Shared Collision Avoidance Control With Nonlinear Risk Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing shared control collision avoidance systems face model mismatch issues and inefficiencies due to the use of linear vehicle dynamics models, leading to inaccurate risk assessments and slow solution efficiency in forward collision avoidance scenarios for autonomous vehicles.
Innovation Solution
An intervention-based shared control method that utilizes a nonlinear vehicle dynamics model, incorporating a virtual collision avoidance control algorithm to generate optimal nominal collision avoidance trajectories and perform comprehensive risk assessments, integrating steering input data and safety constraints to enhance prediction accuracy and solution efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a linear vehicle dynamics model is used in the system, then the optimal solution efficiency of the model predictive control is ensured, but the prediction of the vehicle state may not be accurate enough, resulting in risk assessment distortion
Solution Approach 1:
The patent segments the vehicle dynamics model into linear and nonlinear components. The linear model is used for real-time control optimization to ensure computational efficiency, while the nonlinear model is used for accurate risk assessment and prediction. This segmentation allows each model to be applied where it is most effective, resolving the contradiction between speed and accuracy.
Solution Approach 2:
The patent introduces a risk assessment module as an intermediary between the control system and the vehicle dynamics. This module uses nonlinear models to evaluate risks accurately while the main control system maintains linear model efficiency. The intermediary layer reconciles the conflicting requirements by translating accurate but slow nonlinear predictions into efficient control commands.
2Measurement precision
If a nonlinear vehicle dynamics model is used in the system, then the risk assessment of vehicle state prediction is more accurate, but the efficiency of optimal solution of model predictive control may be very slow and difficult for practice
Solution Approach 1:
The patent divides the computational tasks between linear and nonlinear models based on their respective strengths. The nonlinear model is used specifically for risk assessment where accuracy is critical, while the linear model handles the iterative optimization process where speed is critical. This segmentation enables the system to achieve high risk assessment accuracy without sacrificing real-time control efficiency.
Solution Approach 2:
The patent applies different model qualities to different functional requirements. High-fidelity nonlinear modeling is applied locally to the risk assessment function where precision matters most, while simplified linear modeling is applied to the control optimization function where computational speed matters most. This local differentiation resolves the efficiency-accuracy tradeoff.
3Ease of operation
If the system makes an optimistic prediction on vehicle operation and driver's behavior, then the minimum intervention principle is feasible, but the vehicle will break through the safety constraints preset by the system due to model mismatch
Solution Approach 1:
The patent applies preliminary anti-action by using conservative risk assessment that anticipates potential model mismatches and driver behaviors that could lead to safety violations. Rather than relying on optimistic predictions, the system pre-calculates safety margins and constraint boundaries that account for worst-case scenarios, preventing the vehicle from breaking through safety constraints while still allowing minimum intervention when appropriate.
Solution Approach 2:
The patent implements continuous feedback between the risk assessment module and the control system. The risk assessment continuously monitors predicted vehicle states and driver behaviors, providing feedback that adjusts the intervention level dynamically. When risk is low, minimum intervention is applied; when risk increases or model mismatch is detected, the system increases intervention to maintain safety constraint compliance.
Data Source
AI summary
An intervention-based shared control method and apparatus in forward collision avoidance scenario of autonomous vehicle includes: acquiring vehicle state data of the autonomous vehicle and inputting the vehicle state data into a constructed forward collision avoidance control model to obtain an optimal nominal collision avoidance trajectory of the autonomous vehicle; and acquiring steering input data of a driver in controlling the autonomous vehicle, and obtaining the shared control method in the forward collision avoidance scenario of the autonomous vehicle. The present disclosure proposes a vehicle model decoupling method for control solution and risk prediction in a high-velocity forward collision avoidance scenario.