Resource Management for Multifunction Multichannel Cognitive Radars

A modern radar may be designed to perform multiple functions, such as surveillance, tracking, and communications. A radar resource management (RRM) module makes decisions on parameter selection, prioritization, and scheduling of such tasks. RRM becomes especially challenging in overload situations, where some of the tasks may need to be delayed or even dropped. In general, task scheduling is an NP-hard problem. Finding the optimal solution has high computational complexity, and heuristic methods, while having low complexity, provide relatively poor performance. In this work, we use machine learning-based techniques to address this issue. We develop a cognitive scheduler which is trained through the interactions of the radar with the environment. Specifically, we propose an approximate algorithm based on the Monte Carlo tree search method. A policy network is also used to help to reduce the width of the search. Such a network can be trained using reinforcement learning techniques. Furthermore, we consider the implementation of the above methods with constraints on channels and tasks, e.g., nonhomogeneous channels, blocked channels, periodic tasks, etc. We have modified the MCTS method to make sure that the constraints are met. The efficiency of the proposed methods are shown using simulation results.