Adverse weather conditions have a significant detrimental impact on autonomous driving. Incorrect perception information caused by adverse weather can lead to erroneous inputs in the decision-making module. Therefore, it is crucial to develop a decision-making model that can resist the influence of erroneous information. In this paper, we propose a lane-changing decision model that combines the self-attention mechanism and Deep Q-Network (DQN) to filter out irrelevant data and focus attention on vehicles with higher risks. Through experimental simulations, it is observed that the model effectively avoids collisions and ensures vehicle safety. A comparison with other algorithms demonstrates that even in situations where input data is incorrect, the safety rate of the model remains above 95%.
Decision Model of Lane Changing for Autonomous Vehicles Considering Weather Influence
28.10.2023
1483536 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Investigation of lane-changing decision-making models for autonomous vehicles
British Library Conference Proceedings | 2021
|XGBoost Lane-Changing Decision Model Considering Driving Style
Springer Verlag | 2024
|