Performance analysis of clustering methods for balanced multi-robot task allocations
研究了带平衡约束的多机器人任务分配问题，用聚类方法将任务分组并规划路径，通过基准数据集比较了K-means、高斯混合模型和层次聚类的性能，发现K-means最适合复杂拓扑且可扩展。
This paper models the Multi-Robot Task Allocation (MRTA) problem with a balance constraint to improve the utilisation (completion time) of the robots. Our balancing constraint attempts to minimise the travel distance difference among the robots as well as allocates an equal set of tasks to these robots. The clustering-based approach is employed to solve the Balanced Multi-Robot Task Allocation (BMRTA) problem for two principal reasons. That is, this approach clusters given tasks into groups using various clustering techniques for each robot and sequences the route for each robot using the travelling salesman problem (TSP) conhull algorithm. This work analyses the suitability and performance of the clustering techniques with respect to the balancing criteria using a benchmark dataset. Our findings suggest that K-means clustering is the most suitable for the solving BMRTA problem with complex topologies and it is scalable to deal with any number of tasks and robots compared with Gaussian Mixtures Models (GMM) and hierarchical clustering methods.