提出一种系统结构为内部状态生成需求、需求指导感知和控制决策、感知为决策提供信息、行为执行结果通过感知反馈至内部状态的循环结构。内部状态作为实际物理量和抽象感念的交汇点;需求用于连接系统内的认知模块和物理模块;根据决策理论建立了需求指导感知的标准,运用多目标决策算法实现了以子需求为目标,在行为原子集合中选择合适的行为序列。通过与约束满足实现方法的比较,验证了本方法的有效性。
The architecture of the system is a cycle of Internal-State generating Desire, Desire guiding Perception, Decision-making based on Perception, and the Executing results feed-backing to change the Internal-State. In the proposed system, internal-state works as a point of intersection to integrate the physical factors with the conceptual factors ; desire plays as a bridge to link the modules of cognitive level and physical level. Besides, a set of criteria for guiding perception is proposed according to the game theory; and finding solutions from the atomic actions to meet the demand of sub-desires is fulfilled by a kind of multi-objective decision making algorithm. Comparing with another method, constraint satisfying algorithm, the results indicate the validity of this new approach.