Performance Test and Evaluation of Vehicle IMU Installation Angle Calculation Considering Road Surface State

Wheel speed odometry and nonholonomic constraints are two commonly used methods to suppress error divergence in Global Navigation Satellite System (GNSS)/Inertial Navigation System (INS) integrated navigation systems during prolonged GNSS signal outages. Accurate in-vehicle inertial measurement unit (IMU) mounting attitude is a prerequisite for applying wheel speed odometry and nonholonomic constraints. Traditional mounting angle calibration methods perform well on ideal road surfaces, but the validity of their core kinematic constraints heavily relies on ideal tire-ground contact conditions. In actual complex driving environments, different road surface conditions can disrupt the fundamental assumptions of these constraints by inducing abnormal vehicle motions, leading to degraded performance or even failure of online mounting angle estimation algorithms. To investigate the impact of different road conditions and driving states on IMU mounting attitude estimation algorithms, this study conducts simulation analyses and in-vehicle experiments focusing on three scenarios: road bumps, prolonged small-angle turns, and short-duration large-angle turns. By comparing the performance of velocity observation models and position observation models across these scenarios, the influence of different road surface conditions on the accuracy and robustness of IMU mounting angle estimation is analyzed. Experimental results show that in road bump scenarios, the position observation model achieves higher estimation accuracy than the velocity observation model, improving pitch mounting angle estimation by 76% and heading mounting angle estimation by 67%. In prolonged small-angle curve driving scenarios, the velocity observation model performs better, enhancing pitch mounting angle estimation by 32% and heading mounting angle estimation by 57%. However, in large-angle sharp turn scenarios, rapid changes in vehicle heading generate significant lateral velocity and displacement, which violate the constraint conditions. Therefore, in such high-dynamic scenarios, enhanced dynamic constraints and error compensation are required to achieve stable and accurate mounting angle estimation.