基于学习机制的移动机器人动态场景自适应导航方法

Mobile Robot Adaptive Navigation in Dynamic Scenarios Based on Learning Mechanism

针对在单一学习机制中,移动机器人自主导航一般只适用于静态场景,适应性差的问题,提出一种动态场景自适应导航方法.该方法通过激光测距仪(LRF)获取周围环境的距离信息,在基于增量判别回归(IHDR)算法的单一学习机制导航的基础上,提出了最远距离优先机制的局部避障环节.该导航方法克服了传统导航方法对环境模型的过度依赖,并且本文提出的基于最远距离优先机制的局部避障算法,解决了基于单一学习机制的导航方法对动态场景适应能力不足的问题.本文将动态场景自适应导航方法应用到了MT-R机器人中,与基于单一学习机制的导航方法进行了对比实验,并且运用提出的局部避障算法,对实验中的激光数据进行了算法性能分析.实验结果证实了该方法的可行性,并显示了该方法在动态场景下的良好表现.

Mobile robot navigation based on a simple learning mechanism is generally applied to static scenarios and has poor adaptability. Therefore,we propose a method of adaptive navigation under a dynamic scenario. In the method,we propose a local obstacle avoidance link to the maximum distance priority mechanism,on the basis of a simple learning mechanism,using an incremental hierarchical discriminant regression（ IHDR） algorithm,and acquire environmental distance information with a laser range finder（ LRF）. This overcomes the over-dependence on the environmental model in traditional navigation methods,and simultaneously resolves the problem of poor adaptive capacity in dynamic scene navigation with a simple learning-based mechanism,using the proposed local obstacle avoidance algorithm. We apply the proposed navigation method to an MT-R robot,and compare this with the experimental results from a learning-based navigation method. In addition,an algorithm analysis experiment is performed on LRF data using the proposed local obstacle avoidance algorithm. The results illustrate the feasibility of the proposed method,and reveal its effectiveperformance in dynamic scenarios.