Robot Reachable Manifold Learning for Fast Inverse Kinematics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current robot motion planning systems face challenges in validating the feasibility of planned motions within kinematic and dynamic constraints, often resulting in time-consuming operations that can fall into local minima, especially for sequential motion planning.
Innovation Solution
A system comprising a neural network architecture with forward, inverse, and density network modules is trained to learn the reachable manifold of a robot's end effector, allowing for kinematic feasibility validation and diverse inverse mapping solutions without iterative processes, using babbling data to generate training samples and avoid self-collision and joint limit constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If optimization operations are performed to validate robot motion feasibility, then motion planning accuracy is improved, but computational time increases significantly
Solution Approach 1:
The patent pre-computes and stores the reachable manifold of the end effector in a lookup table during an offline phase. This preliminary action allows the system to avoid time-consuming optimization operations during online motion planning, as feasibility can be directly queried from the pre-computed manifold, thus resolving the contradiction between validation accuracy and computational time.
Solution Approach 2:
The patent replaces the traditional iterative optimization-based feasibility validation with a direct lookup approach using a pre-computed reachable manifold. This substitution eliminates the need for repeated optimization operations during motion planning, significantly reducing computational time while maintaining validation accuracy.
2Reliability
If iterative optimization is used for sequential motion planning, then kinematic feasibility is validated, but the system falls into local minima
Solution Approach 1:
The patent creates a simplified copy of the robot's kinematic constraints in the form of a reachable manifold representation. This copy allows for direct feasibility checking without the complexities of iterative optimization, avoiding local minima while maintaining reliable kinematic validation through the pre-computed manifold structure.
3Speed
If offline manifold modeling is performed, then query speed is improved, but training data requirements increase
Solution Approach 1:
The patent samples only the essential boundary and key configurations of the reachable manifold during the offline phase, rather than exhaustively computing the entire configuration space. This partial action approach achieves sufficient query speed improvement while reducing the amount of training data required compared to complete manifold modeling.
Data Source
AI summary
A system includes: a first module configured to, based on a set of target robot joint angles, generate a first estimated end effector pose and a first estimated latent variable that is a first intermediate variable between the set of target robot joint angles and the first estimated end effector pose; a second module configured to determine a set of estimated robot joint angles based on the first estimated latent variable and a target end effector pose; a third module configured to determine joint probabilities for the robot based on the first estimated latent variable and the target end effector pose; and a fourth module configured to, based on the set of estimated robot joint angles, determine a second estimated end effector pose and a second estimated latent variable that is a second intermediate variable between the set of estimated robot joint angles and the second estimated end effector pose.