A Q-Learning-Based Broadcast Scheduling Approach for Multi-Hop Wireless Sensor Networks

Broadcast scheduling is a core problem in multihop wireless sensor networks (WSNs), especially in time division multiplexing (TDMA) systems, where minimizing the number of time slots to cover the entire network is crucial for energy efficiency and latency. In this work, we propose a novel reinforcement learning-based approach using Q-learning to solve the broadcast scheduling problem in TDMA-based WSNs, in which the modeled state is the set of nodes that have received the message. Specifically, we transform the problem into a sequential decision-making process. Herein, the agent proceeds with action by selecting a single transmitting node at each time slot, then the network applies a greedy rule on state transition to maximize the immediate coverage. Simulation results show that the proposed algorithm can learn an efficient scheduling policy, achieve comparable or superior performance to classical heuristic methods, and show high potential in future smart sensor networks.