Laser SLAM loopback detection method

Abstract

The invention relates to a laser SLAM loopback detection method. The method comprises the following steps: 1, acquiring single-frame three-dimensional point cloud data and synchronous IMU information of the single-frame three-dimensional point cloud data; 2, segmenting the point cloud of the current frame into unit spaces, and respectively storing data of the unit spaces into a matrix so as to construct a first descriptor; storing the segmented point cloud data and IMU information into a B vector to construct a second descriptor; 3, constructing a KDTree by using the second descriptor of the historical frame in the step 2; 4, performing nearest neighbor search by using the second descriptor, and finding n candidate similar frames from the KDTree; and matching the candidate similar frame with the current point cloud according to the first descriptor, judging whether a closed loop exists in the candidate similar frame or not according to whether a matching result accords with a threshold value or not, and if so, obtaining the relative rotation change of the current frame and the loopback frame. The loop detection and map construction method with higher precision and higher robustness of the mobile robot in an outdoor scene and a scene with high local road section similarity degree is realized.

Description

technical field [0001] The invention relates to the field of mobile robots, in particular to a loop detection method for laser SLAM. Background technique [0002] With the development of the times, people pay more and more attention to the use of mobile robots in the actual environment. The research direction of the robot to automatically construct the surrounding environment map (Simultaneous Localization And Mapping, SLAM) and successfully locate itself is the core issue in the field of mobile robots. Many advantages, but the robot's ability to automatically map its surroundings and successfully localize itself is still very limited. At present, the SLAM method based only on lidar is prone to error drift when running for a long time when the robot moves quickly and the outdoor environment changes little. [0003] The loop closure detection method can effectively solve the problem of motion drift in the field of laser SLAM. Existing loop closure detection methods mostly s...

AI Technical Summary

Problems solved by technology

Existing loop closure detection methods mostly start with the feature vectors of point cloud features, such as histograms, etc., which are computationally intensive and have limitations when the outdoor scene changes little and the robot travels in the opposite direction to the same road section.

Images