in

Lyft engineer sees self-driving as long game, calls for public prediction data

Lyft engineer sees self-driving as long game, calls for public prediction data
Before vehicles and other machines fully take over driving, a Lyft engineer says there must be more work on prediction data sets. (Getty Images)

The auto electronics supply chain has been affected by COVID-19 and the crushing downturn in car sales, but there are indications that R&D for AI for future self-driving and assisted vehicles remains intact.

And, according to at least one expert a Lyft, plenty of research remains into some fundamental questions, including how self-driving machines can or should make predictions.

“Generally speaking, developing autonomous vehicles is a long-term game, a question of years, or even decades,” said Vladimir Iglovikov, senior computer vision engineer at Lyft. “The pandemic on this time scale is a short-term event. It does not affect the long-term roadmap.”

Advances in AI and Machine Learning are adding an unprecedented level of intelligence to everything through capabilities such as speech processing and image & facial recognition. An essential event for design engineers and AI professionals, Engineering AI sessions during Fierce AI Week explore some of the most innovative real-world applications today, the technological advances that are accelerating adoption of AI and Machine Learning, and what the future holds for this game-changing technology.

Iglovikov wouldn’t discuss particular plans at Lyft for software or hardware AV development, but did offer comments to Fierce Electronics about the general sector comprised of multiple companies undergoing research and testing for Advanced Driver Assistance Systems and autonomous driving, including various plans for robo-taxis and more. He works in the Level 5 (the highest level in the SAE scale) self-driving division at Lyft and holds a PhD in physics from UC Davis. He has held prior positions as a data scientist and speaks to self-driving research largely from that data science perspective.

“For sure, COVID makes the demand for L5 greater,” he said via email. “Fewer people interact in person with fewer chances for the disease to spread. “

He conceded It’s possible that quarantines and other limitations on movement and economies will be more than temporary and will represent a “new reality,” which would heavily impact AV roadmaps at all the companies involved in AV work. “Hopefully, that will not be the case,” he said.

“It is unclear how everything will turn out with the pandemic. It is possible that people will commute less or that quarantines will have a seasonal nature, say every winter,” he added.

Some AV developers have suggested that now is the time to run road tests for future AV vehicles amid an economic lull instead of pulling back as some government leaders have argued. “I believe that AV testing should be as close to the real scenario as possible,” Iglovikov said. “It would be better to test AVs with cars, pedestrians, bicyclists on the streets.”

There are still fundamental research questions for self-driving that go beyond efficient sensors or faster accelerator chips. These questions go to the very nature of how predictions are made by the human brain, and specifically, the human driver, and how that knowledge can be applied to machines.

“There are plenty of problem in self-driving everywhere, but I believe the largest blocker is that we, as a total humanity, do not know how to solve the problem of prediction. V2X is an interesting technology, but it does not really help to solve problems that the autonomous industry is facing,” he said.

“For prediction, you know the map, you know how every car and pedestrian was moving in the last N seconds and you need to predict how will they move in the next M seconds. The largest blocker lies in the research plane. We need more good public datasets and a lot of researchers focusing on the topic.”

For the more immediate future, a few analysts are asking semiconductors makers like Nvidia and Intel about their plans to continue research on future chips that will enhance AI work mainly through acceleration of compute speeds. Research firm Omdia recently surveyed about 30 companies about the impact of COVID-19 on auto electronics, as a follow-up to a similar survey in April.

Omdia’s April survey of 46 auto integrated circuit suppliers found that nearly 65% were expecting a delay in technology deployments in upcoming product launches. Final results of the latest survey are still being tallied.

“It seems as if L4 (high automation) is being impacted more than L2+ (partial automation or ADAS) in terms of development,” said Phil Amsrud, an Omdia analyst in advance of the full survey results. “That was already happening before COVID-19 based on the message from CES in January and before the impact of COVID-19 in 2020.”

“I suspect there will be fewer resources available in general in 2021 and 2022, so the industry will reprioritize everything,” he added. “I did talk to one semiconductor supply who summed it up as, “Those who believe AV is a key to their success will still see it that way, so one is going to abandon it.”

Vladimir Iglovikov will appear with other experts on a Fierce AI Week panel on Wednesday August 12 at 11:30 a.m. EST. The virtual event is free and starts Monday. A full schedule and registration are online.

Source: www.fierceelectronics.com

What do you think?

486 points
Upvote Downvote

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Aurora’s AVs Keep On Trucking Through Pandemic

Aurora’s AVs Keep On Trucking Through Pandemic

Tata's JLR developing anti-nausea software for AVs

Tata’s JLR developing anti-nausea software for AVs