Abstract :
A new mechanism, termed Behavior Arbitration Network (BARNET) is presented. BARNET is a dynamic adaptive network which is used for arbitration among behaviors and for combining reflexive behaviors and reasoning-driven behaviors in high-functionality intelligent autonomous systems (IASs). BARNET is structured as a networked team of stochastic learning automata, operating in multiple stochastic nonstationary environments as game of automata. The automata team selects the optimal combination of behaviors and their respective level of activation, while properly reacting to continuously changing and unexpected situations and events. BARNET enables a truly hybrid architecture which combines the advantages of both traditional reasoning-driven hierarchical architectures and behavior-based architectures, while lacking their inherent drawbacks. BARNET is embedded within the intelligent autonomous hyper-controller (IAHC) of the IAS. The motivation for developing BARNET, its architecture and algorithms, as well as simulation results, are described
Keywords :
adaptive systems; inference mechanisms; intelligent control; learning systems; stochastic automata; BARNET; Behavior Arbitration Network; behavior-based architectures; dynamic adaptive network; intelligent autonomous hyper-controller; intelligent autonomous systems; reasoning-driven hierarchical architectures; stochastic learning automata; Adaptive systems; Context modeling; Delay; Educational institutions; Intelligent networks; Intelligent systems; Learning automata; Predictive models; Robustness; Stochastic processes;
