基于多控制智能体的改进shepherding行为

Improved shepherding behavior with multiple shepherds

针对目前shepherding行为生成方法中,shepherd的运动方式缺乏速度和时间约束的缺陷,提出了一种改进的shepherding行为生成方法。通过在shepherd的路径规划中,把基于概率路径图(probabilistic roadmap,PRM)的多智能体解耦式规划方法和基于速度调节的时变规划算法相结合,使得shepherd的运动满足时间和速度的约束,同时利用优先级策略实现shepherd之间的信息交互能力,因此产生的shepherding行为更加真实。仿真结果表明,改进的shepherding行为更符合骚乱事件中真实人的运动特性。

According to current shepherding behavior, motions of the shepherds lack of the restrictions of time and velocity. An improved method is presented to solve this problem. By combining multi agent decoupled planning based on probabilistic roadmap and time varying planning based on velocity-tuning method, the mo tions of the shepherds satisfy the restrictions. Communication capacity among shepherds is also realized by means of priorities. Thus, a more realistic shepherding behavior is achieved. Simulation tests show that the im proved shepherding behavior accords with the characteristics of the real people in riot events more ideally.