Multi-Agent Intelligent Adaptive Coordinated Robotic System

摘要

The objective of this project is to study cooperative behaviors in the frameworkof mobile manipulatory agents, observer agents and a human operator for semiautonomous material handling in unstructured environments. A direct impact of using distributed agents (which are cheaper, faster and lighter) is to perform tasks more reliably, in that a broken agent can be replaced by another, or the remaining agents can be reconfigured to carry out a task. Conventionally, physical interactions/contacts between agents are regarded as accidents and are avoided if at all possible. In this approach to multiagent cooperation, physical interactions are positively exploited for the dual purposes of: (1) cooperative manipulation; and (2) communication. The information communicated through physical interactions includes the relative position/orientation, and interactive forces/moments. Sensing gives each agent independence, while communication is a prerequisite to cooperation. The assumption is that different agents, including the human agent, have different degrees of knowledge and views of the world with varied spatiotemporal resolution and capabilities of interaction. Agents ask for and take advice (communicated information), using it for control and completion of their task, unless they decide their sensory input is more appropriate for the given subtask.