Slashdot Mirror
How Much Will Autonomous Cars Really Help? (theconversation.com)
An anonymous reader writes: An opinion piece at The Conversation questions the common belief that autonomous vehicles will easily solve a host of problems with road-based travel, including safety and traffic. "Assuming autonomous vehicles were one meter apart and traveling at 100 kilometers per hour (an aim that has been stated as the ultimate hope) this would mean around 25,000 people per hour could be taken down a freeway lane. While impressive, this movement capacity is only half that of a train. But getting to this capacity means 100% of vehicles are under control of a guidance system, with none under independent control. As soon as one car does this, the whole system would slow down considerably, as is seen on freeways now." The writer argues that a better role for autonomous cars might be to take passengers to and from hubs for public transportation.
Something north of 90% of accidents are preventable; take a look at table 8 here: http://www-nrd.nhtsa.dot.gov/P...
That table shows the 'critical event' in an accident, which is what made it unavoidable. Just 1.4% of accidents are from an object or animal in the road. Likewise, only 1.2% are due to a vehicle problem, although a large percentage of those are improper maintenance, which would be solved by some autonomous vehicle business models where they are owned and maintained by a fleet company (such as Uber).
So we can prevent 90% of accidents, but you think it's not worthwhile because the other 10% still happen?
Furthermore, if the fleet model is adopted, it actually becomes more likely that safety improvements will make more financial sense; far fewer cars are needed in the fleet, so the costs are amortized over more people. But in either case, safety standards are set by the government, and we can choose to raise or lower them as we see fit, completely orthogonally from whether cars are autonomous or not.
This won't solve all accidents, but it will certainly improve the situation.
Indeed. The naysayers seem to forget that self-driving cars already have millions of miles of testing. If they were accident-prone, the data would show that, and it does not. Driving in close formation, or "platooning" is well tested. I remember seeing test car platoons on I-5, north of San Diego, in the 1990s. TFA is mostly nonsense and conjecture. It says that a single non-autonomous car will "slow the system down considerably". I see no reason that would be true. A human driven car would just mean one car would have a normal gap, but that wouldn't slow down other cars. I have heard the opposite: That even a few autonomous cars can make a big difference in preventing congestion, since they have more information about traffic conditions ahead, and can react quicker, so they smooth out the "accordion effect" for themselves as wells as all the cars behind them.
Comparing self-driving cars to trains is idiotic. I can't take the train to the grocery store. A stream of self-driving cars may have half the bandwidth of a passing train, but not if you consider the gaps between trains, which are usually far more than the length of the train. A mile of passenger rail costs about $100M. A lane of asphalt costs about $1M per mile.