A Dual-Event-Driven Fusion Estimation Method for Road Friction Coefficient Based on Image and Vehicle Dynamics

The road friction coefficient reflects the tire's ability to grip the road surface and directly affects a vehicle's acceleration, braking, and steering performance. Therefore, real-time and accurate detection of the road friction coefficient plays an irreplaceable role in ensuring safe vehicle operation and enabling intelligent traffic management. To address limitations such as the excitation constraints of single dynamic estimators, the low accuracy and high computational demand of standalone image recognition, this paper proposes a dual-event-driven fusion estimation framework that integrates visual perception and dynamic information. Deep fusion is performed at the data, model, and decision levels. In the image recognition module, kinematic constraints are applied to image data, and a Multi-Patch Fusion Decision Network (MFDNet) is designed. Additionally, a dual-event-driven (DED) mechanism is introduced by defining a visual confidence factor and a state evaluation metric, which jointly trigger decision-level fusion.During model fusion, a Dual-Event-Driven Adaptive Cubature Kalman Filter (DED-ACKF) is developed. It dynamically adjusts the filter gain based on residuals between image-based and dynamics-based estimation results using a time-varying scaling factor. Simulation and road test results demonstrate that the proposed estimation system achieves both predictability and accuracy, demonstrating robustness across all driving conditions.