Law compliance decision making for autonomous vehicles on highways

As autonomous driving advances, autonomous vehicles will share the road with human drivers. This requires autonomous vehicles to adhere to human traffic laws under safe conditions. Simultaneously, when confronted with dangerous situations, autonomous driving should also possess the capability to deviate from traffic laws to ensure safety. However, current autonomous vehicles primarily prioritize safety and collision avoidance in their decision-making and planning. This may lead to misunderstandings and distrust from human drivers in mixed traffic flow, and even accidents. To address this, this paper proposes a decoupled hierarchical framework for compliance safety decision-making. The framework primarily consists of two layers: the decision-making layer and the motion planning layer. In the decision-making layer, a candidate behavior set is constructed based on the scenario, and a dual layer admission assessment is utilized to filter out unsafe and non-compliant behaviors from the candidate sets. Subsequently, the optimal behavior is selected as the decision behavior according to the designed evaluation metrics. The decision-making layer ensures that the vehicle can meet lane safety requirements and comply with static traffic laws. In the motion planning layer, the surrounding vehicles and the road are modeled as safety potential fields and traffic laws potential fields. Combining the optimal decision behavior, they are incorporated into the cost function of the model predictive control to achieve compliant and safe trajectory planning. The planning layer ensures that the vehicle meets trajectory safety requirements and complies with dynamic traffic laws under safe conditions. Finally, four typical scenarios are used to evaluate the effectiveness of the proposed method. The results indicate that the proposed method can ensure compliance in safe conditions while also temporarily deviating from traffic laws in emergency situations to ensure safety.