Steering Wheel Torque Estimation Using Hybrid AI Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current steer-by-wire steering systems lack a realistic mechanical feedback, making it difficult to replicate the familiar steering feel of traditional electric power steering systems, and existing methods for adjusting this feedback are time-consuming and costly.
Innovation Solution
A method and device that estimate steering wheel torque using a parameterizable steering model and artificial intelligence to account for nonlinearities, allowing for a more realistic and adjustable steering feel by combining physical and AI-based estimations, eliminating the need for a lengthy learning phase.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a parameterizable steering model is used to estimate steering wheel torque, then the steering feel can be adjusted by changing parameters, but the model cannot accurately capture nonlinearities of the steering system
Solution Approach 1:
The patent combines a parameterizable steering model with an artificial intelligence model to create a hybrid estimation system. The parameterizable model provides adjustability through parameters, while the AI model captures nonlinearities that the physical model cannot represent, thus merging the advantages of both approaches to simultaneously achieve adaptability and precision.
Solution Approach 2:
The steering torque estimation system uses a composite approach by integrating two different modeling methodologies - the traditional parameterizable physical model and the data-driven artificial intelligence model. This composite structure allows the system to leverage the interpretability and adjustability of physical models while incorporating the pattern recognition capabilities of AI to handle complex nonlinear behaviors.
2Measurement precision
If artificial intelligence methods are used to estimate steering wheel torque, then nonlinearities of the steering system can be captured, but time-consuming and costly learning phases are required
Solution Approach 1:
Instead of using a complete AI model that requires extensive learning phases, the patent applies AI only to estimate the nonlinear components of the steering system. By partially applying AI to handle only the nonlinearities while relying on the parameterizable model for the linear/adjustable aspects, the system achieves high precision without the full time and cost burden of training a complete AI model.
Solution Approach 2:
The steering torque estimation problem is segmented into two parts: the adjustable linear/parameterizable components handled by the parameterizable model, and the nonlinear components handled by the AI model. This segmentation allows each component to be optimized independently, with the AI model focused only on capturing nonlinearities rather than the entire steering system behavior, reducing training requirements.
3Extent of automation
If steer-by-wire steering systems are implemented, then mechanical decoupling between steering wheel and steering system is achieved, but realistic mechanical feedback cannot be provided
Solution Approach 1:
The patent replaces the mechanical coupling traditionally used to provide steering feedback with an electronic control system that uses a hybrid estimation model (parameterizable + AI). Instead of direct mechanical connection, the system uses sensors, processors, and actuators to generate artificial feedback torque based on the combined model output, substituting mechanical interaction with electronic control while maintaining realistic steering feel.
Data Source
AI summary
A method for estimating a steering wheel torque for a mechanical feedback at a steering wheel of a steering system of a motor vehicle is disclosed, comprising: receiving and/or detecting at least one current measurement value of at least one state variable of the motor vehicle by suitable input means; estimating the current steering wheel torque by means of a controller; outputting the estimated steering wheel torque as a steering wheel torque signal, wherein the estimation is carried out by means of a parameterizable steering model, wherein current measurement values of at least a subset of the at least one state variable are fed as input data to the parameterizable steering model for this purpose, and wherein nonlinearities of the steering system are estimated based on the current measurement values of at least the subset of the at least one state variable using a method of artificial intelligence.