Research On 3D State Sensing And Autonomous Localization Of Quadruped Robot In Unstructured Environment

Quadruped robot is widely used in detection,rescue,explosion-proof,transportation and other aspects.It can surmount obstacles in the process of movement and penetrate into the harsh unknown environment that human cannot enter.In order to complete the task autonomously,quadruped robot must be able to provide accurate information of its own state,and at the same time be able to sense the 3D environment and build a map.However,there are random errors and calculation drift in inertial sensor,and impact vibration in foot sensor leads to calculation error,which further affects the accuracy of state parameter calculation.For these shortcomings,this paper studies the perception and state estimation methods of quadruped robot in 3D environment,and carries out corresponding experimental verification.By analyzing the working requirements and motion characteristics of quadruped robot,combing the system and designing the environment sensing and positioning system which combines encoder,radar,inertial navigation,etc.This paper analyz es the system ontology,simulation platform,system software and hardware framework of quadruped robot,and then determines the requirements and performance indicators of the perceptual positioning system;designs the specific sensor scheme of the perceptual positioning system,including sensor selection,measurement model derivation,sensor error calibration and data processing,and introduces error evaluation calculation.The design scheme gives a clear,reasonable,efficient and complete perception and positioning system framework of quadruped robot.Aiming at the walking and moving mode of quadruped robot,a 3D st ate estimation method combining internal sensor and inertial measurement unit is proposed.In order to improve the accuracy of the odometer,t he error Kalman filter algorithm is used to fuse and compensate the error state value;the simulation environment of quadruped robot is built and the accuracy and motion performance of the odometer are verified.The scheme can provide the state information of quadruped robot reliably and verify the feasibility of leg mileage calculation method.Based on the fusion of multi-sensor,such as laser,foot force sensor,inertial measurement unit and encoder,a method of map construction and localization of quadruped robot in unstructured environment is proposed.In order to fully express and utilize the environmental information,different types of environmental maps are established;the global pose estimation is carried out by combining the external sensors with high accuracy and the global map;in order to ensure the balance of frequency and accuracy,the high-frequency and high-precision odometer information is output by Kalman filter algorithm;the mapping and positioning experiment are carried out in the simulation environment.The scheme can realize the fusion of redundant sensor data,and further improve the odometer frequency and accuracy.Finally,the experimental system of environment perception and location of quadruped robot is built and verified.In order to reduce the measurement error,the state estimation method using internal sensors and the positioning method based on multi-sensor fusion are verified.The experimental results show that the algorithm is effective and has high accuracy and frequency,and can provide reliable state information for motion control and navigation.