AUV-aided computing offloading for multi-tier underwater computing: A Stackelberg game learning approach

The Internet of Underwater Things (IoUT), which connects underwater devices, such as sensor clusters and Autonomous Underwater Vehicles (AUVs) to monitor the underwater environment, is an important component of ocean activities. Due to the limited computing resources of Cluster Heads (CHs) of sensor clusters, AUVs can provide additional computing power for CHs. In this paper, an AUV-aided offloading framework is proposed for multi -tier underwater computing, where the Edge Intelligence Service Platform (EISP) manages computing resources. The EISP provides computation services for CHs through multiple AUVs, buoys, and a Low Earth Orbit (LEO) satellite at a certain price. Afterwards, we formulate the game process between CHs and the EISP as a Stackelberg game. Considering the impact of communication energy consumption and communication delay in the process of task offloading, we establish the optimization models of EISP and CHs. Then, the backward induction method is applied to prove that the proposed game has a unique Stackelberg Equilibrium (SE). Furthermore, a Game Solving Algorithm based on Improved Differential Evolution (GSA_IDE) is designed to achieve SE in the proposed game. Finally, the simulation results show that the proposed GSA_IDE can achieve SE and outperform other algorithms.