进化学习模糊规则实现移动机器人的自适应导航

Learning Fuzzy Rules by Evolution for Mobile Robot Adaptive Navigation

给出了一种轮式移动机器人导航的方法,机器人由一组动态变化的模糊规则集控制,通过遗传算法在线调整和学习模糊规则.根据机器人的运动模型构造了模糊控制器,采用变长度编码方法对规则编码,减少了染色体的尺寸和复杂度,提高了学习速度.通过竞争型小生境遗传算法解决了模糊规则的学习问题法,并分析了设计中遇到的如多条模糊规则同时激活的信度分配等问题.学习过程在二维仿真环境下完成,在自行开发的全局视觉平台上对学到的规则进行了验证.实验结果证明,该方法是正确可行的.

A learning mechanism for wheeled mobile robot navigation was proposed. The robot is controlled by a dynamic set of fuzzy rules and the rule set is tuned and learned using a genetic algorithm. The fuzzy controller is constructed based on the kinematic model of the robot. Messy coding method is used to reduce the size and complexity of chromosome and increase the learning speed. By using niche genetic algorithm the learning problem of fuzzy rules is solved. The problems encountered in the design, such as the credit assignment of rules that are fired simultaneously, were discussed. After the rules were learned in a 2-D coordination simulated environment, they were tested in a globe-vision-system. The experimental results prove that the learning mechanism is correct and feasible.