Optimization of channel allocation in wireless body area networks by means of reinforcement learning

We propose a novel algorithm for channel assignment in wireless body area networks. The proposed approach is based on the machine learning sub-domain known as reinforcement learning, and is named reinforcement learning — channel assignment algorithm (RL — CAA). RL — CAA interacts with the environment in an unsupervised way and selects the optimal frequency channel for the wireless sensor nodes. RL-CAA also takes into consideration the traffic load conditions and assigns the optimal number of channels to fulfill the minimum throughput requirement of the system. The algorithm is evaluated by means of a MATLAB simulator tool based on IEEE 802.15.6 specifications. It allows comparing our algorithm with classical static channel assignment algorithm. The proposed algorithm gives better error rate performance which is on average 30% better than static channel assignment. Since the error rate is reduced, the algorithm also proves better in terms of throughput by giving an average of 77.3 kbps over static channel assignment. Our proposed algorithm proves better in the terms of traffic load considerations and is more robust to the change in the load.