Dynamic Slope Measurement and Positioning Accuracy Optimization Using Improved Particle Filter in Complex Environments

Abstract

In complex environments, the drastic terrain changes and the superposition of sensor measurement noise result in insufficient accuracy of dynamic slope estimation by traditional positioning methods, thus limiting the overall positioning performance. To solve this problem, this paper adopts an improved particle filter (PF) algorithm based on the temporal attention mechanism. By applying a weighted modeling mechanism of historical observations, the particle importance weight distribution is reconstructed to enhance the responsiveness of state estimation to nonlinear dynamic disturbances. In the state prediction stage, a terrain-aided inertial navigation (TAIN) correction model is introduced to guide the particle distribution to converge to the terrain-consistent area with the slope prior information. At the same time, a Bayesian estimation framework integrating nonlinear terrain constraints is constructed to realize the joint reasoning of slope and position state. Experimental results show that this method achieves a root mean square error of 1.27° in slope estimation, reduces the confidence boundary of positioning error to 1.9 m, and reduces the particle degradation rate to 31.7%. This method achieves efficient coordination between dynamic measurement and nonlinear estimation, significantly improving navigation accuracy and reliability in complex environments.