Most of the existing shared-ride path algorithms only focus on the least time and the lowest cost, and lack the definition of the proportion of passengers’ willingness to share rides and the acceptable detour deviation range. Aiming at this phenomenon, a dynamic sharing policy model considering passenger preference was constructed. The passenger and the vehicle were respectively processed by the search module and the screening module was classified, and the vehicles that did not meet the requirements were deleted to reduce the time of dynamic real-time matching of passengers and vehicles, and then the insertion algorithm was used to obtain the optimal path matching. The results show that the algorithm can quickly and efficiently match passengers to the optimal vehicle that meets the requirements. Compared with the conventional algorithm, A carpooling algorithm considering passengers' willingness to ride together and detour deviation, the average waiting time of passengers is reduced by 7.0%; the total social benefits are increased by 44.7%; the dynamic carpool matching efficiency is increased by 34.2%. In addition, the length of the time window for passengers is changed can also significantly improve the ride sharing effect.