Reinforcement learning for delay tolerance and energy saving in mobile wireless sensor networks

Reinforcement Learning (RL) has emerged as a promising approach for improving the performance of Wireless Sensor Networks (WSNs). The Q-learning technique is one approach of RL in which the algorithm continuously learns by interacting with the environment, gathering information to take certain actions. It maximizes performance by determining the optimal result from that environment. In this paper, we propose a data gathering algorithm based on a Q-learning approach named Bounded Hop Count - Reinforcement Learning Algorithm (BHC-RLA). The proposed algorithm uses a reward function to select a set of Cluster Heads (CHs) to balance between the energy-saving and data-gathering latency of a mobile Base Station (BS). In particular, the proposed algorithm selects groups of CHs to receive sensing data of cluster nodes within a bounded hop count and forward the data to the mobile BS when it arrives. In addition, the CHs are selected to minimize the BS tour length. Extensive experiments by simulation were conducted to evaluate the performance of the proposed algorithm against another traditional heuristic algorithm. We demonstrate that the proposed algorithm outperforms the existing work in the mean of the length of a mobile BS tour and a network's lifetime.