论文部分内容阅读
为自主地对2D 激光雷达感知的环境进行特征提取,提出一种改进的遗传聚类算法.首先将测距数据的空间近邻信息和模糊聚类相结合,提出一种加权的模糊聚类算法进行特征提取.针对聚类数目难以事先获得的问题,利用多种有效性索引对不同聚类算法的有效性进行计算评估,选取一种适合于测距数据有效性分析的索引函数作为遗传算法的适应度函数.同时,为解决聚类中局部最优问题,通过增加群体多样性,改进竞争择优的遗传算子来改进算法,以便提高局部搜索能力,加快收敛速度.通过相关算法的性能比较,本文方法的有效性得以验证.
In order to autonomously extract features from the 2D laser radar environment, an improved genetic clustering algorithm is proposed.Firstly, spatial neighbor information of ranging data is combined with fuzzy clustering to propose a weighted fuzzy clustering algorithm Feature extraction.Aiming at the problem that the number of clusters is difficult to obtain in advance, the effectiveness of different clustering algorithms is evaluated and evaluated by using multiple effective indexes. An index function suitable for the validity analysis of ranging data is selected as the genetic algorithm to adapt Meanwhile, in order to solve the local optimal problem in clustering, this paper improves the algorithm by increasing the population diversity and improving the genetic algorithm with competitive preference in order to improve the local search ability and speed up the convergence rate.Through the performance comparison of the relevant algorithms, The validity of the method is verified.