Skip to content
Open access

Multi-objective optimization of lane-changing trajectories for autonomous vehicles based on deep reinforcement learning

Oct 2026 · Frontiers of Mechanical Engineering · 0 citations · 23 references

Abstract

The viability of autonomous driving technology largely depends on driving efficiency and driving safety, in which the accuracy of lane-changing trajectory planning is particularly important. Existing methods still suffer from low prediction accuracy and insufficient multi-objective collaborative optimization. A multi-objective optimized lane-changing trajectory planning method for autonomous vehicles based on the improved deep deterministic policy gradient is proposed. A trajectory prediction module is constructed by combining the temporal convolution network, the spatial multi-head attention mechanism, and a Transformer encoder–decoder, and the driving risk field is integrated into the multi-objective reward function. A performance comparison analysis was conducted between the trajectory prediction module and other methods, and the prediction root mean square errors for going straight, turning left, and turning right were 1.59, 1.19, and 1.27 m, respectively. The proposed method achieved obstacle avoidance and target advancement success rates of 98.92% and 98.26%, respectively, both of which outperformed the comparison method. In the lane-changing overtaking and lane-changing obstacle avoidance scenarios, the vehicle lane-changing correction was more timely, the vehicle speed fluctuation was smaller, and the front wheel angle control was smoother, all of which were better than those of the comparison method. The proposed lane-changing trajectory planning method effectively improves the trajectory prediction accuracy and planning performance, and can provide a theoretical basis for related research fields.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.