Developing a Framework for Biological Intelligence Modules in Autonomous Social Robots

Autonomy implies the ability of an agent to be self-directing, thus independently handling tasks and responding to stimuli, as well as make their own decisions. Current mobile social robots are not autonomous as demonstrated by the frequently observed Frozen Robot Problem (FRP) as demonstrated when a robot freezes when encountering crowds and unexpected pedestrian behavior. Developing an autonomous system that can naturally react to such unplanned experiences is both an interdisciplinary and integration activity that involves sensing, artificial intelligence (AI), and robotics. In this paper, the authors document the development of a modular testbed whole system approach to autonomous mobile robots, their modeling and simulation (M&S), and demonstrate the system. The system developed herein leverages a biologically inspired cognitive architecture for decision making and memory through the Semantic Pointer Architecture (SPA) for context-based decision making and motor control for a robot. Through this approach, the autonomous system thus employs reactive goal-driven navigation which enables a robot to move toward a desired location by dynamically avoiding obstacles without relying on traditional path planning. Results are presented from M&S through factorial experimental designs to explore architecture and environment complexity as well as live testing in similar environments. The entire approach presented herein provides a framework for developing, testing, and validating autonomous robotics systems.