An integrated particle filter and potential field method applied to cooperative multi-robot target tracking

We describe a novel method whereby a particle filter is used to create a potential field for robot control without prior clustering. We show an application of this technique to control a team of mobile robots to cooperatively locate and track a moving target. The particle filter models a probability distribution over the estimated location of the target, providing robust tracking despite frequent target occlusion. This method extends previous work in particle-filter-based tracking in two important ways. First, the particle cloud is never clustered to find a single estimate of target location. Instead, robot motion is guided by a potential field generated directly from the particle cloud. Secondly, effective coordinated multi-robot searching and tracking can be achieved by simply assigning a subset of the particles to each robot. Simulation trials and real robot experiments demonstrate the method successfully locating and tracking targets, and experiments show that multiple coordinated robots outperform a similar but uncoordinated team.