Action-oriented sensor data integration and its application to control of an autonomous vehicle

We describe the control system structure of our test autonomous vehicle for indoor movement. The control system consists of several independent "sub-controllers". Each sub-controller has its own map and planner and is tuned for one particular environment. As the vehicle moves around and encounters a different environment, a different sub-controller becomes active and plans vehicle's action. The data processing in a sub-controller becomes simple, as it copes only with one particular kind of environment. This approach enables the machine to cope with various environments by adding different sub-controllers.<>