Robot Calibration Error Estimation Using Friction and CoM Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Robots lack the sophistication to accurately duplicate human interactions required for complex tasks due to inadequate calibration, leading to inaccuracies in movement and control.
Innovation Solution
A computing system that divides sensor data into training and test data to determine friction parameters and center of mass estimates, using actuation prediction data to assess the accuracy of robot calibration by comparing residual data against an error parameter threshold.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If robot calibration is performed using traditional methods, then the calibration process can be completed, but the accuracy and precision of robot movement control remain insufficient
Solution Approach 1:
The patent implements feedback by using sensors to collect actual movement data and actuation data, then comparing predicted actuation values with actual measurements to calculate residual values. This feedback loop enables continuous refinement of friction parameters and center of mass estimates, thereby improving robot movement accuracy and calibration reliability simultaneously.
Solution Approach 2:
The patent replaces traditional mechanical calibration methods with a computational approach using machine learning models. Instead of relying solely on physical adjustments and mechanical measurements, the system uses algorithms to predict actuation values, compare them with actual sensor data, and iteratively optimize calibration parameters, achieving higher precision and reliability.
2Adaptability or versatility
If traditional calibration methods are used, then the process is simpler, but the sophistication required to duplicate human interactions is insufficient
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models using training datasets before actual robot operation. The friction parameters and center of mass estimates are determined in advance through iterative training processes, enabling the robot to perform complex human-like interactions with high accuracy without requiring complex real-time calculations during operation.
3Reliability
If more sophisticated calibration is implemented to improve robot sophistication, then the ability to duplicate human interactions improves, but the system complexity increases
Solution Approach 1:
The patent segments the calibration system into distinct functional modules: sensor data collection, training data preparation, model training, prediction generation, and error analysis. Each module performs a specific function, making the overall complex system manageable and maintainable while achieving high calibration quality through coordinated operation of these specialized components.
Data Source
AI summary
A computing system and method are presented. The computing system may store sensor data which includes: (i) a set of movement data, and (ii) a set of actuation data. The computing system may divide the sensor data into training data and test data by: (i) selecting, as the training data, movement training data and corresponding actuation training data, and (ii) selecting, as the test data, movement test data and corresponding actuation test data. The computing system may determine, based on the movement training data and the actuation training data, at least one of: (i) a friction parameter estimate or (ii) a center of mass (CoM) estimate, and may determine actuation prediction data based on the movement test data and based on the at least one of the friction parameter estimate or the CoM estimate. The computing system may further determine residual data, and determine a value for an error parameter.