Slashdot Mirror
Ex-Apple Engineers Unveil a Next-Generation Sensor For Self-Driving Cars (theverge.com)
An anonymous reader quotes a report from The Verge: Aeva, a Mountain View, California-based startup founded only just last year, has built what its two-cofounders claim is a next-generation version of LIDAR, the 3D mapping technology that has become instrumental for how self-driving cars measure the distance of objects and effectively see the road in front of them. And today, the company is officially unveiling its product, a tiny box that can more directly measure objects in a given scene and the distance and velocity of those objects relative to one another.
Aeva's technology is able to separate objects based on distance and whether the object is moving away from or toward it. It's also able to measure the velocity of the object, which enables the software to predict where cars and pedestrians are going. The company even says its sensing system is capable of completely shutting out interference from other, similar sensors -- including those from other companies -- and operating in all weather conditions and in the dark, thanks to a reflectivity sensor. Not only is Aeva's version of LIDAR superior to the variety found in most self-driving test vehicles on the road today, the company says, but the lightweight, low-power box it's housed in also contains all the other types of sensors and cameras necessary for an autonomous vehicle to see and make sense of every component within its field of vision. Aeva's new system sounds a lot more promising when you consider the company's co-founders, Soroush Salehian and his business partner Mina Rezk, are former Apple engineers who both worked on Apple's "Special Projects" team. Although they will not say so, they likely helped progress the company's secretive autonomous car division. The Verge notes that Salehian also "worked on developing the first Apple Watch and the iPhone 6, while Rezk is a veteran of Nikon where he worked on optical hardware."
Aeva's technology is able to separate objects based on distance and whether the object is moving away from or toward it. It's also able to measure the velocity of the object, which enables the software to predict where cars and pedestrians are going. The company even says its sensing system is capable of completely shutting out interference from other, similar sensors -- including those from other companies -- and operating in all weather conditions and in the dark, thanks to a reflectivity sensor. Not only is Aeva's version of LIDAR superior to the variety found in most self-driving test vehicles on the road today, the company says, but the lightweight, low-power box it's housed in also contains all the other types of sensors and cameras necessary for an autonomous vehicle to see and make sense of every component within its field of vision. Aeva's new system sounds a lot more promising when you consider the company's co-founders, Soroush Salehian and his business partner Mina Rezk, are former Apple engineers who both worked on Apple's "Special Projects" team. Although they will not say so, they likely helped progress the company's secretive autonomous car division. The Verge notes that Salehian also "worked on developing the first Apple Watch and the iPhone 6, while Rezk is a veteran of Nikon where he worked on optical hardware."
With the current trends I just read the title as "Next-Generation Censor".
What are we supposed to make of this credential? Ex Apple engineer makes new sweatshirt, Ex Google engineer makes new ice cream scoop.
If you're fired from Silicon Valley, is the next move to make anything, advertise it online and describe as having been invented by an ex big-tech engineer?
Apple people do not have some magic power to invent better stuff. I have to think much of these sensors will more likely end up in something less then self driving vehicles. All of them will have to obtain a squeaky clean test record before anyone begins to trust this technology.
LIDAR is not a new invention. What would differentiate this device from other LIDAR devices already on the market would be cost. If this is substantially cheaper than other LIDAR implementations, it is relevant. If it is the same price, then nobody cares. There are two LIDAR arrays promised under $50 and one under $100.
What is the cost per device anticipated to be? That is literally the only relevant factor.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
If only there was a material or a location sensors could sit in that protect other sensors that don't like getting wet or dirty..
I just think lidar is misguided. It's an easy way to get "part of the way there", but no further. It has the same weaknesses as your visual cameras; why would you run your sensors in the same frequency spectrum and subject to the same weaknesses (such as weather)?
The big thing LIDAR has over radar is resolution (particularly vs. lower frequency radars). But... well, "and"? You still need visual processing from cameras to identify the object either way, so how critical is the resolution on your ranging system anyway? You just need a return, you don't need it to identify the return for you. There are various other advantages and disadvantages to each, in that they can each be "tricked" in various ways and some things are very visible to one but not to the other (e.g. small chunks of jagged metal in the road = very visible to radar, not lidar; moderate-sized chunks of plywood or fibreglass = very visible to lidar, not radar). But if you have your cameras working on the visible spectrum, you really want some other spectrum giving some other insight. Radar also has the advantage of being simply-steered (via a phased array), rather than requiring a rotating dome.
And there's real potential for other insights for radar, beyond just seeing metal debris very well, seeing through weather, in some cases being able to see past the car ahead of you, etc. For example, radar can be used to measure the texture of a surface, on a scale proportional to the wavelength of the beam (the rougher the surface, the more the backscatter). So depending on what frequency you probe with, you could be measuring anything from potholes to ice. Seeing where a shoulder looks weak, seeing where something might be slick, etc, etc - important things in terms of deciding how to drive, and a sense beyond that of humans. No systems (AFAIK) do this today, but the potential is there, and it's some serious potential.
Also, the resolution issue can be (to some extent) overcome. The minimum resolution you can see is proportional to the radar's aperture. If automakers wanted to, they could dramatically increase the effective horizontal resolution by having radar antennas on opposite sides of the front of the car (or the rear, for rear-facing radar). Now you have a virtual antenna the width of the entire vehicle.
I mean, if LIDAR can be implemented cheaply enough, with minimal power consumption, and it's just a low-cost, no-drag-added addition to your cameras... sure. But that's not even close to the description of LIDAR at present. Maybe time-of-flight cameras would be a suitable replacement.
"Close the door! What, were you born in a barn?" -- Police chief, "Jesus Christ Supercop"
Aeva is an interesting company, but one of 50+ new entrants in the automotive LiDAR in recent years.
Fusing camera with LiDAR? Aeye has been doing this for a long time.
Coherent LiDAR detection that measures velocity as well as distance? Blackmore and Strobe (and others) do this too.
Marketing departments need to do what marketing departments do, I guess.
It appears that the Aeva device moves the calibration, computation, and sensor fusion into one black box that includes other sensors. The advantage is reduced risk and domain expertise for the end system integrators (car manufacturers). LIDAR is optical; optical sensors do not work well in fog; the Aeva box has other sensors too. With Aeva in control of the sensor fusion, there may be less risk of programming error in the sensor fusion from car manufacturers that don't have the incentive to spend time on sensor fusion code. I can see this being a repair advantage too -- when a sensor breaks, swap out the entire Aeva black box for a factory-calibrated Aeva black box, instead of installing a new sensor and having to recalibrate the entire car. Some of the claims are weak -- "we can tell whether an object is moving away or towards the car" really means they compare the LIDAR returns to current vehicle speed interpreted by the INS (GPS + many-axis accelerometer).
Sensors will never "self clean" the way your eyeballs do when you blink.
Automobiles will never replace the horse and buggy.
"That's the way to do it" - Punch
Put them behind the windshield then.
That's what I've been saying for years. Self-driving cars (en masse) will not happen in our lifetimes.
what about "sensor fusion" between radar, lidar, and FLIR?
That way the system can have a better picture of what's going on.
There are literally tens of thousands of those. There are so many it's commodity. For some reason it count as streetcreds to be ex Apple engineers?
I would think it means something if talking about ex-CEOs..... .... Actually forget it... one is dead and the other one still alive is a loser and general douchebag.....ok he dead one is a douchebag too, but not a loser.
Seems like any optical system will stop working the first time a bug, leaf, rain drop or snow hits it's lens and sticks. You can coat the lens to make it easier for the wind to blow them off, but in general the optics could be instantly obscured at any time, which basically means the car would have to stop immediately.
You say that but even having sensors inside the wind screen would not protect them. All that crap you have to clean off the inside of your windows and wind screen are oils and polymers exuding from the plastics breaking down inside the passenger cabin, combined with all the crap you breathe out and sweat when you're in the vehicle.
http://www.pitt.edu/~drn16/conferencePaper.pdf
This is the theory paper on the tech. and as further proof, here is AEVA's hiring site looking for an FPGA engineer.
https://jobs.lever.co/aeva/84cdf11c-0f55-4181-be5d-3e682cb4b314
your welcome all.
Put them behind the windshield then.
Specifically, between the rear view mirror and the windshield, where it won't affect the driver in any way. Or give it its own wiper and washer, like euro cars do with headlights. Either way it's a trivial problem in the grand scheme.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"