Multi-Target Radar Search and Track Using Sequence-Capable Deep Reinforcement Learning

2025-02-19

AI Summary

To address the challenge of joint optimization for multi-target search and tracking in complex radar environments, this paper establishes a 3D Active Electronically Scanned Array (AESA) radar simulation framework integrated with multi-object tracking (MOT) algorithms to enhance observation quality, and proposes a deep reinforcement learning (DRL)-based unified task management architecture. Its key contributions include: (i) the first empirical validation of multi-head self-attention mechanisms for modeling dynamic radar task sequences, demonstrating substantial performance gains; and (ii) a dual pretraining strategy combining behavioral cloning and autoencoding to mitigate sparse reward issues and policy cold-start problems. Experiments show that the proposed architecture significantly outperforms RNN- and CNN-based baselines: under high-density clutter and target maneuver conditions, multi-target identification accuracy improves by 12.6%, and tracking continuity increases by 23.4%. These results substantiate the efficacy and robustness of sequence modeling for intelligent radar task scheduling.

Technology Category

Search and Optimization: Learning to SearchPlanning, Routing, and Scheduling: Activity and Plan RecognitionMachine Learning: Transfer, Domain Adaptation, Multi-Task Learning

Application Category

Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingUser Modeling, Personalization and Recommendation: Fairness-aware retrieval and rankingResponsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

The research addresses sensor task management for radar systems, focusing on efficiently searching and tracking multiple targets using reinforcement learning. The approach develops a 3D simulation environment with an active electronically scanned array radar, using a multi-target tracking algorithm to improve observation data quality. Three neural network architectures were compared including an approach using fated recurrent units with multi-headed self-attention. Two pre-training techniques were applied: behavior cloning to approximate a random search strategy and an auto-encoder to pre-train the feature extractor. Experimental results revealed that search performance was relatively consistent across most methods. The real challenge emerged in simultaneously searching and tracking targets. The multi-headed self-attention architecture demonstrated the most promising results, highlighting the potential of sequence-capable architectures in handling dynamic tracking scenarios. The key contribution lies in demonstrating how reinforcement learning can optimize sensor management, potentially improving radar systems' ability to identify and track multiple targets in complex environments.

Problem

• Optimize sensor task management for radar systems
• Efficiently search and track multiple targets
• Enhance radar systems in complex environments

Innovation

• Deep Reinforcement Learning
• Multi-Headed Self-Attention
• 3D Simulation Environment