Adaptive horizon model predictive control based sensor management for multi-target tracking

The flexibility of modern multi-sensor data fusion systems can only be exploited through on-line sensor management. This paper discusses a sensor management architecture based on model predictive control theory which optimises the configuration of sensors tasked with tracking dynamic targets using predictions of future target states. Methods for defining the optimisation horizon are discussed and the relationship between the accuracy of the predictions and the quality of the resulting utility estimates is discussed. A novel adaptive horizon approach, which exploits knowledge of the current target state estimate uncertainty, target dynamics, target noise properties and filter performance is proposed. This increases estimation performance in scenarios with high levels of uncertainty