensors installed in their Advanced driver assistance systems (ADAS) and autonomous vehicles (AVs), they remain vulnerable when driving in the dark, or into the sun, or in fog, heavy rain or snow.</p><p>“When the sun is up and birds are singing, yes, autonomous vehicles can mostly drive themselves,” said Eyal Levi, co-founder, executive vice president, business development &amp; product at BrightWay Vision Ltd. (Haifa, Israel) But a lot of driving routinely occurs after sunset and the weather is bad.</p><p>AV companies claim to have already nailed highly automated driving. However, “we know that those same companies are all still looking for solutions” to cope with stormy weather, said Levi.</p><p>Whether a car is driven by human or machine, many challenges trip up drivers, among them the element of surprise, when an emergency crops up suddenly. Bad weather and dark of night, however, are hardly a surprise.</p><p>Demonstrations/presentations offered by technology suppliers at the “Destination ACM” event last month in Detroit made clear that the next frontier in the race to autonomy is all about technologies that can improve perception and expand operational design domain (ODD), said Phil Magney, founder and president of VSI Labs. In his opinion, advancements in perception hold the key to steadily improve ADAS, “which will last over the next 20 to 25 years.”</p><p>Here are a few companies which strutted their stuff.</p><p><strong>Brightway Vision</strong></p><p>Brightway Vision (Haifa, Israel) offers a gated vision camera system with a Near-infrared CMOS image sensor and a pulse-based illuminator.</p><p>The ten-year-old startup, spun out of Elbit Systems — one of the largest defense electronics companies in Israel — has developed an “all weather automotive camera system,” poised to be automotive qualified by mid-2022.</p><p>The system, called Visdom, is no ordinary camera. It comes with a “smart flash” and a CMOS image sensor with a built-in “smart accumulator.” The setup supports integration of reflected light from multiple illumination pulses, according to Brightway Vision’s Levi. What makes the flash IR sensor smart is “its ability to place a specific amount of light at each range,” he said, “without affecting other ranges.”</p><p>This brings two major benefits, noted Levi. First, “it creates an eerily uniform lighting across all ranges — looking more like a daytime image than a standard active-illumination night image.” Second, “the use of sophisticated control of the pulse illuminator makes it possible to effectively eliminate backscatter.” The backscatter is the reason why main headlights of a vehicle can’t penetrate fog.</p><p>Brightway Vision uses Vertical-Cavity Surface-Emitting Laser (VCSEL) as its illuminator. Unlike typical lidars using VCSEL to measure distance, Brightway Vision deploys VCSEL for “collecting light,” Levi said.</p><p>By pulsing the shutter 1,000’s times per frame, the camera captures “slices” of an image and accumulates its multiple exposure — uniformly lit across all ranges — into a single image. The camera “combines these slices of images in the analog domain,” noted Levi. One interesting use of slices is that they allow “background removal” at any desired range. The result is significantly improved image contrast, enabling AI and humans to detect objects much better, Levi claimed.</p><p>Brightway Vision, nonetheless, believes its camera is the “ultimate” complementary automotive sensor to color cameras in ADAS cars or AVs. Many experts agree that cameras are the key sensor for autonomous driving, but “the shortcomings of regular cameras are low light, and they are adverse to bad weather,” Levi noted.</p><p>Visdom will change that equation, claimed Levi, because it can offer the same level of performance at low visibility, as in clear weather.” The main functional shortcoming of Birghtway Vision’s Visdom camera is that it cannot provide color.</p><p>“There will be no single sensor solutions for AVs,” said Levi. The company’s CMOS-based NIR gated sensors can be also used “as redundancy or verification” of objects detected but not classified by other types of sensors. Assume, for example, a radar picks up a blip on the road. Brightway Vision’s cameras can shed more light on that blip — captured on a slice. Given that the API for its camera is open, communication with other sensors is possible.</p><p>VSI Labs’ Magney, whose team has been testing Brightway technology, noted that the startup’s Visdom “shows just how good a camera can be in challenging conditions.” His team has had on-the-road experience in a vehicle equipped with Visdom when the car was suddenly engulfed in torrential rain. Thus far, he said, “nothing is better” under such circumstances than this camera. However, he added, Visdom is “more complex than a regular camera since it requires emitters.”</p><p>Visdom’s camera system requires carmakers to deploy both a camera and illuminators, which adds complexity in integration. Some OEMs and Tier Ones consider it “challenging,” acknowledged Levi.</p><p>However, there are some advantages in the system integration. The Visdom camera can be placed behind the windscreen, just like any other CMOS image sensors. It means that carmakers do not need to worry about cleaning and cooling the cameras, or fender bender minor accidents damaging the camera. Moreover, the illuminators can be located in the car headlights, thus protecting and concealing them.</p><p>That said, the uniqueness of Brightway Vision’s technology could make Visdom a hardsell to some risk-averse OEMs and Tier Ones. They hesitate when they find out that Brightway Vision is the only company combining a custom-made CMOS image sensor and a pulse-based illuminator.</p><p><strong>Nira Dynamics</strong></p><p>Meanwhile, Nira Dynamics (Linköping, Sweden), also at the Destination ACM to tout its technology, unveiled software that can presumably better prepare drivers (machine or human) of AVs and ADAS for sudden changes in road conditions in adverse weather.</p><p>Nira is a world leader in indirect tire pressure monitoring systems. Nira’s software — installed in many vehicles — dynamically measures road surface info such as tire grip, air pressure and loose wheels. Fusing information from anti-lock braking systems (ABS), microslip, accelerometers and steering angles, Nira’s software can detect friction with the road, collect data on road moisture and send it to car OEMs’ cloud. From this, Nira anonymizes the data and sends alerts of upcoming road conditions back to the driver.</p><p>Magney noted, “Nira’s dynamic algorithm will enable AVs to operate in slippery conditions. It also works well for ADAS.”</p><p>An advantage is that all vehicles, ADAS, AV and otherwise, can use data collected by Nira without adding yet another sensor devoted to safety, explained and Gregory Weber, business development manager at Nira.</p><p>While knowledge about road conditions enables safer driving, Nira also sees a growing opportunity in offering its data to municipalities, for example. States or cities can in turn use it to make road maintenance more environmentally friendly and managing more efficiently such tasks as snow-plowing and pothole mending. Once Nira’s road information is overlaid atop weather information, it could be helpful to trucking companies planning routes, for example, Weber added.</p><p><strong>Aeye</strong></p><p>Aeye believes that its lidar belongs to a class of its own, because it has “bistatic lidar architecture.” Under the patented architecture, transmission and reception channels for lasers are kept separate. Aeye’s chief marketing officer Stephen Lambright claimed that his company’s iDAR, can bring “more resolution where it is needed.”</p><p>As each laser pulse is transmitted, the receiver is told where and when to look for its return. This makes it possible to introduce “deterministic artificial intelligence into the sensing process at the point of acquisition.” Describing the process as “biomimicry,” Aeye explained that it allows iDAR to focus on what matters most in a vehicle’s surroundings.</p><p>One of the known issues of LiDAR sensors, although not broadly acknowledged, is the degradation of its performance in rain. If a lidar beam intersects with a raindrop close to the transmitter, it can reflect enough of the beam back to the receiver to register rain as an object. The droplets can also absorb some of the emitted light, degrading the sensors’ range of performance.</p><p>Demonstrating iDAR at the American Center for Mobility (ACM), Aeye even prepared a rainmaking machine to show off the sensor’s object detection capability in wet weather. As things turned out , it actually rained on Aeye’s parade, helping iDAR prove its point.</p><p>Aeye’s customers can create a library of deterministic, software-configurable scan patterns at design time, each addressing a specific use case, according to the company. In addition to creating different scan patterns for highway, urban, and suburban driving, or an “exit ramp” pattern, Aeye explained that the customer can create scan patterns for those same driving environments but optimized for bad weather — such as “Highway rain scan pattern” vs “Highway sunlight scan pattern.”</p><p>Earlier this year, VSI Labs oversaw the testing of Aeye’s idea system.</p><p>VSI Labs confirmed that 1) the AEye sensor could detect targets with a substantial number of points from farther than 1000 meters without compromising frame rate, 2) AEye’s sensor produced more than 175 points on an 18% target from 200 meters, doing so with 1600 points per degree squared, 3) iDAR scanned at better than 200Hz for full frame and field of view (FOV), and 4) iDAR senses effectively behind a windshield glass at various angles.</p><p>Asked about iDAR, Magney summed up, “It is pretty cool because it can see up to a kilometers and the iDAR can operate behind the windscreen.” He added, “Long range lidar is going to be important for trucking especially.”</p><p>— <em>Junko Yoshida, the former bureau chief and editor-in-chief of EE Times and most recently, global editor-in-chief of Aspencore, is now an independent technology and business journalist.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e55daca9e5d7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Born-Again NHTSA Forces Auto Industry Choices]]></title>
            <link>https://medium.com/@jyoshidaparis/born-again-nhtsa-forces-auto-industry-choices-b0ee044f5fe1?source=rss-9453c74879da------2</link>
            <guid isPermaLink="false">https://medium.com/p/b0ee044f5fe1</guid>
            <category><![CDATA[safety]]></category>
            <category><![CDATA[automated-driving]]></category>
            <category><![CDATA[nhtsa]]></category>
            <category><![CDATA[autonomous-cars]]></category>
            <dc:creator><![CDATA[Junko Yoshida]]></dc:creator>
            <pubDate>Fri, 02 Jul 2021 15:22:16 GMT</pubDate>
            <atom:updated>2021-07-02T15:22:16.846Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dS0-a-jhaUK3BTcLMm0SEg.png" /></figure><p>The National Highway Traffic Safety Administration’s (NHTSA) order this week requiring carmakers and tech companies to submit crash reports involving vehicles equipped with highly automated features — including both Automated Driving Systems (ADS)and Advanced Driver Assistance Systems (ADAS) — is a sensible move.</p><p>It’s also wildly out of character. Known for its hands-off approach to automakers, NHTSA has traditionally permitted automakers a lot of leeways including self-certifying their own vehicles with new technologies.</p><p>Under the new order, NHTSA enforcement is, at least for now, reborn. It is asking vehicle makers to “within one day of learning of a crash, report crashes involving a Level 2 ADAS or Levels 3–5 ADS-equipped vehicle that also involve a hospital-treated injury, a fatality, a vehicle tow-away, an air bag deployment, or a vulnerable road user such as a pedestrian or bicyclist. An updated report is due 10 days after learning of the crash.</p><p>Phil Koopman, co-founder of Edge Case Research and professor at Carnegie Mellon University, noted in his LinkedIn post, “Good to see National Highway Traffic Safety Administration NHTSA being more proactive on automated vehicle safety. This is a nice first step in that direction.”</p><p>While safety experts have applauded NHTSA’s more rigorous approach, it is hard to imagine that automotive and tech companies will take the hint. Expect resistance, a lot of questioning or even attempts to foil the order.</p><p>At stake is the data.</p><p>The issue is not only how, or if, carmakers will follow orders and submit timely crash reports. It’s about how NHTSA intends to use the information.</p><p>Phil Magney, founder and president of VSI Labs, told us, “NHTSA is a little vague as what they are going to do with the data.”</p><p>Today, the collection of crash reports on ADS is entirely left up to each state. For example, in California, the department of motor vehicles (DMV) requires manufacturers testing autonomous vehicles to report within ten days any collision that resulted in property damage, bodily injury, or death. As of June 21, 2021, according to DMV, California has received 316 Autonomous Vehicle collision reports.</p><p>Under current regulations, the department relies completely on the testing company’s account of what happened. More important, it remains unclear what exactly anyone can learn, at a timewhen many of such crash reports filed by Waymo or Cruise tend to blame the other vehicle.</p><p>Further, the Calif. DMV does not collect crash reports from carmakers who market vehicles with Level 2 ADAS features.</p><p>NHTSA, presumably alarmed by a host of accidents related to Tesla’s so-called Auto Pilot (Level 2) features, might have decided to “get decent data on accidents related to Auto Pilot,” speculated Magney. After all, “there is nothing in place now that addresses this,” he added.</p><p>But even if there is no outright revolt from automakers, NHTSA’s order presents logistical problems for car OEMs. The foremost issue is how the industry (OEMs) is supposed to keep track of accidents, Magney pointed out. “The order says that once the OEM becomes aware, they are required to report within 24 hours. OEMs may routinely deny submissions [of crash reports]” because nobody told them.</p><p>“Most OEMs don’t have knowledge of one of their L2 vehicles being involved in a qualifying reportable event,” said Magney. “Most don’t get timely information about crashes, much less information as to whether L2 ADAS was engaged at the time.” Based on what we learned from Tesla’s past accidents, “Tesla has this information readily available, but this is not the case with most other OEMs.”</p><p>Magney predicts that the NHTSA order will “cause some turmoil among OEMs who don’t have a means of monitoring the performance of their ADAS/AD systems.” Combing through police reports is an option, but Magney cautioned, “This likely yields more questions than answers. Here again the police reports don’t capture the technologies in the cars or whether those system were active or not.”</p><p>In a statement, Steven Cliff, NHTSA’s acting administrator, noted, “By mandating crash reporting, the agency will have access to critical data that will help quickly identify safety issues that could emerge in these automated systems.” Clearly, the NHTSA’s administrator’s heart is in the right place, with a conviction that “gathering data will help instill public confidence that the federal government is closely overseeing the safety of automated vehicles.”</p><p>Now the ball is in the industry’s court. Will tech companies and automakers comply with NHTSA? Or would they rather wait until the ADAS/ADS death toll gets high enough to make them look bad?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/729/1*KE6uZR3wN-rBDyMF2_pFiQ.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b0ee044f5fe1" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>