A slotted CSMA based reinforcement learning approach for extending the lifetime of underwater acoustic wireless sensor networks

Underwater acoustic wireless sensor networks (UA-WSNs) are capable of supporting underwater missions. Due to the harsh environment, replacing or recharging battery for underwater sensors are difficult or costly, thus UA-WSN systems must be energy efficient. Although a large number of energy efficient schemes have been proposed for terrestrial wireless sensor networks, the fundamental differences between underwater acoustic channel and its terrestrial counterparts make those schemes perform poorly in underwater acoustic communications. In this work, we present an energy efficient architecture for UA-WSNs, which employs a reinforcement learning algorithm and a slotted Carrier Sensing Multiple Access (slotted CSMA) protocol. Due to the reinforcement learning algorithm, the proposed system is capable of optimising its parameters to adapt to the underwater environment after having been deployed. Simulation results show that the lifetime of the network is extended significantly with the proposed architecture by lowering the number of collisions and retransmissions of data packets.