Abstract: To reduce the potential losses caused by the inaccuracy of threat assessment and sensor radiation in the process of aerial target threat assessment, a risk-based multi-sensor management approach is proposed in this paper. First, a sensor management model based on partially observable Markov decision process is built. Second, the belief-state-based prediction methods for threat assessment risk and radiation risk are proposed to quantify the potential losses. Then, a non-myopic objective function based on multi-step risk prediction value is built and the objective is to obtain the minimal sum of threat assessment risk and radiation risk. Furthermore, to efficiently obtain the optimal solution, the sensor management problem is transformed into a decision tree search problem, and a branch-and-bound-based uniform cost search algorithm is designed. The simulation results show that the proposed algorithm can find high-quality solution while greatly reducing the computational time and memory consumption compared with the classical algorithms. The proposed management approach can accurately predict the risk, and has better risk control effect compared with the existing sensor management methods.

Key words: sensor management, threat assessment, risk control, partially observable Markov decision process, decision tree search