Cobot Model Generation Using Generic Robot Trajectory Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current cobot programming methods, such as programming by demonstration and autoencoder-style networks, are not scalable for diverse environments with multiple cobot models and require complex training processes, precise robot models, and expert supervision, limiting their applicability and accuracy.
Innovation Solution
A neural network-based system that transfers trajectories from a generic robot model to a user-specified cobot model using user references and prior task information, incorporating Conformal Geometric Algebra for inverse kinematics, and iterative refinement to ensure accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If programming by demonstration is used for cobot programming, then the cobot can be programmed without specialized robotics expertise, but the approach is not scalable for multiple robots and requires one robot at a time
Solution Approach 1:
The system creates a universal library of motion skills that can be applied across multiple cobot models and manufacturers. Instead of programming each robot individually, the same motion skills library serves multiple robots with different kinematics, achieving scalability while maintaining ease of use.
Solution Approach 2:
The system captures motion skills from demonstration and creates reusable copies that can be transferred to different cobot models. The motion skills are copied and adapted to target robots with different degrees of freedom through inverse kinematics solving, enabling scalable deployment.
2Measurement precision
If inverse kinematics solvers are used with precise robot models, then accurate motion solutions can be obtained, but expert supervision is required and the process is complex
Solution Approach 1:
The system performs self-calibration by automatically adapting motion skills to target cobot models without requiring expert supervision. The calibration process autonomously handles the complexity of inverse kinematics solving and model adaptation, making the system easier to deploy while maintaining accuracy.
3Productivity
If autoencoder-style networks are used for batch inverse kinematics, then multiple solutions can be found quickly, but complex training from scratch is required for each cobot and inference requires expensive GPUs
Solution Approach 1:
The system performs preliminary work by creating a universal motion skills library during the calibration phase. This pre-computed library contains adapted motion skills that can be directly applied to target cobots without requiring complex training, enabling fast inference on standard hardware.
Solution Approach 2:
The system extracts the essential motion patterns from demonstration and separates them from robot-specific details. The motion skills are extracted in a robot-agnostic form and then adapted to target robots, eliminating the need for retraining on each new cobot model.
4Adaptability or versatility
If a generic robot model is used for inverse kinematics, then a single model can serve various cobots, but the model is limited to revolute joints, does not consider joint limits or self-collisions, and has low position accuracy
Solution Approach 1:
The system applies local quality by adapting the generic motion skills to the specific characteristics of each target cobot. The motion skills are locally adjusted to account for joint limits, collision constraints, and specific kinematics, achieving high position accuracy while maintaining versatility across different cobot models.
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
An apparatus, including: an interface configured to receive a target end-effector pose of a cobot; processing circuitry configured to: generate in a generic robot model a joint trajectory based on the target end-effector pose; employ a trained neural network model to map the joint trajectory generated in the generic robot model into a cobot model; and generate a movement instruction to control a movement of the cobot based on the joint trajectory mapped to the cobot model, wherein the generic robot model has a number of degrees of freedom that is equal to or greater than that of the cobot model.