Constrained Multi-Sensor Control Using a Multi-Target MSE Bound and a δ-GLMB Filter.

Area of Science:

• Control Systems Engineering
• Signal Processing
• Data Fusion

Background:

• Existing multi-sensor control algorithms for multi-target tracking (MTT) rely on distributed processing and the generalized covariance intersection (GCI) rule.
• A significant limitation of GCI is the lack of a reliable basis for setting normalized fusion weights, impacting posterior density accuracy.

Purpose of the Study:

• To propose a novel constrained multi-sensor control algorithm using a centralized processing architecture to overcome GCI limitations.
• To introduce a multi-target mean-square error (MSE) bound as a cost function for optimizing sensor control commands.

Main Methods:

• Developed a centralized multi-sensor control algorithm to avoid the GCI rule.
• Defined a multi-target MSE bound as the cost function, minimized by control commands.
• Utilized the generalized information inequality for RFS observations and a δ-generalized labeled multi-Bernoulli filter for multi-target Bayes recursion.

Main Results:

• The proposed algorithm provides an online indication of achievable MTT precision limits.
• Suboptimal algorithms (mixed penalty function and complex methods) were employed to reduce computational cost.
• Simulations demonstrated superior MTT precision compared to GCI-based methods, especially under varying observation performance.

Conclusions:

• The new constrained multi-sensor control algorithm offers enhanced MTT precision by avoiding GCI limitations through a centralized approach.
• The mixed penalty function and complex methods offer efficient computation for large-scale sensor systems with minimal impact on tracking accuracy.