Abstract: ‍ ‍The mobile computation offloading scenario based on an edge server architecture avoids the network latency of core network links and improves the processing efficiency of computing tasks to a certain extent. However， in a scenario where mobile users are densely distributed， this computation offloading architecture may face a high cellular-network access delay. In order to improve the experience of mobile users， this study investigated the multi-user computation offloading problem based on D2D communication in a scenario where mobile users are densely distributed. Users were divided into two categories based on whether they had task processing needs. Users with intensive computing tasks to handle were defined as demand users， while users without tasks to handle and who could provide computing resources were defined as idle users. It was assumed that the task of a demand user could be offloaded to one idle user in a binary manner for auxiliary processing. Jointly considering the benefits to demand users and idle users， we constructed a utility function for each user and formulated corresponding optimization problems. A computation offloading and resource allocation algorithm was proposed based on the Stackelberg game and genetic algorithm to improve each user’s utility. The algorithm had a two-layer structure， with the inner layer optimizing the computing resource rental unit price decisions of demand users and computing resource allocation decisions of idle users based on the Stackelberg game. It could also be proved that a unique Nash equilibrium existed for the strategies of the demand and idle users. The outer layer used a genetic algorithm to optimize the task offloading decisions of the demand users. The optimal task offloading decisions of the demand users could be solved based on the feedback results of the inner layer. To verify the performance of the algorithm， an experimental simulation was conducted in this study. The simulation results showed that compared with other schemes， the algorithm proposed in this paper could effectively improve the utilities of demand and idle users. At the same time， in order to verify the effectiveness of our proposed algorithm， we also conducted a simulation analysis of the impacts of different parameters.