Abstract: To deal with the problem of sensor resource scheduling for continual multi-object tracking with low earth orbit constellation, the object tracking tasks are first classified into two sets, i.e., the higher precision task set and the lower precision task set, and their state transition procedures are analyzed. Then, a sensor scheduling model for each task set is established based on the dynamic priority theory. Furthermore, a novel real-time sensor scheduling algorithm based on a multi-model is proposed. Simulations executed in diverse scenarios indicate that the applicability of the proposed algorithm exceeds that of the traditional methods which consider the dilution of precision as the optimal objective or threshold constraint. This is especially true in the case of multi-object scenarios of centralized distribution. The drop-out ratio of the proposed method is greatly reduced as compared with that of the traditional methods, although a few objects show slightly larger tracking errors.

Key words: low earth orbit constellation, sensor network, scheduling algorithm, multiobjective optimization, model, tracking