Priority-based Task Scheduling in Dynamic Environments for Cognitive MFR via Transfer DRL

A radar resource management module in a cognitive multifunction radar manages the resources by first prioritizing and then scheduling the tasks. Apart from scheduling the tasks, the task scheduler of a cognitive radar requires the scheduling to be adaptable to the changing environment. We formulate a gen-eral model for the distributions of task parameters, specifically, task priorities and delay tolerance, to ensure priority-based task scheduling. We develop the use of transfer learning (TL) within a deep reinforcement learning (DRL) framework to address the challenge of adaptability to a varying environment. Our approach builds on using a Monte Carlo Tree Search (MCTS) aided by a deep neural network (DNN). We show that TL allows accelerated training by transferring the policy learned by training the D NN-based MCTS on initial parameter distribution (environment) to the policy required for a new environment. Our results show that the high priority tasks are least delayed and dropped with the new formulation, whereas TL ensures the respective adaptation to the dynamic environment.