Tire-Road friction coefficients adaptive estimation through image and vehicle dynamics integration

Current methods for identifying tire-road friction coefficient (TRFC) often struggle with accuracy and multiple interference issues. Addressing this, this paper proposes a novel TRFC estimation method that adapts to complex road conditions and performs self-diagnosis by the fusion of image and vehicle dynamics information. First, we compiled a comprehensive database comprising numerous images of extreme road surfaces along with corresponding vehicle dynamics data. A multi-temporal image fusion method was then developed. This method enables the segmentation and integration of road surface images for both the left and right-side wheels by tracking historical data, which ensures that each image is enriched with adequate road surface details. Subsequently, the road surface condition for each side is identified using a pre-trained transformer model. Following the image analysis, the TRFC is estimated in real-time using a residual adaptive unscented Kalman filter (UKF). High-confidence image estimation results serve as key adjusters for the UKF, enhancing estimation accuracy. The real vehicle test results demonstrate that our method accurately identifies the TRFC with enhanced robustness. Additionally, these tests confirmed the adaptive estimation method’s ability to detect faults and adjust steady-state values amidst model distortion, effectively maintaining accuracy despite image estimation declines caused by environmental interferences.