Robot Gait Motion Generation Using Human Capture and ZMP Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing two-leg walking robots do not accurately imitate the natural gait of humans, and their calculated gait differs significantly from human gait, lacking the ability to reproduce swing leg representations such as dancing.
Innovation Solution
A motion data generating system for robots that includes a test subject motion data acquiring unit, a representative point extracting unit, a scale converting unit, a ZMP setting unit, and an interpolation generating unit to generate motion data that imitates human gait by capturing and converting leg and waist motion data, setting a target zero moment point, and interpolating postures to satisfy the target ZMP.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the robot executes calculated gait control with mirror image relationship between swing leg and support leg, then stable walking control is achieved, but the gait differs significantly from natural human gait
Solution Approach 1:
The patent captures actual human gait motion data using motion capture equipment and directly applies it to the robot, copying natural human movement patterns rather than calculating synthetic gait. This allows the robot to reproduce authentic human gaits including dancing movements while maintaining stability through the inherent correctness of human motion patterns
Solution Approach 2:
The patent replaces the mechanical calculation-based gait generation system with a data-driven approach using motion capture and reproduction. Instead of computing gait parameters through mathematical models, the system directly imports measured human motion data, substituting mechanical calculation with empirical observation
2Reliability
If the robot uses adaptive upper body tilting according to target walking speed, then walking control is improved, but the overall gait still lacks natural human characteristics
Solution Approach 1:
The motion capture system captures comprehensive data including upper body, lower body, and gait information simultaneously. This universal data collection approach enables the robot to reproduce various types of human movement including walking, running, and dancing with natural coordination of all body parts, not just isolated limb movements
3Manufacturing precision
If the robot generates motion data through calculation and interpolation, then control precision is achieved, but the motion lacks smooth natural transitions
Solution Approach 1:
The patent performs preliminary motion capture of human subjects before robot execution, recording complete natural motion sequences including all transitional phases. This preliminary data capture ensures that smooth transitions are inherently present in the source data, eliminating the need for artificial interpolation that may create discontinuities
Data Source
AI summary
A motion data generating system for a robot including a waist and a leg, comprising: a test subject motion data acquiring unit that acquires captured waist motion data from motion of a leg separated from motion of a test subject; a representative point extracting unit that performs peak detection by smoothing the captured leg motion data; a scale converting unit that converts the capture leg motion data and the capture waist motion data at a representative point into a scale of the robot; a ZMP setting unit that sets a target zero moment point (ZMP of the robot; an interpolation generating unit that generates an attitude of the leg and the waist of the robot by interpolation; and a motion data generating unit that generates motion data in which the attitude of the leg and the waist is corrected so as to satisfy a target ZMP.