Path Design for NOMA-Enhanced Robots: A Machine Learning Approach with Radio Map

A communication enabled indoor intelligent robots (IRs) service framework is proposed, where the non-orthogonal multiple access (NOMA) technique is adopted to enhance the data rate and user fairness. Build on the proposed communication model, motions of IRs and the down-link power allocation policy are jointly optimized to maximize the mission efficiency and communication reliability of IRs. In an effort to find the optimal path for IRs from the initial point to their mission destinations, a novel reinforcement learning approach named deep transfer deterministic policy gradient (DT-DPG) algorithm is proposed. In order to save the training time and hardware costs, the radio map is investigated and provided to the agent as a virtual training environment. Our simulation demonstrates that 1) The participation of the NOMA technique effectively improves the communication reliability of IRs; 2) The radio map is qualified to be a virtual training environment, and its statistical channel state information improves training efficiency by about 30%; 3) The proposed algorithm is superior to the deep deterministic policy gradient (DDPG) algorithm in terms of the optimization performance, training time, and anti-local optimum ability.