UTS-based foresight optimization of sensor scheduling for low interception risk tracking

摘要

In this paper, we are concerned with the problem of non-myopic sensor scheduling as a partially observable Markov decision process in which available sensors are assigned dynamically to observe targets for the trade-off between the tracking accuracy and the interception risk. Our scheduling problem is difficult to solve using traditional methods due to continuous and high-dimensional state space. However, foresight optimization restricts the sequence of action mappings to be a stationary action sequence and gives rise to the upper bound of the objective function value, leading to an approximate solution. Furthermore, unscented transformation sampling combined with extended Kalman filtering is proposed to provide an approximate evaluation of the upper bound that results from an action sequence. To determine the best action sequence efficiently, a pruning algorithm is incorporated into the tree-search techniques. The feasibility and effectiveness of our non-myopic schemes are verified in a simple simulation experiment that involves multiple radars tracking a single target.