Long-Horizon Constraint-Aware Collaborative Scheduling for Multiple Phased-Array Radars Using Mamba Temporal Encoding...

Area of Science:

• Radar Systems Engineering
• Artificial Intelligence
• Operations Research

Background:

• Phased-array radar networks face complex real-time scheduling challenges due to coupled constraints.
• Long-horizon scheduling is difficult as current decisions impact future radar states like power and tracking uncertainty.

Purpose of the Study:

• To propose CS-Mamba, a novel long-horizon, constraint-aware collaborative scheduling framework for homogeneous multiple phased-array radars.
• To address the challenges of real-time task scheduling under resource, beam, time-window, and power limitations.

Main Methods:

• Formulated the scheduling as a finite-horizon constrained decision process with hybrid actions (radar-task matching and power allocation).
• Introduced a Mamba-based temporal encoder for efficient summarization of long scheduling histories.
• Developed an action-dependent radar model linking power, SNR, detection, noise, and tracking covariance.

Main Results:

• CS-Mamba improved the normalized cost-effectiveness score from 0.62 to 0.78 compared to MAPPO in the Medium scenario.
• Reduced end-to-end decision latency from 24.5 ms to 12.8 ms per step.
• Ablation studies confirmed the effectiveness of temporal encoding, structured matching, and feasible power projection.

Conclusions:

• CS-Mamba offers a significant advancement in collaborative scheduling for phased-array radar networks.
• The framework demonstrates superior performance in cost-effectiveness and decision latency, validating its components and training methodology.