Robot non-trace quick simultaneous localization and mapping (SLAM) method based on multiple fading factors

Abstract

The invention provides a robot non-trace quick simultaneous localization and mapping (SLAM) method based on multiple fading factors, and belongs to the technical field of autonomous navigation of mobile robots. The robot non-trace quick SLAM method comprises the steps that system initialization is conducted to determine the initial position and posture an initial state estimation mean and a covariance matrix of a mobile robot at the k moment,; time updating is conducted, and the predicted position and posture and covariance of particles at the moment of k+1 are calculated; observation data arecorrelated with predicted observation data; the multiple fading factors are calculated according to a particle observation value, position and posture estimation means and covariance of the particlesare calculated in a measurement updating mode, Gaussian distribution functions are constructed, and sampling is conducted; resampling is conducted, a new particle set is obtained, and robot positionand posture estimation is completed; and according to the result after data correlation, a new guidepost is updated through an EKF algorithm, and map features are estimated. According to the robot non-trace quick SLAM method, the error introduced in the linearization process is lowered, the filtering accuracy of a system is improved, the strain adjusting ability of a filter when the system state changes suddenly is improved, and the robustness of the robot navigation and localization system is enhanced.

Description

technical field [0001] The invention belongs to the technical field of autonomous navigation of mobile robots, and in particular relates to a traceless and fast synchronous positioning and mapping method for robots with multiple fading factors. Background technique [0002] Autonomous navigation technology is a difficult and hot spot in the field of mobile robots. Leonard and Durrant-Whyte clearly pointed out in the literature that the problem of autonomous navigation of mobile robots can be solved by solving the following three problems. First of all, the first question "where am I?" (Where am I?), this question is a self-localization problem, the robot must determine its own current position to make decisions for subsequent tasks, which is the basis of the latter two questions. The second question "where is the destination? (Where is the destination?)", that is, the task target position of the mobile robot. The third question "How do I get there? (How do I get there?)" is...

AI Technical Summary

Problems solved by technology

First of all, the first question "where am I?" (Where am I?), this question is a self-localization problem, the robot must determine its own current position to make decisions for subsequent tasks, which is the basis of the latter two questions

The second question "where is the destination? (Where is the destination?)", that is, the task target position of the mobile robot

The third question "How do I get there? (How do I get there?)", that is, path planning and decision-making problems

Because the EKF-SLAM algorithm combines the pose vector and the map feature vector, the complexity of the algorithm will increase significantly under the conditions of more complex environments with many features, which seriously affects the real-time performance of the SLAM algorithm, so EKF-SLAM is only suitable for small environments. robot localization and mapping

[0004] The second mainstream SLAM algorithm is the FastSLAM algorithm based on particle filter. The SLAM problem is decomposed into two parts: pose estimation and map feature estimation through the conditional independence of probability theory. Particle filter is used for robot pose estimation, and extended Kalman filter is used for map analysis. Feature estimation, because of the existence of particle filtering, the problem of particle degradation cannot be avoided. The root cause of the degradation problem is that it cannot be sampled from the particle posterior distribution function, but obtained from the constructed proposed distribution function. The difference between the two leads to the final degradation problem

[0005] In practical applications, the established robot motion model and observation model may be different from the real model, and at the same time, uncertain disturbances in the environment may have a certain impact on the robot's own state, which will lead to the problem of a decrease in the filtering accuracy of the SLAM algorithm.

Images