Abstract: To address the issue of significant positioning errors in traditional vehicle-mounted integrated navigation methods under high-dynamic environments, caused by the instability of GNSS signals and the drift of inertial navigation systems (INS), this study proposes an error compensation strategy for vehicle-mounted integrated navigation positioning based on an improved particle swarm optimization (PSO) algorithm. The improved PSO algorithm constructs a fitness function by integrating information variance and the root mean square (RMS) of positioning errors, and divides the population quality based on this fitness function. It incorporates a gravitational guidance mechanism to enhance collaboration and optimization between superior and ordinary particles, and sets nonlinear dynamic inertia weights to balance the algorithm's local exploitation and global exploration capabilities. Additionally, it improves error compensation performance by optimizing the gain of the Kalman filter. Validation tests were conducted under various paths, including straight paths and closed-loop paths with S-curves, as well as different operating conditions. The results demonstrate that, in complex path scenarios, the positioning root mean square error (RMSE) of the proposed method is reduced by 40% to 53% compared to the comparative methods. Compared to traditional methods, the heading angle accuracy is improved by up to 98.8%, significantly enhancing the performance and robustness of vehicle-mounted integrated navigation positioning error compensation.