Autonomous Vehicle Steering Control With Hybrid Dynamics Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous vehicle steering systems face challenges in accurately modeling the complex interplay of power steering and tire forces, leading to inaccuracies in steering control due to unaccounted residual dynamics and proprietary system components.
Innovation Solution
A predictive model combining a physics function and a machine learning model is used to improve the accuracy of steering control, where the physics function models linear kinematics and the machine learning model captures nonlinear residual dynamics, such as lateral dynamics and road-tire interactions, within a model-predictive-control framework.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a first-principles model derived from physics laws is used to approximate steering system behavior, then the model can be developed without proprietary specifications, but the accuracy is insufficient to capture the complex interplay of power steering and tire forces
Solution Approach 1:
The patent combines a physics-based model with a data-driven neural network model to create a hybrid predictive model. The physics model captures fundamental steering dynamics while the neural network learns residual nonlinearities from operational data, achieving both development ease and high accuracy without requiring proprietary specifications.
Solution Approach 2:
The predictive model uses a composite approach by integrating two different modeling paradigms (physics-based and data-driven) into a unified framework. This composite model leverages the strengths of both approaches to accurately represent the complex steering system behavior.
2Device complexity
If a simple physics-based model is used for steering control, then the system complexity remains low, but the model cannot accurately capture residual dynamics and proprietary system behaviors
Solution Approach 1:
The patent segments the steering system model into distinct components: a physics-based model for fundamental dynamics and a neural network model for residual dynamics. This segmentation allows each component to be developed and validated independently while maintaining overall system reliability.
Solution Approach 2:
The neural network acts as an intermediary that bridges the gap between the simplified physics model and the actual complex steering system behavior. It learns and compensates for residual dynamics that the physics model cannot capture, improving reliability without requiring full system complexity.
3Measurement precision
If proprietary power steering control module specifications are obtained for accurate modeling, then steering control accuracy improves, but the system requires access to proprietary information that is generally unavailable
Solution Approach 1:
The system performs self-identification of steering dynamics by using operational data from the vehicle itself to train the neural network model. This self-service approach eliminates the need for proprietary specifications while achieving accurate modeling specific to each vehicle's actual behavior.
Solution Approach 2:
The patent changes the modeling approach from requiring fixed proprietary parameters to learning dynamic parameters from operational data. This allows the model to adapt to specific vehicle configurations without needing manufacturer specifications, improving both accuracy and applicability.
Data Source
AI summary
Model-based control of dynamical systems typically requires accurate domain-specific knowledge and specifications system components. Generally, steering actuator dynamics can be difficult to model due to, for example, an integrated power steering control module, proprietary black box controls, etc. Further, it is difficult to capture the complex interplay of non-linear interactions, such as power steering, tire forces, etc. with sufficient accuracy. To overcome this limitation, a recurring neural network can be employed to model the steering dynamics of an autonomous vehicle. The resulting model can be used to generate feedforward steering commands for embedded control. Such a neural network model can be automatically generated with less domain-specific knowledge, can predict steering dynamics more accurately, and perform comparably to a high-fidelity first principle model when used for controlling the steering system of a self-driving vehicle.