Multi-sensor Fusion-based Quadruped Robot Mapping Method

The quadruped robot has received extensive attention because of its good passability,and accurate map construction is one of its key prerequisites for walking and navigation on complex roads.However,the walking mode of the leg-and-foot robot causes the body and the sensor to shake significantly.The traditional image-based and inertial measurement unit(IMU)-based method is prone to the problem of odometer failure or posture drift required for mapping.The problem of noisy map construction.In this paper,the leg odometer is innovatively introduced,and the Kalman filter algorithm based on the error state is used to achieve a more robust attitude estimation.Secondly,the method based on aggregated information is used to filter the point cloud,and the concept based on outlier is proposed.Noise compensation provides a more accurate and complete point cloud map.The specific research contents are as follows:1.The four-legged robot shakes violently,which makes the visual odometers have a large gap between the collected image frames,which leads to difficulty in feature extraction and matching,and the IMU that is often introduced is prone to drift.In view of the above problems,this article introduces the leg model based on the motion model combined with the visual odometer to observe the IMU error information,modeling the error state as offset error and walking noise and constructing a prediction model to form a Kalman filter based on the error state The algorithm increases the robustness and accuracy of pose estimation.2.In view of the problem that the motion of the robot will blur the collected image,which causes a lot of noise in the traditional mapping method,this paper proposes a point cloud filtering method based on aggregate information to estimate the degree of aggregation of a single frame of point cloud.In order to remove the noise with higher dispersion and aggregation,noise filtering is realized.In addition,in order to avoid the loss of information caused by too many filtered points,this paper uses the aggregated information to propose the concept of outliers,and uses a single frame point cloud noise direction estimation method based on point cloud principal component analysis technology to quantify the deleted noise.make up.3.Build a simulation platform for quadruped robots,and collect sensor data for fusion algorithm verification.First,build a physical collision model of a real elbow-knee type quadruped robot in Solidworks and introduce Webots simulation software to establish a walking environment,add parameters such as an open power engine and friction and gravity,then write a stable small running controller,and finally build a sensor virtual Collect the information after adding the noise variables of various sensors to the model.This article first analyzes the mapping system and key technologies applied to the quadruped robot to form the content framework of this article.Then,this paper uses sensors commonly used in quadruped robots to establish three types of sensors and establish an error state Kalman filter.Then,a filtering method based on aggregated information is proposed to filter the established point cloud image,and the concept of outlier is established to carry out effective noise compensation.Finally,a simulation platform was built to realize the trot gait of the quadruped robot,collect information and conduct experiments.