一种优化的联合极大似然SLAM数据关联方法

提出一种将联合极大似然方法(JML)和遗传算法相结合解决SLAM数据关联问题的方法,简称GAJML.该方法采用"关联门"缩小数据关联的解空间范围,提高搜索效率;利用数据关联解的联合极大似然值作为适应度值,种群的初始化采用了自适应策略以提高算法计算速度.与单匹配最近邻(ICNN)和JML方法的对比实验表明该方法相比于ICNN方法耗时增加很少实时性好,数据关联正确率接近JML准确度高,并能够有效克服闭环问题引起的定位累积误差增长.

A Joint Maximum Likelihood Method Optimized Data Association Approach to SLAM Problem

A data association approach for SLAM is proposed by combining genetic algorithm and joint maximum likelihood (JML) method,named "GAJML" for short.In that approach,"association door" is used for reducing data association solution space to improve search efficiency.Joint maximum likelihood value of data association solution is taken as fitness value.A self-adaptation strategy is implanted into population initialization to improve computation efficiency.Experimental result indicates that its instantaneity is well with only increasing little time-consumption compared with ICNN(individual compatibility nearest neighbor),the data association correct rate is very high even close to that of the JML method,and can effectively overcome "close loop" problem caused localization accumulated error growing.