Cooperative 3-D Active Multi-Robot Multi-Target Tracking

Abstract

In this paper, we present a novel algorithm for the 3-D active multi-robot multi-target tracking problem in a distributed manner. For a team of robots equipped with sensors, we design the estimation framework by integrating a distributed target state estimation algorithm with the cooperative visual inertial odometry (CVIO) algorithm. This approach allows moving robots to continuously update their localization and target estimates by using available measurements in their neighborhoods. Motion planning for the robots is tackled through a distributed optimization paradigm as a function of the estimation results. In the distributed optimization setting, all robots cooperate to find their control actions using local information. Cost functions are defined using a differentiable field of view concept for the targets' uncertainty reduction, alongside potential functions for collision avoidance and connectivity maintenance.