Research on Maglev Pipeline Trains Speed and Positioning Based on Improved Kalman Filtering Algorithm

This study presents the A novel Adaptive Kalman Filter Algorithm (APRKF) to address system noise, measurement noise, and sensor anomalies arising from train vibrations in multi-sensor fusion for maglev pipeline trains velocity and positioning. The algorithm focuses on rapid train vibration changes, estimating actual innovation sequence covariance using a constrained limited-memory approach. An adaptive factor is derived from the ratio of actual to theoretical innovation covariance. Further, an anomaly detection threshold restricts adaptive factor variations, constraining the scaled covariance matrix size and eliminating abnormal data. The adaptive factor dynamically corrects measurement noise and predicted state covariance in real-time, enhancing noise adaptability and mitigating filter divergence risk. This approach effectively mitigates the impact of sensor data distortion and noise variations on system state estimation. Simulation reveals that APRKF outperforms traditional Kalman Filter (KF) and Sage-Husa Adaptive Filter (SHKF) in velocity estimation, closely approximating true values, reducing oscillations, and efficiently filtering abnormal data to converge rapidly. For position estimation, APRKF excels by effectively reducing integration-induced errors. Comparative analysis demonstrates the superior accuracy and stability of APRKF in velocity and position estimation through standard deviation, mean squared error, and mean value metrics.