摘要
提出一种基于代价地图和最小树的多区域覆盖方法.首先,确定先验静态地图的栅格代价,并将静态地图转化成动态代价地图;然后,将代价地图分割成若干个区域块,对区域块再作Boustrophedon单元分割处理,得到子区域信息;最后,根据子区域间的邻接关系构建图,图中的各节点代表一个子区域.将基于最小树的子区域规划算法应用于该图得到子区域规划序列,应用基于代价地图和Dijkstra算法的区域转移算法得到区域间转移路径,子区域内使用往返覆盖策略,以此实现区域的全覆盖.将本文提出的多区域覆盖方法在线应用于多种结构场景中,实验结果表明,相对于传统的区域覆盖方法,本文方法可以有效提高移动机器人区域覆盖的效率.
A multi-region coverage method based on cost map and minimal tree is proposed. First, the cost of cells in priori static map is determined, and then the static map is turned into a dynamic cost map. Then, the cost map is divided into several regions, which are decomposed into sub-regions using Boustrophedon cellular decomposition. Finally, a graph is built according to the adjacency relationship between the sub-regions, in which each node represents a sub-region. The sub-region planning method based on minimal tree is applied to the graph to generate a sub-region planning sequence, the algorithm based on the cost map and Dijkstra algorithm is used to plan a transfer path among different regions, and back and forth coverage strategy is applied to every sub-region to complete the coverage. The multi-region coverage method is applied to completing coverage in different environments on-line, and experimental results show that our algorithm has advantages in improving the efficiency of coverage over traditional methods.
出处
《机器人》
EI
CSCD
北大核心
2015年第4期435-442,共8页
Robot
基金
国家自然科学基金资助项目(61375079)
