A Non-myopic Scheduling Method of Mobile Sensors for Maneuvering Target Tracking

Multi-sensor cooperative scheduling has been the main mean to gather intelligence information due to the complexity of the battlefield environment and variability of targets. This study presents a non-myopic scheduling method of mobile sensors for manoeuvring target tracking. Within the partially observable Markov decision process framework, the sensor scheduling model is formulated since the target state cannot be observed directly, and the cost function is given based on the posterior Carmer-Rao lower bound. The multi-step scheduling cost is predicted with a certain number of particles generated by the unscented sampling method, which can reduce the computation complexity. For multi-target tracking cases, a target threat degree function is presented to assess the target threat. The scheduling problem is transformed into a decision tree optimisation problem by discretising the manoeuvring direction, and an improved decision tree searching algorithm is proposed to solve the sensor scheduling scheme quickly based on the branch-and-bound technique.