2023_programme: Adaptivity in Multi-Modal Underwater Mobile Networks

• Session: 19. Underwater Communications and Networking
  Organiser(s): Charalampos Tsimenidis, Paul Mitchell and Konstantinos Pelekanakis
• Lecture: Adaptivity in Multi-Modal Underwater Mobile Networks [invited]
  Paper ID: 1986
  Author(s): Tomasi Beatrice, Pottier Antony, Bouvet Pierre-Jean, Pelletier Leo-Paul
  Presenter: Tomasi Beatrice
  Abstract: In this paper, we consider a network of underwater mobile nodes, equipped with both acoustic and optical underwater communication systems. Both communication systems are proven to have complementary range versus rate capabilities underwater. We assume that the mobile nodes operate around an infrastructure, for example to monitor it, to surveil it, or to maintain it. This infrastructure is also equipped with the same set of communication systems. During their operations, the mobile nodes may work at short and long distances from the infrastructure. In addition, the environment physical properties of the water, such as temperature and turbidity affect the acoustic and optical propagation, thus making the communication systems unreliable or even unavailable for some periods of time and in some locations. For this reason, we first formulate a stochastic optimization problem whereby the achievable network throughput is the maximization function with respect to the chosen communication mode and the random channel conditions. \n\nWe then pose the adaptivity problem within the reinforcement learning framework (RLF). The paper proposes a reinforcement learning algorithm to be used by the master to decide which communication mode to select based both on current sensed environmental conditions and the communication performance statistics collected for a specific link. \n\nIn this RLF, the actions correspond to the choice of the communication. The system state corresponds to the current environmental conditions, and the link state statistics. The performance metric is the effective data rate. We compare the performance obtained with this algorithm with that obtained by conventional adaptive algorithms based for example on hysteresis thresholding. The evaluation will be based on numerical simulations of both the network scenarios and the underwater acoustic and optical communication channels. We then quantify the gain when both environmental measurements and link quality statistics are integrated into multi-modal underwater modems to enable effective adaptivity. \n
• Corresponding author: Dr Beatrice Tomasi
  Affiliation: NORCE Norwegian Research Center AS
  Country: Norway
  e-mail: