October 19, 2010

Are the Autonomous Cars in an Open Loop?

Along with Google’s autonomous cars there is discussion of other autonomous car projects in development, such as the driverless taxi project in Berlin.

With laser scanners, radar and cameras the cars have an impressive set of tools to detect their environment and adjust their motion accordingly.  This works well when the objects in the cars path are static (road, trees, etc) or objects (other cars) moving in a well-behaved manner.

However, all the autonomous cars seem to be in an open loop. There is no active feedback from other cars (in the vicinity) on how they are likely to move, as will be required for a closed loop system. When the other cars do not move in a well-behaved manner (sudden braking by the car in front, a sudden swerve by car on the side etc), the open-loop system will break and the results can be unpredictable.

Google with its compute capacity definitely has the means to control the cars in a closed loop. But it also requires
  • Each car on the road to be connected to a central network
  • An always-on central network that is available to the cars at any location

Each large projects on their own.

Without a closed loop system, the cars cannot be used in a truly autonomous fashion, with no human intervention.