Research on Unmanned Driving Decision-Making Technology Based on IRF-SAC Model

Unmanned driving decision-making is an important field of unmanned vehicle research. Traditional unmanned driving decision-making technology cannot deal with complex road conditions. Therefore, a driving intention-risk composite field (Driving Intention & Risk Field, IRF) decision-making scheme is proposed based on the artificial field situation. Based on the attributes of destinations, vehicles, and road boundaries, they are modeled individually and consolidated in IRF as potential fields. Moreover, the developed IRF model collaborates with the Soft Actor Critic (SAC) maximum entropy model to establish an IRF-SAC unmanned driving decision-making inspection platform. This platform aids in the completion of the unmanned driving decision-making analysis. The high-speed and cross-road scenarios are used to test the effect of the model. During the high-speed scenario decision-making, the IRF-SAC model's accuracy has increased by 32.6% and 30.6% compared to the decision-making state machine model and the rapid expansion random tree model, respectively. In the intersection scene, the decision-making of the IRF-SAC model is more adaptable, leading to better timeliness. The research content has the potential to bring about new advancements and progress in the development and implementation of unmanned driving technology.