Fuzzy Evaluation of an Intelligent-vehicle Driving State Based on a Vehicle-road Collaborative Information Fusion

This study monitored and judged the real-time state of intelligent vehicles running on the road, based on the vehicle's developed onboard sensing system, which obtains information on the vehicle's interior and surrounding environment through a visual sensor, LiDAR, GPS positioning, an onboard sensor system, and a vehicular bus. Vehicle-to-everything (V2X) communication equipment was applied to obtain the traffic information transmitted by a ray-vision integrated machine for roadside, roadside sensors, and meteorological sensors, which was transmitted to a cloud server through V2X communication equipment and a 4G communication module, and used to establish a fuzzy evaluation model. A fuzzy reasoning algorithm, based on credibility, was applied to fuse the environmental and traffic information, and the state of the running vehicles was evaluated. First, the fuzzy evaluation set and parameter membership function for the vehicle driving state were established, the membership degree of each parameter was calculated, and a typical driving-state evaluation parameter dataset was established for each parameter of the vehicle. Second, a fuzzy hypothetical-reasoning method was applied to establish a fuzzy rule database with credibility and threshold based on typical datasets. A fuzzy relation matrix library, corresponding to each rule of the rule database, was established using the Mamdani method. Using the information from the vehicle device and the roadside unit as input, fuzzy reasoning with credibility was conducted using the rules of the rule database. Then, taking the similarity as the matching degree, a threshold was set for the reasoning rule, and the reliability of the conclusion was calculated, considering that the evidence is not equal to the antecedent of the rule. When resolving the conflict of conclusions, the conflict-resolution strategy was to select a conclusion with high reliability. Finally, the reliability of the conclusion was verified using the matching degree, and the algorithm was verified by experiments on real-time vehicles in multiple road scenes. The experimental data analysis shows that the evaluation of the running vehicle state by the algorithm is consistent with the real vehicle state. Moreover, it can sound an alarm for an unsafe vehicle state and intervene in a running state, which has significant practical application to ensure driving safety. © 2022 Xi'an Highway University. All rights reserved.