Continuous Trust Estimation Method Based on Kalman Filtering in Human-Machine Co-Driving

Subjective trust and active takeover behavior characteristics are two important aspects of trust performance in human-machine co-driving cars. However, trust is a subjective, abstract concept that changes over time and is difficult to measure directly. At present, there is a lack of quantitative research on objective trust and dynamic estimation of continuous trust under the influence of different independent variables, which inhibits its further use and development. This study adopts a continuous objective trust estimation method based on driving behavior, which mathematically describes the continuous measurement problem of objective trust and extracts driving behavior indicators in different traffic event research segments. The objective trust state space equation is established, and the objective trust estimation model is constructed based on the Kalman filter algorithm. Through model parameter definition and model verification, the estimation results and subjective trust are divided into trust levels through K-means clustering to verify the accuracy of the model. The results show that this method can estimate the continuous change in driver trust, and the error with the driver's subjective trust gradually decreases over time. This study improves the estimation method of trust estimation by combining theoretical calculations with experimental results, fills the gap in the continuous measurement of trust, and has a guiding role in the future research and development of trust in human-machine co-driving cars.