The invention provides a multi-lane space-time trajectory optimization method for an intelligent network connection vehicle. The space-time trajectory optimization algorithm based on a reinforcement learning algorithm is designed, so the optimal trajectory can be quickly matched. The algorithm comprises the following steps of (1) optimizing a space-time track, taking the current position and speedof a vehicle and a target driving-out lane, taking a time period as an input, and taking a set of vehicle accelerations as an output; and (2) optimizing multi-lane cooperative lane changing, taking the current position and speed of the vehicle and the position and speed threatening the vehicle of the target lane as input, and taking a vehicle acceleration set as output. That is to say, after a vehicle initiates a lane changing request, the track of the vehicle cooperative lane changing process can be matched through reinforcement learning, and after lane changing is completed, the space-timetrack at the moment is matched through reinforcement learning to achieve the multi-lane track optimization process. The method can optimize and generate the space-time track of the passing vehicles inthe road section in real time according to different road environments and traffic states, improves the mutual cooperation capability of the vehicles, improves safety of the passing road section andvehicle passing efficiency of the intersection, reduces the energy consumption level of the vehicles, and improves traffic safety of the road section in order to guarantee the traffic safety of the road; and a new solution and a theoretical basis are provided for improving the travel efficiency.