Positioning method based on improved iterative volume particle filtering algorithm

Abstract

The invention discloses a positioning method based on an improved iterative volume particle filtering algorithm. The method comprises the following steps: firstly generating a particle set composed ofa plurality of particles; then performing multiple times of iteratively updating on the particle set based on one or more road signs observed in a known environment, wherein the state mean is updatedfor multiple times during each iteration process, but the state covariance is updated once only after the last iteration of a road sign; finally completing the importance sampling of each particle based on the estimated value of the obtained updated state mean and the estimated value of the state covariance, and achieving the positioning of the mobile robot. The algorithm uses the volumetric numerical integration principle to estimate the Gaussian prior nonlinear transfer density, so that the particle set is concentrated in the tail of the observed likelihood function, which overcomes the problem of particle set degradation in the traditional particle filter localization algorithm, and can achieve high-precision and high-efficiency attitude determination and positioning of mobile robots with less particles.

Description

technical field [0001] The invention relates to a positioning method, in particular to a positioning method based on an improved iterative volume particle filter algorithm. Background technique [0002] Robot positioning technology is the use of motion information and environmental observation information by mobile robots to estimate their own position and attitude in the environment. This technology is widely used in path planning, obstacle avoidance, etc., and is also the most basic, most important question. Among them, the Monte Carlo Localization (MCL) algorithm based on particle filtering is one of the mainstream methods in the current academic circles to study the positioning of mobile robots. MCL uses discrete particle sets to approximate the posterior probability distribution of mobile robot poses, and is suitable for A posteriori reliability estimation for systems with arbitrary system dynamics and arbitrary noise distribution. In the Monte Carlo localization algo...

AI Technical Summary

Problems solved by technology

For example, when the iterative extended Kalman filter method deals with nonlinear problems, it converts the nonlinear problem into a linear problem by performing first-order linear truncation on the Taylor expansion of the nonlinear function and ignoring the remaining high-order terms, but in the process The truncation error is introduced, which is likely to cause the filtering to occur; the volumetric Kalman filter uses the third-order sphere-phase diameter volume rule to approximate the Gaussian integral in the nonlinear Gaussian filter, avoiding the linearization of the nonlinear function by the iterative extended Kalman algorithm However, when the measurement equation is a nonlinear function, when the volumetric Kalman-based particle filter localization algorithm uses the Gaussian approximation filter based on the linear Bayesian estimator to approximate the state posterior probability density function, the effect achieved is not ideal

Images