Steering System Neural Network Emulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current steering systems, especially steer-by-wire systems, face challenges in replicating the familiar steering feel of conventional systems, requiring significant effort and limited success in transferring configurations.
Innovation Solution
A method utilizing a neural network to specify and control steering forces/torques, trained on data from reference vehicles, allowing for the emulation of a reference steering system's behavior in a new system, including variables like steering wheel torque, torsion bar torque, and vehicle conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional methods are used to transfer steering system configurations, then the steering feel can be replicated to some extent, but the process requires considerable effort and achieves only limited success
Solution Approach 1:
The patent creates a virtual copy of the reference steering system's behavior through neural network training. The neural network learns the input-output relationships of the reference system by training with measurement data, effectively copying the steering feel characteristics without physically replicating the mechanical configuration. This allows accurate replication of steering behavior while simplifying the transfer process.
Solution Approach 2:
The patent replaces complex mechanical configuration transfer with a data-driven neural network approach. Instead of manually adjusting mechanical components to match reference systems, the invention uses neural networks to learn and reproduce steering characteristics from measurement data, substituting mechanical tuning with intelligent algorithmic replication.
2Reliability
If a neural network is used to control steering force/torque, then the steering system can reliably replicate reference system behavior, but the system complexity increases
Solution Approach 1:
The neural network performs self-learning by automatically training with measurement data from the reference steering system. The system self-adjusts its internal parameters through the training process, eliminating the need for manual configuration or complex control logic design. This self-service capability reduces the burden on system integrators while maintaining high replication accuracy.
Solution Approach 2:
The patent implements a feedback mechanism where the neural network continuously receives measurement data from the reference steering system during training and adjusts its predictions accordingly. This feedback loop enables the neural network to learn the nuanced relationships between steering inputs and outputs, improving replication accuracy while keeping the control architecture relatively simple.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method for configuring a steering system (10) for a motor vehicle (1). The steering system (10) is designed having a steering wheel actuator (32) for applying a torque and/or a force to a steering mechanism (30) and/or to a steering wheel of the vehicle (1) and having a control unit (50), which is designed to control the operation of the steering wheel actuator (32) and has a steering-force and/or steering-torque specifying unit (37) for this purpose. In order to controllably specify a steering force and/or a steering torque, a a steering-force and/or steering-torque specifying unit (37) having a neural network (40) is used. The neural network (40) is trained and/or conditioned on the basis of base data for a plurality of driving and/or steering maneuvers on at least one reference steering system of at least one reference vehicle as training data. The invention further relates to a method for controlling a steering system (10), and to a steering system (10) and a vehicle (1) as such.