Research on dynamic lane change trajectory planning method integrating environmental potential field
In the research on intelligent vehicle autonomous driving technology,the lane change process has always been a critical issue.While traditional lane change models can effectively plan paths in simple scenarios,they often face challenges in complex and dynamic road environments,such as excessive curvature changes in the trajectory and sudden speed variations,which can affect the smoothness and safety of the vehicle.To address these issues,this study analyzes and compares existing traditional vehicle lane change models and proposes a dynamic planning method based on time-space sampling,integrated with an environmental potential field.By introducing the environmental potential field and an improved trajectory planning algorithm,the stability and safety of intelligent vehicles during lane changes are further enhanced.First,an environmental potential field is established based on an improved two-dimensional normal distribution formula,which includes fixed obstacle potential fields,dynamic obstacle potential fields,and road boundary potential fields.The total potential field value at a given point on the road is used as a measure of the passability of that location.Next,using the transformation relationship between the Frenet coordinate system and Cartesian coordinates,an independent integral system is constructed that decouples the vehicle's lateral and longitudinal motion,greatly reducing the computational complexity of trajectory planning while improving calculation efficiency.The lane change endpoints are sampled in both time and space dimensions,and a set of quintic polynomial trajectories is generated by incorporating boundary conditions into quintic polynomial curves.Further,velocity,acceleration,and curvature constraints are applied to filter out trajectories that do not meet kinematic requirements.A multi-objective cost function is constructed,considering trajectory safety,the degree of deviation from the target lane,planning time,and trajectory comfort.The total cost of each candidate trajectory in the lane change trajectory set is calculated,and priority scores are assigned to the candidate trajectories.The motion trajectories of surrounding vehicles are predicted,and a vehicle OBB(Oriented Bounding Box)collision model is constructed.The cost values of the candidate trajectories are compared,and combined with the collision detection model,the quintic polynomial trajectory with the minimum cost and no collision is selected as the optimal trajectory for the current planning.To validate the effectiveness of the proposed method,an S-shaped dual-lane change simulation scenario was constructed,and active lane change simulations were carried out.During the simulation,not only were the effects of different planning densities on vehicle state changes analyzed,but the optimal trajectory was also updated in real-time based on the vehicle's current state.Simulation results show that,due to the incorporation of the environmental potential field,intelligent vehicles can detect obstacles earlier than traditional methods and react promptly based on the obstacle potential field values.Additionally,through the optimization of key parameters such as the cost function and dynamic planning density,the obstacle avoidance performance of intelligent vehicles in complex road environments is further improved.Specifically,during the lane change process,the curvature of the vehicle trajectory remains consistent with the road curvature,with fluctuations not exceeding±0.02 m-1,maintaining high smoothness and continuity overall.Furthermore,the speed of the host vehicle does not experience significant changes during the lane change,ensuring a safer and more comfortable lane change process.
intelligent vehiclefrene coordinatestrajectory planninglane changecost function
国家科技期刊平台
NSTL
万方数据
