Context-Dependent Structure Control for Adaptive Behavior Selection

This paper presents a bio-inspired two-layer architecture that employs a context-dependent “structure control” mechanism for autonomous robotic agents to achieve adaptive behaviors in dynamic physical-social environments. In this architecture, the bottom layer is an asymmetry mutual inhibition behavior network where different behaviors inhibit each other for behavior selection. The top “behavioral context” layer dynamically sets the behavior selection context by changing the structure of the behavior network, thus modulating an agent’s overall behavior patterns according to different operating conditions. An example of dynamic team formation multi-robot system is presented. Simulation results and properties of this architecture are discussed.