| Due to the huge existing stock of traditional vehicles,there will inevitably be a period of coexistence with Human Driven Vehicles(HDV)during the popularization of Connected and Autonomous Vehicles(CAV)in the process of vehicle networking and CAV technology.The automatic driving dedicated lane reduces the occurrence of mixed flow by physically separating CAV and HDV traffic flow,thus constructing a pure CAV traffic environment and fully exerting the advantages of CAV.CAVs have the need to change lanes when entering or exiting the automatic driving dedicated lane,and CAV vehicles face a complex driving environment of lateral mixed flow when changing lanes.Existing research mostly considers the application of CAV’s collaborative control ability in this environment,but ignores the coupling relationship between CAV and the preceding and following HDVs,and there is still room for improvement in the efficiency of traffic operation.At the same time,the existing research separates the lanechanging decision-making and lane-changing control,and the level of integration can be further improved.To address these issues,this paper constructs a vehicle cooperative lane-changing control method based on the Model Predictive Control and Deep Learning(MPC-DL)coupling framework,considering the coupling relationship between HDVs and CAVs in mixed flow environments,reducing the impact of CAV lane-changing on HDVs,and improving traffic flow efficiency under automatic driving dedicated lane conditions.This paper works as follows:(1)To analyze the coupling effect of CAVs on HDVs,this paper designs a short-term prediction method for HDV trajectories based on a kinematic model and LSTM to analyze the lane-changing behavior influence zone.For different driving environments of HDVs,trajectory prediction models are respectively constructed for the following state and cooperative lanechanging state.Based on the demand for constructing trajectory prediction models,the NGSIM dataset is preprocessed,and driving data for the two states mentioned above are extracted.Finally,the prediction performance of the constructed trajectory prediction model is analyzed,and the results show that it can meet the data requirements of the MPC-DL coupling control model for HDV trajectories.(2)To meet the cooperative control requirements of CAVs taking into account the degree of HDV influence,this paper constructs a lane-changing control model based on the MPC-DL coupling,achieving active and passive cooperative control between CAVs and HDVs.This paper adopts a lateral control process based on a sine curve and defines the constraint space of the model predictive control part by specifying the constraints on vehicle dynamics,safety,and performance during the lane-changing process.By considering factors such as longitudinal follow efficiency,traffic flow stability,and passenger comfort,a target function is constructed.Based on the CAV-HDV coupling effect established in previous research,this paper establishes an MPC-DL coupling control model,which generates the optimal control output in real-time through the rolling solution of the model predictive control algorithm and neural network.(3)To address the feasibility of lane-changing control and the decision-making requirements for lane-changing,this paper conducts relevant theoretical analysis.Regarding the feasibility of lane-changing,this paper analyzes and demonstrates the initial gap feasibility and process sequential solvability separately.In terms of the initial gap feasibility,this paper analyzes the conditions that must be met for lane-changing to be feasible based on the initial state of the participating lane-changing vehicles.In terms of the process sequential solvability,this paper mathematically proves that the process can be solved sequentially by meeting the threshold requirements of the undetermined coefficients in the safety constraints.Regarding the lane-changing decision-making requirements in this study,this paper proposes a method for selecting the optimal lane-changing gap based on the initial utility function of the control algorithm to solve the gap selection problem in a driving environment containing multiple feasible lane-changing gaps.(4)This article conducted simulation experiments based on the Python language environment and the IPOPT solver,designed four lane-changing scenarios considering the different attributes of the preceding and following vehicles on the target lane,and verified them through multiple experiments.A comparative experiment was also designed.The experimental results show that the proposed method in this paper can effectively reduce the impact of HDV caused by lane changing while ensuring the safe and comfortable lane changing of CAV.The lane-changing scenarios designed in this paper include multiple feasible lane-changing gaps,and the correctness and effectiveness of the proposed optimal lane-changing gap selection method are demonstrated through simulation experiments. |
