虚拟人自适应行为的多Agent协调控制算法  

Multi-Agent Cooperative Manipulating Algorithm for Virtual Human’s Adaptive Behavior

作　　者：王朝晖[1] 郑国磊[1] 朱心雄[1] 

机构地区：[1]北京航空航天大学机械工程及自动化学院,北京100191

出　　处：《工程图学学报》2009年第3期46-54,共9页Journal of Engineering Graphics

摘　　要：提出了基于Multi-Agent的虚拟人自适应行为控制算法。算法定义了模糊多目标人体舒适度评估模型,设计了行为协调的控制机制,将行为规划问题转化为分阶段的最优化问题。由于良好的工作姿态是人在生理上多方面利益互相妥协的结果,所以纳什均衡理论被用于控制姿态优化,使得虚拟人各方面的舒适度能够达到纳什均衡下的帕雷托最优。本算法可以支持连续的行为仿真,协助完成产品的人机工程分析。A Multi-Agent based algorithm is presented to realize virtual human＇s adaptive behavior manipulating. This method defines a fuzzy multi-objective evaluation model for analyzing human＇s comfort level. A cooperative control mechanism is designed to organize mannequin＇s actions. In this method, the problem of body＇s behavior planning is converted to an optimization problem with multiple stages. Because one satisfied gesture is a compromise among utilities of multiple physiological factors, the theory of Nash equilibrium is applied to optimize virtual mannequin＇s posture and make the Pareto-optimum under the Nash equilibrium on comfort level utilities. The method may realize sequential behavior animation, and be helpful to ergonomic evaluations of product.

关 键 词：计算机应用 虚拟人行为仿真 模糊多目标决策 舒适度 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]