Self-Evaluating Space and Robotic Agents

Abstract : In the past three years, we focused on self-evaluative methods for agents that interact with other agents and dynamic environments. We started with the observation that it would be beneficial for agents to sense prevailing qualities that stem from their interactions such as situation awareness, sociability, coordination, autonomy, failure tolerance, timeliness, and purposefulness. Agents usually have access to constraining requirements over these qualities. Additionally, many of these qualities are conflicting but a balance is desirable for a given domain and agents can discover that in operation. We have shown that an agent could attempt adjustments in it interactions to bring about favorable global changes. Such abilities require agents to have capabilities at the architectural level. Theoretical results include various models of relationships among social notions, and several models of autonomy, trust, and Power Simulation results include two implemented multiagent systems as testbeds. Over 60 published reports listed at the end of this report present our theoretical developments and reports of experiments from simulations. In the next two sections we briefly outline theoretical developments and implemented simulations.