An architecture for autonomous vehicle navigation

The Autonomous Land Vehicle group at CMU is building an intelligent sensor-equipped robot vehicle. We have produced the first demonstrations of a system that uses vision to follow roads. This paper first briefly describes the hardware, including the vehicle, sensors, and the computers. We then present an overview of the sensor interpretation processes. The main part of the paper is a discussion of the blackboard architecture that ties all the processes together, and our experiences with a real implementation. Finally, we discuss our plans for the future.