Communication-Efficient Reinforcement Learning in Swarm Robotic Networks for Maze Exploration