Lidar Company Velodyne Debuts $100 Auto Safety Sensor

https://www.forbes.com/sites/alanohnsman/2020/01/07/lidar-pioneer-velodyne-debuts-100-auto-safety-sensor-as-self-driving-cars-pace-to-market-slows/#1483897e6cbc

by finphil • 7 years ago

103 81 7 years ago

forbes.com

rtkwe2• 7 years ago

It's really nice to see LIDAR getting cheap I've wanted to play around with them for ages but the buy-in for anything other than cheap 2d units has always been way too high for me to justify. I do wish there were at least a few specs in this though.

Animats2• 7 years ago

Yes, there's nothing about the resolution. You can get single-beam scanners for about that price now. If it's only a few beams, it's not a big deal. You can get a small number of beams automotive LIDAR from Continental right now.

If you took a single-beam LIDAR and reflected it off a reflective prism where each face is at a slightly different angle, you'd have a somewhat slow multi-beam scanner with only one emitter and detector. I looked into doing that for the DARPA Grand Challenge, We were thinking of a mod to a rotating scanner used in laser printers to scan the page. There are optical companies that can cut and polish custom prisms. We were too small a team to build our own scanner, though.

KaiserPro2• 7 years ago

Yeah I notice that they are not showing a sample pointcloud.

cr0sh• 7 years ago

Another way to do it is to mount a 2D LIDAR to scan in the vertical plane, and then pan that around. A 180 degree pan will net you a dome/spherical volume (although in a slow manner).

You could do this easily with cheap parallax sensors (like you find on Neato vacuum cleaners, and which are now sold as separate units - RPLidar and such). This has also been done using SICK coffeepot-style 2D LIDAR units (such rigs tend to be large and heavy, tho).

Something I've often considered is the idea of a "stochastic" scanner - that is, don't worry about strict angles, just let the sensor scan at random angles, and note the angle and reading. Over time you'd build up a complete scan. Just an idea I've rolled around in my head; the idea was to eliminate the need for syncing and timing of the scan hardware (2D or 3D), at the price of not necessarily getting a complete perfect scan all the time. It was something of a thought experiment I had while thinking up ways to DIY a LIDAR sensor in a very cheap manner, beyond what has already been done.

Accujack• 7 years ago

Do you happen to know if there's a LIDAR system (I'm aware that many of them just use LEDs) that operates at UV frequencies near the transition to visible?

I'm looking for a LIDAR that can provide basic terrain and obstacle mapping underwater.

Someone1234• 7 years ago

So Lidar is getting cheaper, but the advantage of Lidar (Vs. optical) may be diminishing.

Lidar worked really well because it was computationally impractical to process visual images in near real-time onboard. Lidar simplifies the informational inflow, which reduces computational cost. But the computational cost and practicality of processing optical data onboard has changed radically, it is now practical and affordable (both due to substantially improved software and new/cheaper hardware).

Keep in mind humans use "optical sensors with parallax" (i.e. our eyeballs). Cars are already optimised around that assumption (e.g. headlines to improve visibility in the visible light spectrum). Lidar still has advantages, but also a bunch of drawbacks (like dispersion and sunlight disruption), optical sensors do too (e.g. glare) but they're a lot more intuitive for humans because we share them.

I guess what I am saying is: Are you betting on better sensor tech (Lidar) or better computational tech (Optical)? I think Lidar will hit a ceiling after the "easy wins" have been consumed (and we're approaching that point), whereas optical has no real ceiling (even over and above human's innate abilities). With optical you can understand the world as humans visibly see it, Lidar sees the world fundamentally differently, seeing both less (bad) and more (good).

Obviously it is a somewhat false choice, but if people are investing dollars into development of both techs it is a choice that matters.

Infinitesimus• 7 years ago

... optical sensors do too (e.g. glare) but they're a lot more intuitive for humans because we share them

That's an important point. The long term safer option will be to use both types of sensors before together, that can provide a more holistic picture of the scene around the car (much like we use radar for adaptive cruise control which would be much harder with just cameras ... I think).

I hope we keep investing in making all these sensors cheaper and good enough for 30k cars to start packing them

UI_at_80x24• 7 years ago

Jumping on the 'agree' bandwagon. >The long term safer option will be to use both types of sensors before together

This 100% is the real target but I'd like to add to it.

Camera-based systems rely on ambient photons arranged in a pre-determinded configuration hitting a sensor. Lidar-based systems rely on reflected photons hitting a sensor.

What I would prefer to see is a LiDAR/RADAR combination. When both systems rely on "photons" you suffer the same inherited weaknesses of both. i.e. fog

But split it up and you can be SAFER.

KaiserPro2• 7 years ago

If we want to be safer, then we need to blend sensors.

We do this as humans. We use our ears for sound and inertial detection.

Lidar, Radar, colour and near infrared optical all have a place in autonomous driving.

I dont think we are really anywhere near optical working reliably as a single source. I am immediately suspicious of anyone who suggests otherwise, because they either believe to much in the state of AI, or haven't thought enough about life critical systems to be let near anything autonomous. Worse still, they may be out to make a quick buck.

The thing that lidar has is speed and accuracy. 100hz update rates, and 1/100th second latency is not unthinkable. With optical systems you're lucky to get 25hz with sub 100ms latency. Thats just for depth estimation using semantics, you still need to feed it into your driving model.

Lidar works way better at night, it works a tonne better with unknown objects. However, like optical its shit in rain/snow. Hence why decent resolution radar will be needed as well.

ganeshkrishnan2• 7 years ago

If we want to be safer, then we need to blend sensors.

isnt that what a kalman filter does? I thought most autonomous cars have sensor fusion to blend all sensors

lmm3• 7 years ago

Humans mostly don't use their ears when driving cars (to the frustration of the rest of us using the road).

lightedman2• 7 years ago

Radar is pretty garbage in rain and snow as well.

https://www.weather.gov/media/publications/front/14dec-front...

My father was a missile radar guidance technician for a living. He says radar on a car makes almost no sense, especially for navigating in inclement weather, due to how radar tends to reflect off of anything to some degree.

sllabres• 7 years ago

When I'll drive down a motorway I doubt that I use sound or inertial for planing my further track. I presume that one could drive 'remote' with only visual, if it would be the same quality as when one was in the vehicle.

andbberger• 7 years ago

It's not just the computational cost of real-time visual inference, there are some fundamental hurdles to be overcome we've attained truly human level computer vision.

It's hard to imagine a pure vision system ever really matching the performance of a fused sensor system.

Why infer when you can measure?

jmole• 7 years ago

we use optical sensors with parallax, yes, but they also have 6-dof motion capabilities. Cameras on a car are stuck in one place. Take some time and notice how often your head moves while driving. Even just the small motion of our eyes (rotation during fixation and tracking) adds spatial information to the visual stream that cameras simply can't achieve.

A lot of the parallax information we infer is not because we have two eyes, but because we're able to move our eyes/head around, and "fill in the gaps" in a bayesian sort of way.

lazyjones• 7 years ago

Cameras on a car are stuck in one place. Take some time and notice how often your head moves while driving.

That's why multiple cameras are used and positioned farther apart than you could move your head while inside the car.

m4633• 7 years ago

Except when stopped, car visual systems can do parallax just like pigeons do (by moving the camera forwards).

SomewhatLikely• 7 years ago

Cars can be driven remotely using only cameras, so the motion is likely not very critical. https://youtu.be/9sgetWQGYxY

threeseed4• 7 years ago

Trying to emulate human eyesight in self driving cars is hilariously misguided.

1. We have a computer behind our eyes so advanced that we may never be able to come close to replicating it. It is capable of identifying, tracking and predicting the behaviour of multiple objects in real-time even in reduced visibility and can infer new objects without training e.g. a green firetruck or a RV with a satellite dish.

2. Our eyes are connected to a very adaptive and movable object i.e. our head. In order to perceive depth and identity objects e.g. an actual person versus a photo of a person we continuously move our head around in multiple dimensions. A car can't do this.

I would refer everyone to the countless examples of Tesla's Autopilot recognising humans in bus signs, sides of trucks etc and attempting to do auto-avoidance. That is an unsolvable problem with only optical cameras.

ebg132• 7 years ago

continuously move our head around in multiple dimensions. A car can't do this

Doesn't isn't can't. There's no reason why car-mounted hardware can't move just as much as eyes do.

Dylan16807• 7 years ago

Head movement is a useful trick but it's a pretty minor improvement. Just having a few cameras with good separation will beat it.

NotSammyHagar• 7 years ago

That was an interesting observation, that lidar might actually be unnecessary by use of visual image processing. I always thought Tesla made a terrible choice in precluding (expensive) lidar and said if humans can drive safely without lidar, just using visual processing, so should cars (eventually). Surprising but I guess it may turn out that way.

trianglem• 7 years ago

I have serious doubts visual processing under all conditions could ever be better than a combination of both sensors.

mrfusion• 7 years ago

Maybe lidar could be great for training optical?

m4633• 7 years ago

This will only augment current optical technology, not replace it.

choppaface• 7 years ago

Lidar worked well (and continues to do so) because it’s the most accurate proximity sensor today for cars. If you don’t want your car to hit anything, you need Lidar. Radar and vision-based stereo are useful too, but 10%+ distance error and overall false negatives make that combo pretty unsafe (especially off the freeway). Computational cost is a far smaller factor than safety in the value of a demo.

Cheap lidar (and probably radar) will earn a permanent place on ADAS of the future if only because it works much better than vision at night. You’ll see them together because they have complimentary failure modes and the denser lidar point cloud is worth a $100 add-on.

While many of the autonomy efforts are still struggling with basic detection, the major challenge facing the leaders is in forecasting future motion, especially complex motion. The new cheap sensors are a boon but not a game changer.

lootsauce• 7 years ago

how does one buy this?

ThrustVectoring2• 7 years ago

From Velodyne, presumably, and their website says it will be available mid-2020

ycombonator• 7 years ago

My 2021 prediction: And the Chinese already got the specs and should start selling $5 units in a year.

finphil• 7 years ago

You beat me to it, but I was going for $50 ;)

brixon• 7 years ago

It will be $5 in China and $50 in the US.

sanguy• 7 years ago

Very cool to see this price point. If Velodyne really can manufacture and sell it in volume and be profitable it's an amazing evolution for Velodyne.

If it's a loss leader to buy attention and stay relevant versus their fast moving competitors it will become obvious eventually.

ape4• 7 years ago

Its solid-state. Old Lidar needed an expensive revolving sensor

jklinger410• 7 years ago

Would be really cool to retro-fit these on an older vehicle. Replace the center console with a tablet, some OS handles the lidar.

herogreen• 7 years ago

Better link: https://velodynelidar.com/press-release/velodyne-lidar-intro... states "60-degree horizontal FoV x 10-degree vertical FoV" I wonder whats the resoltion though. Judging by the name "velabit" it could be low.

m4633• 7 years ago

It wouldn't have to be too much to overcome the resolution of current front-facing radar and ultrasonic sensors, which are 1x1 resolution ;)

stevehawk• 7 years ago

I'm not a big hardware nerd, how do this and that $16,000 one from earlier this week compare? They're both roof top, spinning sensors it seems but at very different price points. Is this one not as far distance as the other?

mdorazio• 7 years ago

LIDAR units have different effective ranges and, most importantly, number of points they capture. These cheap units generally have a small number of points. Looks like the Velabit has a range of about 100 meters, but FOV is 60 degrees horizontal by 10 degrees vertical, so you would need to mount several of these around a vehicle to get the equivalent coverage of a 360 degree unit. The press releases don't say how many points are in its point cloud, but I'm guessing it's probably 1/10th of a larger unit's.

robocat• 7 years ago

To add to above comment: comparison of some different models for Ouster (a competitor):

https://ouster.com/blog/128-channel-lidar-sensors-long-range...

Discussion about above:

https://news.ycombinator.com/item?id=21970796

ThrustVectoring2• 7 years ago

It'd be pretty cool to use this as part of an assistance tool for people with visual impairments. Convert the positional data to something you can learn to parse with your ears, and people could get a better sense of obstacles around them. I'm really not sure what the right conversion strategy or user interface is, but it seems like the sort of thing that'd be relatively straightforward.

ipsum2• 7 years ago

Microsoft has something similar: https://www.microsoft.com/en-us/ai/seeing-ai

Lidar Company Velodyne Debuts $100 Auto Safety Sensor

Related Stories