Heuristically accelerated reinforcement learning for channel assignment in wireless sensor networks
Résumé
In wireless sensor networks (WSNs), multi-channel communication represents an attractive field due to its advantage in improving throughput and delivery rate. However, the major challenge that faces WSNs is the energy constraint. To overcome the channel assignment problem in an energy-efficient way, reinforcement learning (RL) approach is used. Though, RL requires several iterations to obtain the best solution, creating a communication overhead and time-wasting. In this paper, a heuristically accelerated reinforcement learning approach for channel assignment (HARL CA) in WSNs is proposed to reduce the learning iterations. The proposal considers the selected channel by the neighboring sender nodes as external information, used to accelerate the learning process and to avoid collisions, while the bandwidth of the used channel is regarded as an important factor in the scheduling process to increase the delivery rate. The results of extensive simulation experiments show the effectiveness of our approach in improving the network lifetime and performance.