Attentive Monitoring Strategies in a Behavior-Based Robotic System: An Evolutionary Approach

In this paper, we propose an evolutionary approach to develop an attentive executive control system for a robotic agent. We consider a behavior-based system endowed with simple attentional mechanisms that regulate sensors sampling rates and action activations. In this context, the overall emergent attentive behavior is specified by a finite set of parameters regulating a process of adaptation w.r.t. the internal processes and the external environment. In this work, we propose the deployment of a Differential Evolution algorithm to set these parameters. As a case study, we consider an agent operating in an adaptation and survival domain. The collected results show that the generated attentional control systems perform better then the hand-tuned ones and are general enough to remain effective and robust when deployed in different environments.