Robot Behavior Generation With Constraint Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for converting human body motion into robot behavior fail to optimize motion indices with constraint conditions, resulting in suboptimal correspondence between human and robot behavior.
Innovation Solution
A method involving generating human and robot models, associating them, acquiring normative motion data, setting evaluation standards, selecting constraint conditions, calculating minimum reproduction errors, and controlling the robot using optimized motion data to enhance accuracy in reproducing human behavior.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If motion indices of a human body model are converted into robot behavior and then corrected to reflect constraint conditions, then the robot can move within its physical limitations, but the correspondence between human behavior and robot behavior is not optimized
Solution Approach 1:
The patent applies preliminary action by incorporating constraint conditions directly into the evaluation function before motion conversion. The evaluation function includes terms for joint range limits, center of gravity constraints, and contact point requirements, allowing the optimization process to simultaneously satisfy constraints while minimizing reproduction error, rather than applying corrections after the fact
Solution Approach 2:
The patent changes parameters by reformulating the motion conversion problem as an optimization task that adjusts motion indices based on constraint satisfaction. The evaluation function weights different constraint parameters (joint limits, center of gravity, contact points) to find optimal motion trajectories that balance fidelity to human motion with adherence to robot physical constraints
2Ease of operation
If motion indices are corrected after conversion to reflect constraint conditions, then the robot behavior becomes physically feasible, but the correspondence between human and robot behavior deteriorates
Solution Approach 1:
The patent merges the motion conversion process with constraint satisfaction by combining the reproduction error minimization objective with constraint conditions into a single unified evaluation function. This allows the optimization to simultaneously achieve accurate behavior correspondence and physically feasible robot motion without separate correction steps
Data Source
AI summary
A robot behavior generation method includes: generating a human body model; generating a robot model; creating associations between the human body model and the robot model; acquiring first motion data indicating movement of a human body model; selectively choosing movement feature indices having first unknown motion data that indicates movement of the human body model, and setting an evaluation standard for a reproduction error with respect to the normative motion data; selecting a constraint necessary for a robot to move, included in second unknown motion data that indicates the movement of the robot model; calculating first unknown motion data and second unknown motion data the reproduction errors of which based on the evaluation standard are minimum under the association and the constraint; and using the second unknown motion data to control the robot.