Abstract: With the increase of vehicle ownership, traffic congestion has become one of the most serious problems faced by urban road networks. To address this issue, this paper proposes a dynamic path planning method for mean field multi-agent vehicles based on graph attention mechanism, which achieves dynamic path planning through effective collaboration between intelligent vehicles and improves the traffic efficiency of the road network to a certain extent. Firstly, construct a graph based on the distance and influence relationship between vehicles; Secondly, using attention mechanism to aggregate the observed states of the current intelligent vehicle and neighboring intelligent vehicles; Finally, based on the mean field theory, the joint actions of surrounding intelligent agents are taken into account, and the Q-value function is updated to select the optimal route. This article is implemented on a simple custom road network and a road network in Binjiang District, Hangzhou. The results show that under different road traffic conditions, the average travel time of intelligent vehicles can be shortened.