基于人工物理法的多智能体协同控制

 对多智能体的协同控制和队形重构技术进行研究，采用人工物理法完成多智能体的队形建立与队形保持。人工物理法是一种分布式的控制方法，对传感器信息和通信的依赖程度较低，并可以很好地移植到大规模系统中。人工物理法通过设定虚拟的物理力，完成机器人速度和方向信息的解算，并用这些信息进行实时控制。同时为保证在障碍物区域内的智能体避障，采用沿墙跟踪策略，设计双输入、单输出的模糊控制器实现智能体的避障。为克服多模态运动的转换瞬态，利用四点平滑抑制模态转换瞬态。通过传感器探测外界环境，获得相关环境信息并确定障碍物的位置，由此依据判决条件完成多智能体控制的模态转换，进行队形重构。通过复杂环境中的数学仿真及多智能体协同控制的实验验证，表明该系统具有良好的鲁棒性和适应性。

Cooperative Control of Multi-agent Based on Artificial Physics Method

The synergic control and the formation reconstruction of multi-agents are studied in this paper. Artificial physics method is used for the formation establishment and holding of multi-agents. Artificial physics concerns a kind of distributed control strategy, which does not very much depend on the message of the sensor and communication. Multi-agents carry out the complex tasks by local cooperation. At the same time, artificial physics method is easy to be implemented in a giant system. By constructing the virtual physics force in the movement area of a robot, the speed and the direction of the robot are determined and used to drive the robot with time. In order to deal with the obstacle, a fuzzy controller with double inputs and single output is designed according to the wall-following strategy. One thus can make sure that mobile robot keeps a certain distance from the obstacle and goes along the obstacle. The orientation error can be reduced in this manner. The mode switching transient is restrained by the four point smoothing method. The information about the circumstance and the position of obstacle obtained by the sensor can be used to make the mode switch according to some conditions and to make the formation reconstruction. The results of the math simulation and the multi-agents cooperative control experiment in complex circumstances show that the method proposed in this paper enjoys good stability and adaptability.