3 ms·

Really great point about cars only taking off once they paved the roads. The government should get involved in building out infrastructure that will make it eas

Really great point about cars only taking off once they paved the roads. The government should get involved in building out infrastructure that will make it easier for self driving cars to work better, and get the technology production ready faster. Such an investment in next-gen infrastructure could be a huge competitive advantage for the US economy.

Shivetya 10y ago
I guess my question is how do you denote road ways in such a way that is detectable in all weather types without being sight pollution or requiring external power sources. I can only figure embedding something in the road in such a pattern that it indicates direction of travel. Yet I don't know what can be buried that can work that doesn't require power but can be detected by a car traveling past it.
- lttlrck 10y ago
  Perhaps magnets or passive transponders.
- jessriedel 10y ago
  Several projects have had good results with embedded magnets: http://www.gizmag.com/volvo-road-magents-autonomous-cars/31172/ http://www.gizmag.com/volvo-road-magents-autonomous-cars/311... It wouldn't be too hard to pattern the magnets to communicate other info to the car too, such as an upcoming sharp turn. More abstractly, magnets could mark the car's position extremely precisely, and the car could then acquire regularly updated info about upcoming road conditions from the internet, or a short-range local radio broadcast. Magnets seem very cheap compared to the cost of resurfacing. I don't understand why we haven't been putting these in for years. > The magnet-based system relies on magnets embedded along the center of the lane 1.2 meters apart to define the roadway. The car tracks the magnets with less than 7.5 centimeters of error. Although this requires doing something to the road, it is very cost-effective to spend about $10,000 per lane mile to buy and install the magnets compared to spending from $1 million to $100 million per mile to build new lanes. And the life of the magnets is 30 to 50 years -- longer than the best pavements. http://www.fhwa.dot.gov/publications/publicroads/97novdec/p97nov32.cfm http://www.fhwa.dot.gov/publications/publicroads/97novdec/p9...
  - nojvek 10y ago
    $10000 per mile is still a huge cost. If human is able to do it then AI should be able to do it. The cost should be no more than reprinting blurred out lanes. If a car can work if the roads are well marked, run it can work in many different countries too. That's a huge boon for car makers.
    - jessriedel 10y ago
      Whether AI can replicate all human abilities is an empirical question, not a matter of "should". And it's clear it can't currently. Given that 30k people per year die of preventable traffic accidents, we obviously shouldn't wait to deploy AI until it is perfect. Adding magnets to all 160,000 miles of the US highway system would be a one-time cost of roughly $10B, which is small in comparison to the hundred of billion we lose every year that autonomous cars are delay. Additionally, humans do fail to keep track of lane lines in bad weather, and this result in accidents that could be prevented if magnet-directed computers were driving the car.

Shivetya 10y ago

I guess my question is how do you denote road ways in such a way that is detectable in all weather types without being sight pollution or requiring external power sources. I can only figure embedding something in the road in such a pattern that it indicates direction of travel. Yet I don't know what can be buried that can work that doesn't require power but can be detected by a car traveling past it.

lttlrck 10y ago
Perhaps magnets or passive transponders.
jessriedel 10y ago
Several projects have had good results with embedded magnets: http://www.gizmag.com/volvo-road-magents-autonomous-cars/31172/ http://www.gizmag.com/volvo-road-magents-autonomous-cars/311... It wouldn't be too hard to pattern the magnets to communicate other info to the car too, such as an upcoming sharp turn. More abstractly, magnets could mark the car's position extremely precisely, and the car could then acquire regularly updated info about upcoming road conditions from the internet, or a short-range local radio broadcast. Magnets seem very cheap compared to the cost of resurfacing. I don't understand why we haven't been putting these in for years. > The magnet-based system relies on magnets embedded along the center of the lane 1.2 meters apart to define the roadway. The car tracks the magnets with less than 7.5 centimeters of error. Although this requires doing something to the road, it is very cost-effective to spend about $10,000 per lane mile to buy and install the magnets compared to spending from $1 million to $100 million per mile to build new lanes. And the life of the magnets is 30 to 50 years -- longer than the best pavements. http://www.fhwa.dot.gov/publications/publicroads/97novdec/p97nov32.cfm http://www.fhwa.dot.gov/publications/publicroads/97novdec/p9...
- nojvek 10y ago
  $10000 per mile is still a huge cost. If human is able to do it then AI should be able to do it. The cost should be no more than reprinting blurred out lanes. If a car can work if the roads are well marked, run it can work in many different countries too. That's a huge boon for car makers.
  - jessriedel 10y ago
    Whether AI can replicate all human abilities is an empirical question, not a matter of "should". And it's clear it can't currently. Given that 30k people per year die of preventable traffic accidents, we obviously shouldn't wait to deploy AI until it is perfect. Adding magnets to all 160,000 miles of the US highway system would be a one-time cost of roughly $10B, which is small in comparison to the hundred of billion we lose every year that autonomous cars are delay. Additionally, humans do fail to keep track of lane lines in bad weather, and this result in accidents that could be prevented if magnet-directed computers were driving the car.