Situation Assessment for Lane-Changing Risk Based on Driver’s Perception of Adjacent Rear Vehicles

With advancements in sensors and communication technologies, hazardous lane-changing can be readily identified, by analyzing driving behavior and vehicle data. Most traditional approaches to assess hazardous lane-changing depend on surrogate safety measures, such as gap acceptance, minimum safe distance, and time-to-collision (TTC). Unlike these measures, which focus on statistical data or vehicle kinematics analysis, this study proposes a new method for assessing risk when a vehicle changes lane, based on the driver’s perception of adjacent rear vehicles. First, a driver model based on virtual spring theory is used to build a risk perception model. The proposed model realizes comprehensive assessment of risk by car-following parameter amendment when the adjacent vehicle cuts in. Second, model parameters are identified using the recursive maximum likelihood method. Third, the threshold-determining mechanism is established using a first-order grey prediction model, the threshold is adjusted online based on the driver’s braking data, and the information entropy is used to judge the rationality of threshold and search the optimal threshold. Finally, offline simulation of real vehicle test data is adopted to verify the effectiveness of the proposed method. Results show that the algorithm can adapt to individual differences, and performs better than TTC in evaluating hazardous lane-changing.