POTENTIAL-FIELD BASED LOCOMOTION REFLEXES

The integration of sensor perception into the world model of a mobile robot is not performed without delay and loss of information. Path planning in this model can't thus react immediately and error free to environmental dynamics. The reflexive planner, presented in this paper, uses only unprocessed sensor data to control collision free robot motions in an unknown, dynamic environment. The potential field method is used for planning. Local minima reduce the functionality, oscillations at nearby obstacles limit the execution speed inherently. Solutions to both problematics and an integration of reflexive planning into a navigation hierarchy are proposed.