We consider the online planning problem for a team of agents with on-board sensors to discover and track an unknown and time-varying number of moving objects from sensor measurements with uncertain measurement-object origins. Since the onboard sensors have limited field of views (FoV), the usual planning strategy based solely on either tracking detected objects or discovering unseen objects is inadequate. To address this, we formulate a new multi-objective multi-agent model for a predictive control problem based on information-theoretic criteria; cast as a partially observable Markov decision process (POMDP). The resulting multi-agent planning problem is exponentially complex due to the unknown data association between objects and multi-sensor measurements; hence, computing an optimal control action is intractable. We prove that the proposed multi-objective value function is a monotone submodular set function, and develop a greedy algorithm that can achieve an 0.5OPT compared to an optimal algorithm. We demonstrate the proposed solution via a series of numerical experiments with a real-world dataset.
View on arXiv