Adaptive State-Constrained Vehicle Navigation and Positioning in GNSS-Denied Environments

In Global Navigation Satellite System (GNSS)-denied environments, conventional vehicle kinematics-based navigation methods struggle to accurately identify the vehicle’s motion state, leading to improper constraint application, degraded model effectiveness, and reduced positioning accuracy. Existing inertial measurement unit (IMU)-based motion recognition approaches are highly susceptible to sensor noise, road-induced vibrations, and other disturbances, making it difficult to reliably distinguish key motion states such as stationary, straight-line driving, and turning. These limitations significantly impair the effectiveness of constraint-aided inertial navigation frameworks. To address this issue, this article proposes an adaptive state-constrained vehicle navigation and positioning in GNSS-denied environments. A Mamdani-type fuzzy inference system fuses optical flow from an upward-facing fisheye camera, IMU measurements, and dual odometry (OD) to classify vehicle dynamics into stationary, straight-line, and turning states. State-specific error correction models are then applied, a zero velocity update (ZUPT) model mitigates error accumulation during stationary periods, while an adaptive nonholonomic constraint (NHC) model dynamically adjusts constraint thresholds during motion to maximize efficacy within the inertial navigation system (INS). Experimental results demonstrate that, compared with conventional fixed-threshold NHC, the proposed approach improves positioning accuracy by 72.6% on straight segments and 75.22% on curved sections. Both simulations and field tests validate the method’s robust high-precision performance in GNSS-denied environments.