With the increasing quantity of private cars, the traffic congestion problem is getting more and more seriously. Cooperative control strategy for connected and autonomous vehicles (CAVs) has been considered as an effective way to improve road traffic flow quantity. This paper extends the thought which uses deep reinforcement learning (RL) to create controllers for mixed-autonomy traffic. Specifically, based on Bayesian network, a formation strategy is proposed to decide the number and formation of the vehicle fleet. Then the control policies of the single vehicle in the fleet are derived by deep reinforcement learning. Simulation has been carried out to show the performance of the proposed algorithm.

CICTP