Autonomous Driving Behavior Selection With Collision-Risk Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving solutions based on neural network models prioritize human-like driving behaviors over safety, potentially leading to risks in autonomous driving scenarios.
Innovation Solution
A driving behavior determining method that collects and processes data from both vehicles using sensors, deriving safe driving behaviors through a target model and anomaly detection, ensuring no collision risks and selecting the most reliable predictions to determine safe driving actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a neural network model is used to determine driving behavior, then the driving behavior becomes more human-like, but safety is ignored and collision risks increase
Solution Approach 1:
The patent segments the driving behavior determination into two independent modules: a neural network model for generating human-like driving behaviors, and a safety verification module for checking collision risks. This segmentation allows each module to specialize in its function while the system as a whole achieves both human-like behavior and safety.
Solution Approach 2:
The patent introduces a safety verification module as an intermediary between the neural network model and the final driving behavior execution. This intermediary checks the safety of generated behaviors and can request regeneration if collision risks are detected, thus mediating between human-like behavior generation and safety assurance.
2Reliability
If safety verification is added to the driving behavior determination process, then collision risks are reduced, but the complexity of the system increases
Solution Approach 1:
The patent performs safety verification as a preliminary check before executing driving behaviors generated by the neural network model. By conducting this verification in advance, the system prevents unsafe behaviors from being executed, thereby ensuring safety without requiring fundamental redesign of the core driving behavior generation system.
Solution Approach 2:
The patent implements a feedback mechanism where the safety verification module communicates collision risk assessments back to the driving behavior determination system. When unsafe behaviors are detected, the system receives feedback to regenerate behaviors, creating a closed-loop control that ensures safety while maintaining system simplicity through iterative refinement.
3Extent of automation
If the neural network model generates all driving behaviors, then the driving process is fully automated, but unsafe behaviors may be selected
Solution Approach 1:
The patent makes the autonomous driving system dynamic by introducing a conditional execution flow. The system automatically executes neural network-generated behaviors when safety is verified, but dynamically switches to behavior regeneration when safety concerns are detected. This dynamic adjustment maintains full automation while ensuring safety through adaptive response to verification results.
Data Source
AI summary
This application provides example driving behavior determining methods and example related devices thereof. One example method in this application includes obtaining first driving data of a first vehicle and second driving data of a second vehicle. A driving behavior set in which the first vehicle has no risk of collision with the second vehicle is obtained based on the first driving data and the second driving data. The first driving data and the second driving data are processed by using a target model to obtain a predicted driving behavior of the first vehicle. In response to detecting that the driving behavior set includes the predicted driving behavior, the predicted driving behavior is determined as a driving behavior to be performed by the first vehicle.