Thursday, March 8, 2012

Automated and Driverless Cars: Great for Safety, Not So Much for Congestion.

Can automated cars "cure" crashes and congestion? Renown (ex) Stanford University professor Sebastian Thrun who's team won ARPA's $2,000,000 driveless car challenge a few years ago thinks so as he presents the Google driverless Toyota Prius in this TED video.

This is an area where I believe that lawyers and politicians have more impact than engineers and technologists. The US had a fully developed and tested AHS or Automated Highway System in the mid-1990s as the sample article Whatever Happened to Automated Highway Systems? reminds us.

For those of us involved with intelligent transportation systems (ITS) the image below of eight large Buicks developed by California's Partners for Advanced Traffic and Highways (PATH) remains etched in memory. Observe the 0.2 second clearance between the AHS Buicks at 60 mph and the typical 2.0 second clearance in regular traffic.

When success was fully demonstrated, the government cut AHS funding because the issue became liability not technology. However, many of the technologies trickled down to piecemeal applications, some of which I summarize below.

Greyhound buses in the mainland have vibrating steering wheel (modeled after the aviation stick shaker to warn of impending stall) activated by radars if the bus tries to change onto a lane that is occupied by a vehicle. This also serves as an alarm if the drivers becomes drowsy. Daimler has introduced this to Mercedes cars but the system is not available in the US (due to liability.)

Since 2005 one can purchase many luxury vehicles with intelligent cruise control that can follow the car ahead. Some of them will bring a car to a complete stop automatically if the leader car comes to a stop. Some companies brand it as Adaptive Cruise Control and here is a demonstration dating to back 2008 at about 90 mph by a motorist on an autobahn.

In Europe higher priced BMWs will soon be offered with a system that if its driver becomes incapacitated, the car will maneuver itself, at German autobahn speeds, all the way from the fast lane to the right side shoulder, stop and send an SOS.

Many inexpensive cars in Europe in the $20,000 bracket have optical sensors on the bottom side of their exterior mirrors that follow the lane markings. They issue a "lane departure" warning to their driver. A handful of cars brought in the US in 2012 have this option too.

The US federal government has a major research initiative called to further boost these efforts.

And now for the conclusion and why AHS was terminated as a capacity enhancement: On a busy highway most drivers follow each other at a headway of about 1.5 seconds. As a result, the maximum sustained capacity of a freeway lane is 3600 seconds in one hour divided by 1.5 second headway equals 2400 vehicles per hour.

If car technology takes over, this headway can be reduced to 0.5 seconds which triples the capacity of the same freeway lane. So one lane could carry as many cars as an entire 3-lane section of the H-1 Freeway! This is clearly a bargain for our highway infrastructure.

However, if this was ever launched, it would require the presence of a largely empty lane next to the AHS lane (such as a bus-only lane with large gaps between the buses) so that vehicles can be merged in and out the tight AHS platoon; see the empty lane next to the platoon of fast moving Buicks in the picture above.) Only professional race drivers can routinely cope with 0.5 second headways (and they fail almost at every NASCAR race.)

With the press of the AHS button, merging into the tight lane, traveling at 60 mph and exiting the AHS lane will be done entirely by the computer, sensors and servos of the car in dense traffic. Now visuallize such a car with mommy, daddy and two kids in the back on a dusty, rainy or dark environment which may affect sensor performance and image recognition. There is clear risk and because of the tightness of the platoon, one mishap will likely cause large losses. Who is liable? The feds wanted none of this on the federal interstate system.

AHS has a tremendous promise for safety bust much less promise for direct congestion reduction. However, crash reduction does help traffic congestion because by most accounts 30% to 60% of the annual traffic congestion in a metro area is caused by accidents that block traffic lanes. Intelligent systems minimize driver error and accidents, so lanes become closed less often.

Mid-March 2012 update: The Economist publishes Self-driving cars. Safer at any speed? "Another headache will be lawsuits from motorists blaming their car for crashes. Honda is already being sued in America over the collision-avoidance system on its top-end Acura models. Pim van der Jagt, a research chief at Ford, says new laws will be needed to deal with such issues—and cars may need black boxes to record what went wrong in accidents."