Body state estimation method of legged robot based on multi-sensor information fusion

Abstract

The invention discloses a body state estimation method of a legged robot based on multi-sensor information fusion. The body state estimation method comprises the following steps: S1, separately acquiring sport information of various legs through sensors arranged in the target robot, and carrying out single-leg kinematics solution on the various legs according to the acquired information so as to obtain kinematics information of the various legs; fusing the kinematics information of support legs in the various legs so as to acquire a body initial state estimation result; S2, acquiring IMU (Inertial Measurement Unit) information through an IMU system, and carrying out inertial navigation modeling solution based on the acquired IMU information so as to acquire an inertial navigation solutionresult; and S3, fusing the acquired body initial state estimation result with the inertial navigation solution result to acquire a final body state estimation result. According to the body state estimation method disclosed by the invention, by combination of the advantages of IMU and the kinematics solution estimation way, drifting on the estimation result is avoided, and meanwhile, spike noise can be filtered and removed, and the precision of the estimation result is improved.

Description

technical field [0001] The invention relates to the technical field of state estimation of legged robots, in particular to a method for estimating the state of a legged robot body based on multi-sensor information fusion. Background technique [0002] Footed robot is a type of robot designed and manufactured in imitation of humans or animals, and its legs adopt a series multi-joint structure. According to different imitation objects and required tasks, legged robots can be divided into several categories: 1. Single-legged robots, which are used to imitate the single leg of mammals for bouncing function research; 2. Bipedal walking robots, which are used to imitate humans for stabilization Functional research of walking, trotting, and interaction; 3. Quadruped robots, used to imitate quadruped mammals such as dogs, cats, horses, and cheetahs for research on stability, compliance, load capacity, and environmental adaptability; 4. Multi-legged robots It is used to imitate rept...

AI Technical Summary

Problems solved by technology

Perceived information sources are generally various sensors, mainly including inertial navigation device (IMU), GPS positioning system, radar, camera, joint displacement sensor, joint force sensor, foot end force sensor, etc. Currently, for legged robots The state estimation is generally based on IMU's inertial navigation modeling solution or internal sensor-based kinematics modeling analysis. The inertial navigation solution method is accurate in a short time, but it will drift over time, and the drift mainly comes from the IMU device. Deviation, integral algorithm error, etc.; kinematic analysis will not produce drift, but there is a lot of noise, especially the spike noise caused by the vibration and impact of the robot when it lands, which cannot be eliminated by ordinary filtering algorithms

Due to the above shortcomings, inertial navigation calculation or kinematics analysis based on a single type of sensor information is difficult to accurately estimate the state of the robot body, and at the same time cannot meet the real-time control requirements of the robot for long-term motion

[0005] In order to solve the above problems, some practitioners proposed to use multi-source information fusion method to realize the state estimation of the legged robot body, but usually the estimated robot speed is obtained directly from the fusion of inertial navigation information and kinematic information, without considering the position estimation, and fusion The fusion of kinematic internal information is not considered in the process. This kind of fusion method is only at the first level of information fusion. There are still problems such as drift in the estimation method based on IMU and a large amount of noise in the calculation method based on kinematics. The estimation results The accuracy is still not high

Images