Abstract：There are many types of driving-related risks with various characteristics and high randomness. The core problem to be solved is how to provide a unified solution to all possible risks and give full play to the driver's initiative to achieve active avoidance. Based on the reinforcement theory, we adopted the token economy technique, exerting positive reinforcement and negative punishment for shaping exogenous risks (e.g., a sudden drop from other objects) and endogenous risks (e.g., overspeed) avoidance behavior respectively. We found that the token economy can effectively increase safe driving behaviors regardless of whether facing an endogenous risk (e.g., overspeed) or an exogenous risk (e.g., a sudden drop from other objects). Moreover, risky-level matched token economy exhibited better performance than consistently high token economy and consistently low token economy. This project conforms to the international development trend of traffic safety management and has high applicability and feasibility.