A fuzzy integrated robotic behavioral architecture

We propose a variant of the fine-grained subsumption architecture. We use active perception layers instead of whole behavior components. This provides for context sensitive behavior arbitration. Each behavior is tied to an active perception unit that may implement a behavior selection mechanism. Further, information passing between layers in terms of fuzzy variables is suggested. This allows for graceful command arbitration at the action execution layer. The whole robot control system can be viewed as a multistage fuzzy control system. We also provide a brief discussion of our implementation of the architecture. The minimal implementation of the scheme has provided encouraging results and further work is on in the analysis of the architecture.