Non-myopic GOSPA-driven Gaussian Bernoulli Sensor Management

George Jones,Angel Garcia-Fernandez,Christian Blackman

DOI: https://doi.org/10.1109/TAES.2024.3418750

2024-06-27

Abstract:In this paper, we propose an algorithm for non-myopic sensor management for Bernoulli filtering, i.e., when there may be at most one target present in the scene. The algorithm is based on selecting the action that solves a Bellman-type minimisation problem, whose cost function is the mean square generalised optimal sub-pattern assignment (GOSPA) error, over a future time window. We also propose an implementation of the sensor management algorithm based on an upper bound of the mean square GOSPA error and a Gaussian single-target posterior. Finally, we develop a Monte Carlo tree search algorithm to find an approximate optimal action within a given computational budget. The benefits of the proposed approach are demonstrated via simulations.

Systems and Control

What problem does this paper attempt to address?

The paper attempts to address the problem of designing a non-myopic algorithm for sensor management within the Bernoulli filtering framework. Specifically, the study focuses on the sensor management problem in scenarios where at most one target is present. To achieve this goal, the paper makes the following contributions:

1. **GOSPA-based Sensor Management Framework**: The paper proposes a sensor management framework based on the Generalized Optimal Sub-Pattern Assignment (GOSPA) metric, suitable for both myopic and non-myopic planning.

2. **Closed-form Cost Function**: A closed-form upper bound of the Mean Square GOSPA (MSGOSPA) error for Gaussian single-target distributions is developed as a cost function to assign costs to each available action. This closed-form cost function is computationally efficient in linear Gaussian systems.

3. **Monte Carlo Tree Search Algorithm**: A planning algorithm implementation based on the Monte Carlo Tree Search (MCTS) method is developed, enabling the solution of non-myopic planning problems with high computational efficiency.

4. **Detection Probability Approximation**: An approximation of the expected detection probability is provided for sensors with circular Field of View (FOV) and constant detection probability, based on the importance sampling method.

In summary, the main objective of the paper is to propose a non-myopic sensor management strategy that effectively handles multi-step future time-step predictions by combining the GOSPA metric and the MCTS algorithm.