5 ms·
I wonder why pick such vehicle to test this. Why not something with smaller mass to have less momentum on impact. (ie google car).
by me_me_me 7y ago
I wonder why pick such vehicle to test this.
Why not something with smaller mass to have less momentum on impact. (ie google car).
- yifanl 7y agoActually yeah, can we not rig some kind of bicycle with the same sensors and test self-driving that way? The steering mechanism would have to be modified obviously, but surely steering is a trivial part of the problem compared to actually figuring out where to steer to?
- rtkwe 7y agoPart of path planning will involve vehicle dynamics, breaking, acceleration and steering response, and the envelope of the vehicle. All of those will heavily impact how a car should drive.
- lovehashbrowns 7y agoBut that's something the AI can learn fairly easily, isn't it? The difficulty in this case wasn't that the AI had issues figuring out the handling of the SUV, it's that it had issues detecting a pedestrian and a dangerous situation. You can still run into these issues on a bicycle, with a much lesser chance of killing people.
- dkonofalski 7y agoThis is totally true but the issue was more with the methodology of the detection rather than the detection itself. Regardless of the type of vehicle, the software wasn't good enough for real-world testing.
- fgvuyg 7y agocycle dynamics is an unsolved control problem
- umanwizard 7y agoMy personal anecdotal belief is that it’s much, much harder for a robot to ride a bike than to drive a car. It requires human-level bodily/physical intuition that AI is nowhere near. We can barely program robots to walk normally, let alone ride a bike.