Hacker News new | past | comments | ask | show | jobs | submit login

Range sensors that actually work reasonably well outdoors are neither cheap nor ubiquitous. Note that we're talking about imaging range sensors here, not single point measurements.

Time of flight cameras? They predominantly use NIR (900nm or circa) and so they're rubbish in sunlight. Try using your kinect outdoors. Beyond the kinect they're also very expensive, a few thousand each. The kinect 1 also fails outdoors and when the range is > 5m or so.

LIDAR? Well sure, you can buy a rotating line scanner - 100-180deg field of view typical. They're very nice, but also (new) you're looking at around $3-5k. Don't even dream about getting a Velodyne system, like the Google car uses.

Stereo? Yeah... well stereo is nice, but it takes a lot of computational grunt to do it in real time, reliably and in varying illuminations (good luck using it at night).

Thermal? We're getting there, FLIR make some fairly low cost, low resolution sensors, but still you're looking at $500-1000 a pop in OEM quantities.

RADAR/Ultrasonics are reasonably cheap nowadays and they're pretty good for coarse measurements like when you need to know how far you are from the wall in your garage. We're getting to the point where RADAR is good enough for collision avoidance (i.e. auto-braking) on high speed roads.

More cameras are cool, but you need to know how to process the data and present it clearly to the user. I can see the advantage of having a camera for a rear-view mirror particularly for vehicles like vans with no rear windows. Side mirrors are still useful and for most people the cost of installing multiple high resolution video cameras and monitors isn't worth the savings in petrol from drag reduction. If you replace a mirror with a camera you also need a very high resolution display, a wide angle camera that doesn't distort too much and a way of viewing it in any conceivable illumination condition. Mirrors are also nice and mechanical, unless you crash they rarely ever break.




> More cameras are cool, but you need to know how to process the data and present it clearly to the user.

Thanks for the state of the art, good to know.

I assumed the relevant sensor tech was cheap and ubiquitous because every time I rent a car these days, it seems to have both a backup camera and radar. Maybe that's all we've got though, and maybe tricky to point that gear in every direction and have it always on, and as you mention, usefully conveyed to a driver is the big hurdle.

Though I have one crazy idea for that, thanks to George Lucas.

Some people pick on this one scene in Star Wars, when Han and Luke are fighting in the turrets, and you hear all these ships whizzing past, even though there's no sound in space. In one of the more contrived retcons, some fans insist that the sounds of whizzing ships are actually just a different approach to HCI. The ship detects threats, then pumps in naturalish-sounding audio telling you where they are in relation to the Falcon.

I know that's not a mass market idea, but my dream car would use fancy radar to make nearby cars sounds a bit like starfighters. (So long as I can turn it off, probably gets old fast.)

So yes, my primary interest in sensor tech is not reducing drag, or even accidents, but in feeling like I'm piloting a spaceship. Don't judge. :)


> Thermal? We're getting there, FLIR make some fairly low cost, low resolution sensors, but still you're looking at $500-1000 a pop in OEM quantities.

From what I've heard, the FLIR lepton core might be around $~250 in quantities, which could make it affordable as an add-on module.


Yep that's true actually, I forgot about the Lepton. Mike's electricstuff did a teardown/hack of it recently - http://www.electricstuff.co.uk/lepton.html.




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

Search: