Method and device for controlling an autonomous exploring robot

摘要

The values sensed and the commands supplied by a strategy-generating network (SGWN) are sent to another network system with several separate networks (B1,B2) which decides between value pairs (S1/S2) obtained during the training phase and unknown value pairs. The value pairs go to a dynamic analyser (DAN) that decides on value pairs according to the stability of the output values. There is an interactive process in which the result of the dynamic analysis is passed back to the network as an additional input value.