You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
I challenge any of the autonomous car companies to give Portland, Oregon roads a chance:
Foggy conditions
Rain hard enough that you can barely make out your own hood
Rain at night time in a construction zone
Black ice
Driving at freeway speeds when a car coming at you in the opposite direction hits a sheet of water sending a waterfall onto your windshield, blinding the driver and presumably cameras too
Crossing our steel bridges where there are no painted lines
Avoiding a deer crossing the road
A 1/8" layer of snow on the ground, effectively covering all visible lines.
I challenge any of the autonomous car companies to give Portland, Oregon roads a chance:
Foggy conditions
Rain hard enough that you can barely make out your own hood
Rain at night time in a construction zone
Black ice
Driving at freeway speeds when a car coming at you in the opposite direction hits a sheet of water sending a waterfall onto your windshield, blinding the driver and presumably cameras too
Crossing our steel bridges where there are no painted lines
Challenges indeed, but also in my opinion nothing more than that...
My question is actually if the first version of an autonomous car should already handle all these things or would it be allowed to need manual driving in exceptional conditions ? These harsh conditions can than be tackled in v2 or v3.
Driving at freeway speeds when a car coming at you in the opposite direction hits a sheet of water sending a waterfall onto your windshield, blinding the driver and presumably cameras too
We had a technical term for that once upon a time: "boofing".
Good list Daniel offers. Construction, weather, and lots of random motion (try the elementary school drop off line or the sports event parking lot) seem to be the big problems right now.
The issue is not the car necessarily careening out of control, but rather coming to a screeching indecisive halt in the middle of the road. Note the opening of the story from Bernard's link says the car "refused" to drive into unmarked lanes.
Deer dodging is a sport in this part of TX as well. Multiple herds of 10 or more are common right outside my front door every dawn and dusk. I'm not at all confident a sensor can arrive at a solution other than full stop.
We had a technical term for that once upon a time: "boofing".
Good list Daniel offers. Construction, weather, and lots of random motion (try the elementary school drop off line or the sports event parking lot) seem to be the big problems right now.
The issue is not the car necessarily careening out of control, but rather coming to a screeching indecisive halt in the middle of the road. Note the opening of the story from Bernard's link says the car "refused" to drive into unmarked lanes.
Deer dodging is a sport in this part of TX as well. Multiple herds of 10 or more are common right outside my front door every dawn and dusk. I'm not at all confident a sensor can arrive at a solution other than full stop.
I remember rain in Florida. Not so much rain, more the sky opening up and instantly dropping unimaginable volumes of water, resulting in zero visibility. The only possible driver response is to pull over and stop. I agree with Don - in those circumstances automation could not and should not keep the car moving. You risk hydroplaning, flash floods and unpredictable behavior of other vehicles (which you can no longer see). Of course (in deference to Staf) well-designed automation in these cases may well do a better job than human drivers - more safely get you to the side of the road then stop.
I remember rain in Florida. Not so much rain, more the sky opening up and instantly dropping unimaginable volumes of water, resulting in zero visibility. The only possible driver response is to pull over and stop. I agree with Don - in those circumstances automation could not and should not keep the car moving. You risk hydroplaning, flash floods and unpredictable behavior of other vehicles (which you can no longer see). Of course (in deference to Staf) well-designed automation in these cases may well do a better job than human drivers - more safely get you to the side of the road then stop.
As a cyclist I'm just shocked at how quickly and blindly that human auto drivers respond to an emergency vehicle approaching, because they simply swerve to the right and brake rapidly without looking for bikes that are already in the bike lane that are trying to avoid being run over. An automated vehicle hopefully would be smart enough to look before swerving right to make way for the emergency vehicle.
A few thoughts after reading the comments above. Recently I drove from Portland to Mt Hood, and I can say that even during the day in good weather, it is hard to see where the lanes are on the road up the mountain. I've driven it at night with snow and was really challenged to see where to drive. Maybe they need to embed RFID on the road to help out. I'd actually gladly give over the job of driving, IF I thought the car could handle it.
Also for hard rain, I cannot say enough good things about Rain-X. You can actually see the road in the worst downpours and as you pass big-rigs. Maybe Rain-X will be mandatory for autonomous cars. Or, they can use AI to remove the optical distortion from water on glass....
RainX sounds good Tom - for normal rain. I'll definitely look into it. But Florida rain isn't normal. Even if the sheet of rain hitting the windscreen instantly disappears, behind that is another sheet of rain and another and.. It's not so much that you can't see through the windscreen. It's that you can't see through all the rain beyond the windscreen. It's like being underwater!