A Multisensor High-Precision Location Method in Urban Environment

The integration of global navigation satellite system/inertial navigation system (GNSS/INS) technology is prevalent in unmanned vehicles. However, GPS signals in complex environments, such as tunnels and forests, often experience signal degradation or loss, reducing vehicle positioning accuracy. Furthermore, the effectiveness of multisensor fusion strategies based on filtering declines sharply in the presence of non-Gaussian noise. This article proposes a multisensor fusion positioning method that combines visual/INS and INS/GNSS technologies to overcome these challenges. The GPS/inertial measurement unit (IMU)/visual sensor data are preprocessed using the adaptive maximum correntropy criterion extended Kalman filter to obtain a fusion result as a benchmark. Specific data are then extracted from the output position and posture of the GPS/IMU and visual/IMU and fused twice using the Bernoulli formula to obtain the vehicle's position and posture information. We use classic navigation data from Canada and Katwijk to validate our method. When encountering both non-Gaussian noise and GPS signal anomalies, our method significantly reduces longitude and latitude errors by 22% and 40%, respectively, compared to the GPS/IMU/visual method, enabling precise positioning in urban environments.