Policy Learning for Active Target Tracking over Continuous SE(3) Trajectories

2022/12/03 by Pengzhi Yang, Yang, Pengzhi, Shumon Koga +5 · 3 citations
Computer Science · #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Gaussian Processes and Bayesian Inference #Machine Learning (cs.LG) #Reinforcement Learning in Robotics #Robotics (cs.RO)

paper · pdf · doi:10.48550/arxiv.2212.01498

openalex publication_date 2022/12/03 · openalex created_date 2022/12/18 · openalex updated_date 2026/07/28

Abstract

This paper proposes a novel model-based policy gradient algorithm for tracking dynamic targets using a mobile robot, equipped with an onboard sensor with limited field of view. The task is to obtain a continuous control policy for the mobile robot to collect sensor measurements that reduce uncertainty in the target states, measured by the target distribution entropy. We design a neural network control policy with the robot SE(3) pose and the mean vector and information matrix of the joint target distribution as inputs and attention layers to handle variable numbers of targets. We also derive the gradient of the target entropy with respect to the network parameters explicitly, allowing efficient model-based policy gradient optimization.

Cited by

• Reinforcement Learning for Heterogeneous Sensor Selection in Maritime Surveillance
• MATT-Diff: Multimodal Active Target Tracking by Diffusion Policy
• Hierarchical Planning for Long-Horizon Multi-Target Tracking Under Target Motion Uncertainty

Related

• Learning Continuous Control Policies for Information-Theoretic Active Perception
• Non-Myopic Target Tracking Strategies for State-Dependent Noise
• Learning Sample-Efficient Target Reaching for Mobile Robots
• Learning to Track Dynamic Targets in Partially Known Environments
• Information Acquisition with Sensing Robots: Algorithms and Error Bounds