LIDAR industry hits impasse

[diplomat33]

According to this article, LIDAR companies are struggling to find buyers:
"The technical question is whether to wait 10 or more years for computer vision research to solve the ranging problem or for cheap LIDARs to hit the market. However, since the business mindset is to focus on short-term feasible technologies, the lack of serious buyers is the real problem for the LIDAR industry."
https://syncedreview.com/2019/09/14/lidar-industry-hits-impasse-was-elon-musk-right-after-all/

 

[Daniel in SD]

Not surprising. All Level 3-5 vehicles are prototypes, so that's a pretty limited market. It doesn't seem like LiDAR will ever be cost effective for Level 2 systems. If you can get driver assist systems to work 99% of the time with computer vision there's no reason to use LiDAR.
It seems to me that LiDAR only becomes useful in Level 3-5 systems where you're trying to eliminate that last 1% of errors.

 

[Eno Deb]

Yep. I think they, just like much of the rest of the industry, have vastly underestimated the time it will take to get autonomous cars working reliably enough for a mass market. As long as the fundamental software issues aren't solved, there will be no big demand for advanced sensors and other components that are only required for autonomous cars.

 

[diplomat33]

It also does not help, that according to the article, some of the start ups were not producing good quality LIDAR.

 

[MichaelP90DL]

Yep. I think they, just like much of the rest of the industry, have vastly underestimated the time it will take to get autonomous cars working reliably enough for a mass market. As long as the fundamental software issues aren't solved, there will be no big demand for advanced sensors and other components that are only required for autonomous cars.

Seriously, a lot of very weird stuff happens on streets, in alleys, on highways. I will be literally astounded if Tesla finds a way to handle virtually everything. The consequences for screwing up can be enormous.

 

[malcolm]

But what about all of the systems that Mobileye was meant to be installing with LIDAR (because redundancy is important)

Obviously fewer than they would have us believe.

 

[Daniel in SD]

But what about all of the systems that Mobileye was meant to be installing with LIDAR (because redundancy is important)

Obviously fewer than they would have us believe.

Same issue. Their plan is to only use LiDAR and RADAR in Level 4 vehicles. They've encountered the same problem that everyone else has, they haven't figured out how to make a Level 4 vehicle.

 

[electronblue]

According to this article, LIDAR companies are struggling to find buyers:
"The technical question is whether to wait 10 or more years for computer vision research to solve the ranging problem or for cheap LIDARs to hit the market. However, since the business mindset is to focus on short-term feasible technologies, the lack of serious buyers is the real problem for the LIDAR industry."
LiDAR Industry Hits Impasse: Was Elon Musk Right After All?

But what about all of the systems that Mobileye was meant to be installing with LIDAR (because redundancy is important)

Obviously fewer than they would have us believe.

Realistically, who was this industry (or article writer) expecting to be ”there” at this point? I have not heard of anyone other than Audi planning on putting Lidar into production vehicles at any schedule that isn’t years into the future.

Same, I haven’t heard of anyone other than Audi and Tesla talk about putting Level 3 and above (Level 3 for Audi and Level 5 for Tesla) within the 2018-2020 time period.

Tesla does not use Lidar and everyone else is working on prototypes and prototype fleets of Level 4 vehicles that are bound to be small in number.

Audi does ship Lidar in volume for its current Level 2 offering, though, but that doesn’t help more than one supplier I guess. And that’s only because they plan on using the same suite for Level 3. There is much less reason for Lidar on Level 2 if you don’t need ”perfect” collision detection. Other companies are aiming at Level 2 and the PR Level 2+ for the time being so I don’t see many using Lidar for that.

 

[electronblue]

Same issue. Their plan is to only use LiDAR and RADAR in Level 4 vehicles. They've encountered the same problem that everyone else has, they haven't figured out how to make a Level 4 vehicle.

MobilEye has not set any clear timeline or partner plan on when and how their Level 4 prototypes might be in volume production, similar to everyone else in the industry for whom this is still a reseach project.

Only companies that have current public timelines or plans for Level 3+ and above are Audi (Level 3 in Germany) and Tesla (Level 5 no geofence). Can anyone think of others?

Small Level 4 robotaxi fleets are not volume operations and usually part of those research projects. These of course exist in many places, MobilEye in Israel being one, Waymo in the U.S. another and more elsewhere, but small research fleets don’t a market make.

 

[electronblue]

Some point from the article:

Although Light Detection and Ranging has proven its reliability and accurate ranging capabilities, the hardware’s size and cost of some US$10,000 per unit are huge obstacles to its mass commercialization.

Complete nonsense. Lidar does not cost 10k anymore, nowhere near for the products relevant here — let alone in volume anyway.

In order to survive, LiDAR companies need to either secure orders from L4 autopilot teams or integrate with auto manufacturers’ supply chains.

Probably so, but misleading. The real problem isn’t integrating into those supply chains, it is the supply chains simply don’t exist. Nobody has L4 supply chains because nobody has L4 volume products or plans for them in the near future.

Top tier auto parts suppliers such as Bosch, Aptiv, ZFAuto part manufacturers have not exactly embraced the commercialization of L3/L4 self-driving tech such as pricey LiDARs.

Again misleading. It is completely true that auto parts manufacturers — just like the timid auto manufacturers — have not embraced the commercializiation of L3/L4 (much discussed on this forum too and rightfully derided by people like @Bladerskb).

The problem is: this has got nothing to do with not embracing Lidar. In fact almost everyone but Tesla has embraced Lidar for L3/L4 commercialization... the problem is they have not embraced the actual commercialization of L3/L4 products, which would then use those Lidars. No volume products = no parts needed for them in volume.

The unsolved or research problem with these self-driving projects is not Lidar. It is the self-driving code and networks. As long as that is not a commercial volume product, the hardware parts are not needed.

A startup rep told us “car manufacturers are considering many promising sensors for mass production targeting the year 2021… unfortunately, LiDAR is not on the list.”

Yep, I don’t expect widespread adoption of Lidar for Level 2/2+ cars. As long as the year 2021 plans are just Level 2/2+ cars, why would Lidar be on the list?

At this spring’s Tesla Autonomy Investor Day event, Elon Musk once again dissed the “fools” relying on LiDAR for self-driving vehicle development

Turns out, this was just clickbait. The article does not analyze the relevance of Elon Musk’s comments at all. But let me: they are not relevant. Companies have not turned their back on Lidar at all as a technology. They just don’t need it until they have volume Level 3 and above products to sell — and that indeed is something which seems to keep getting delayed and delayed... and with it the need for Lidar.

Now, where Elon has a chance to prove everyone wrong and himself correct, is if Tesla actually delivers on Level 5 no geofence robotaxi feature complete in 2019 and in operation in 2020. That could reset some of the above expectations. But until then, Tesla’s is one more research project amongst many.

Once autonomous (Level 3 and above) is being shipped in volume, we shall really see what technologies it is using and who was ”right”.

 

[voyager]

Regardless of the vehicle autonomy system or principle, I'd say that it is better to ‘reformat’ cars before any autonomous tech goes in.
1. The less space they require, the more margin there is to evade other road users.
2. The easier the auto-pilot’s task of overseeing the vehicle and its surroundings, particularly if it's shaped to avoid the dreaded blind spots.
3. If necessary, the human driver can act as a fail-safe; great outside view will boost involvement as well.

 

[malcolm]

How lidar makers are coping with slow progress of self-driving tech

The classic Velodyne lidar had 64 individually packaged lasers inside, each paired with an individual light sensor to detect return flashes. This complex design contributed to the units' high price. By contrast, Ouster uses semiconductor-based technology to pack 64 lasers onto a single chip, with 64 detectors packed on a second chip. This reduces the cost of Ouster's lidar in much the same way that the microchip revolution enabled the creation of cheap personal computers in the 1970s.

The result: in 2018 Ouster was able to offer its first lidar sensor, a 64-laser unit called the OS-1, for just $12,000. That was dramatically cheaper than Velodyne was charging for its 64-laser models at the time. Since then, Ouster has expanded its product line. The company's products now range from a low-end 16-laser unit for $3,500 to a long-range unit with 128 lasers for $24,000.

In recent years, there has been no shortage of hype about the potential of solid-state lidar—lidar that's fixed in place rather than spinning 360 degrees. There are dozens of companies working on lidar sensors, and most of them are solid-state designs. Yet they don't seem to be getting much use in the real world—at least not in the commercial robotics market. Williams told Ars that "we haven't come across solid state in the wild yet."

That's partly because a number of solid-state companies haven't actually started shipping products to the general public yet. It's also partly because some lidar companies are focused on large-volume sales to carmakers, not one-off sales to smaller robotics companies.

But it's also because spinning lidar has some unique advantages. Most obviously, a single spinning lidar sensor offers 360-degree coverage around a vehicle. To achieve similar coverage with a fixed lidar system you need multiple sensors distributed around the vehicle. That means higher costs and power consumption

 

[kaffine]

There are other industries that are starting to use LIDAR. The industrial safety industry has been using them for some time now as part of interlock systems. Use LIDAR to setup safe zones around industrial equipment to shut it down if anyone gets close.

The security industry is starting to use the same setup as an improved volumetric detection system. It is also used in access control system to verify no piggybacking. Granted it isn't as foolproof as a weight sensor in a man trap for ultra high security, it is getting used in high security areas.