Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Am surprised no one yet cited Elon saying "this is fool's errand" because of price, complexity, etc. I think LIDAR will improve significantly the safety of vehicles (autonomous or not) because it can definitely see better and deeper, even in damning weather conditions (rain, fog, haze).

Alas, the price of these devices has to come down at least one order of magnitude. Maybe even two. Still, I am really thankful that other companies (since Tesla has no interest in it) are considering and further developing LIDAR.



Lidar is solidly worse than the human eye in dust, rain etc. Waymo has struggled w/ dust devils in AZ. Cruise has struggled w/ steam vents in SF. I work at a John Deere subsidiary and we are quite interested in dust performance for field work. From our tests Lidar is low on the list for seeing through small particles.

Lots of things could fix this, less beam divergence, custom signals processing on multiple returns. But out of the box, it’s this statement does not hold true.


I work at an industrial plant we use microwave radar based imaging can get quite detailed surface profiles in very poor conditions including inside reactors etc where dust is big issue. I'm not expert in this field I think systems used are continuous wave based.

For vehicle applications specifically probably worth looking into what they use on autonomous vehicles at mine sites imagine that tech probably useful in agriculture. For example Pilbara here in Australia large autonomous fleets in very dusty conditions.


Hi, I've previously been to a company that makes the hardware for autonomous mining vehicles. They rely mainly on DGPS systems for positioning and a combination of simple camera vision and radar for obstacle detection. Keep in mind they don't drive that fast (around 25 mph), are supposed to be in clear unobstructed (unless they're in a queue to load or unload) and just stop when they detect a vehicle or a person.


There are classes of vehicles where the cost of LIDAR is less of a factor, e.g. freight trucks, taxis, public buses, which get even more economic benefit from Level 5 autonomy and can serve as a stepping stone to ramping up LIDAR production to bring the costs down.

Just because LIDAR doesn't make sense for the Model 3 today, doesn't mean it should be entirely discounted.


As far as I’m aware LIDAR cannot see though fog, as the light is dispersed, and even light rain might reduce range significantly.

Tesla cars have radar which can see through any weather condition and detect transparent surfaces, invisible to LIDAR.


There are LIDAR units that can see through fog. If you get data from multiple returns, not just the first, you can tell the difference between fog, rain, and solid surfaces. "First and last" is a big win. A solid obstacle in fog looks like a repeatable "last" (furthest) return, while rain and fog look like random disconnected points.

I think Google's own unit has 8 stored returns.


> LIDAR cannot see though fog

Simply depends on what wavelength of light you use.

Water-absorbing frequencies are nice because the atmosphere then shields most light, giving you nice SNR from your laser illumination. But better sensors could work around this, using other frequencies that can 'see' through fog.

It's certainly a technological limitation of current systems, but it's not an inherent limitation.


Resolution of normal radar is much lower unless you have a gigant receiver. Also there is a significant delay of 10 to 100ms That's why a combination of lidar and radio radar is desired.


> Also there is a significant delay of 10 to 100ms

I'm curious why there is such an apparently long delay?


It really depends on the type of RADAR being used. If it's an FMCW radar, typically you will get a beat signal whose frequency corresponds to the target range. That frequency will vary with range, and in order to be well resolved you have to observe it for something like 1/period. So that puts a fundamental lower bound on how long you have to integrate. There are lots of tricks to improve things, and there are lots of variants of the standard radar hardware/methods, but I suspect that's what OP was referring to.


That doesn't explain 100ms delays though. Speed of light times 100ms round trip is 15,000 km.


It depends entirely on what configuration of RADAR you've got, and what you're pointing it at. You can build a system that will result in returns with beat frequencies of just about anything, depending on the Tx modulation and the range. The question is whether it will be useful for your application in terms of range/velocity resolution, latency, integration time etc.

Like I said I was just speculating about why OP specifically mentioned 10-100ms. the light does indeed travel pretty quickly (although, as anybody in the radar/lidar industry will tell you, not nearly quickly enough!), however the round trip time is just the minimum latency you have to eat to get any information about your target. Once you have light coming back, you need to integrate for some about of time to achieve your desired SNR. That time could be very small, or it could be infinite if there are no photons coming back. Let's randomly say that you're using a RADAR with a Tx bandwidth situated such that the round trip time is 1us, and that your target range is s.t. the beat frequency of the return is 1kHz. Your job is to estimate that frequency, so you have to observe the waveform (by integrating samples for an FFT, typically) for at least one cycle of the RF wavelength. That would require that you wait 1us for the light to fly, and then wait another 1ms for the RF to cycle once. So your measurement latency is ~1ms. Now that's not 100ms, but perhaps you need more than one cycle to give a good estimate of the frequency, and then even more because the target is faint and there aren't many photons coming back. You could possibly arrive at some much higher number, like 10-100ms.

I'm not sure if that was OP's point, but that's all I'm saying ;-)


Also curious, most of what I’ve read about FMCW radar mentions single-digit millisecond latencies.


Yeah, it definitely can be shorter than 100ms. See my sibling comment. It just depends on the type of radar being used and the target range/velocity. Certainly for shortish range targets and mmwave radars on reasonably reflective targets you can get a signal with decent SNR in shorter time frames.


16k each, and Waymo has 4 on it's cars? I think Elon was right to not care about LIDAR until the price point is right.


Elon was right for Tesla, but that doesn't mean the economics don't work for Waymo, even as they continue to improve.

A $100k fully autonomous taxi would print money over its service life (right now they have remote safety drivers on standby).

The question is how quickly they can expand their currently tiny geofenced area.


And this seems to prove him right with $6k - 16k price range. Far too expensive to be practical. Still the potential is there for it to be a useful component for self driving vehicles, only if the price comes down an order of magnitude.


Ouster CEO here.

This is single unit pricing. Volume discounts apply. Still work to do to get this in every honda civic, but it is possible with our technology.


How much is a chauffeur? Or taxi driver salary?


It's a sensor. Not an AI. The competition is cameras and radar, which are very cheap.


If this sensor puts it over the edge and the rest of the cost isn't high, then the competition is paid drivers.


If that were true, sure. But I don't think it's so much better than other sensors for that to be a plausible scenario. Not my field though, so take it with a grain of salt.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: