Robot Control Model Calibration for Friction and Torque Nonlinearities
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for operating machines, particularly robots, face challenges in accurately learning and calibrating nonlinearities such as friction and drive torque fluctuation, which can result in combined learning errors that complicate model-based control and monitoring.
Innovation Solution
A method involving the use of two filters to isolate and reduce the effects of nonlinearities, allowing for improved machine learning and calibration of models, where one model addresses aperiodic nonlinearities like friction and the other periodic nonlinearities like drive torque fluctuation, using adaptive filters and function approximators like radial basis function networks and harmonic activated neural networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If learning errors are determined based on model values from multiple nonlinear models, then measurement precision improves, but device complexity increases due to multiple filters and calibration processes
Solution Approach 1:
The learning error determination process is segmented into multiple independent filter channels, each handling specific types of nonlinearities (friction, drive torque fluctuation). By dividing the complex calibration problem into separate manageable components, the system achieves higher precision without overwhelming complexity management.
Solution Approach 2:
Filters serve as intermediary components between the raw learning errors and the model calibration process. These intermediaries selectively process different types of errors, enabling precise calibration while maintaining systematic organization that manages complexity.
2Productivity
If multiple nonlinearities are learned simultaneously, then productivity improves through comprehensive model calibration, but manufacturing precision deteriorates due to combined learning errors
Solution Approach 1:
Different nonlinearities (friction, drive torque fluctuation) are separated into distinct filter-processing channels. This segmentation allows each nonlinearity to be learned independently with dedicated calibration processes, preventing error contamination between different nonlinearity types while maintaining overall calibration efficiency.
Solution Approach 2:
Each filter is specifically designed with local quality characteristics tailored to handle particular types of nonlinearities. The first filter optimizes for friction-related errors while the second filter optimizes for drive torque fluctuation errors, ensuring high precision for each specific nonlinearity type.
3Measurement precision
If filters are used to isolate nonlinearity effects, then measurement precision improves for individual nonlinearities, but device complexity increases due to additional filtering components
Solution Approach 1:
Filters act as intermediary components that mediate between raw sensor data and model calibration processes. Each filter is positioned strategically in the signal processing chain to isolate specific nonlinearity effects, providing precise measurement capability while maintaining a structured and manageable system architecture.
Solution Approach 2:
The filter system is designed with universal characteristics where each filter can process multiple types of input data (machine state values, model values, reference values) through a unified filtering mechanism. This multi-functionality reduces overall system complexity compared to having separate specialized processors for each data type.
Data Source
AI summary
A method for operating, in particular controlling and/or monitoring, a machine, in particular a robot includes: a) determining learning error values on the basis of model values, which are determined by a first model and a second model on the basis of machine state values, and on the basis of reference values of the machine; b) filtering the determined learning error values with a first filter and calibrating the first model on the basis of the filtered learning error values; c) filtering the determined learning error values with a second filter and calibrating the second model on the basis of the learning error values filtered by the second filter; and d) operating, in particular controlling and/or monitoring, the machine on the basis of model values determined by the calibrated first model and the calibrated second model on the basis of machine state values.