Autonomous Driving Handover Control Using Driver Gaze Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current level 3 autonomous driving systems lack the ability to accurately determine whether a human driver is ready to take control during a handover, leading to potential safety risks as the system may deactivate autonomous driving when the human driver is not ready to take control when the system fails to recognize the driver's readiness, resulting in unsafe transitions.
Innovation Solution
An autonomous driving control method and device that determines human driver gaze validity and intervention validity to safely perform control-right handovers by using a system of sensors and cameras to assess the driver's attention and steering wheel manipulation, switching to manual mode when necessary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If the autonomous driving system deactivates autonomous driving mode when the human driver is not ready to take control, then the system responds quickly to handover requests, but safety is compromised because the driver cannot immediately assume control
Solution Approach 1:
The system performs preliminary assessment of driver readiness before deactivating autonomous driving mode. It evaluates multiple parameters including gaze direction, steering wheel manipulation, and driver state in advance to determine whether the driver is prepared to take control, thereby preventing unsafe handovers while maintaining responsive operation
Solution Approach 2:
The system continuously monitors driver behavior and system state, providing feedback loops that assess whether handover conditions are met before executing mode transitions. This feedback mechanism ensures that autonomous driving is only deactivated when the driver is genuinely ready, balancing quick response with safety
2Measurement precision
If the system requires the human driver to grip the steering wheel during autonomous driving, then driver readiness can be detected, but the autonomous driving function becomes less convenient and more intrusive
Solution Approach 1:
The system segments the driver readiness assessment into multiple independent detection channels: gaze direction monitoring, steering wheel manipulation detection, and driver state assessment. This segmentation allows comprehensive evaluation without requiring continuous steering wheel contact, maintaining both accuracy and convenience
Solution Approach 2:
The system uses multi-functional sensors that serve multiple purposes: cameras monitor both driving environment and driver gaze, steering wheel sensors detect both control inputs and driver presence. This multi-functionality enables accurate readiness detection without adding separate intrusive requirements
3Reliability
If the system uses multiple parameters to assess driver readiness (gaze, steering wheel manipulation, driver state), then handover safety is improved, but system complexity increases
Solution Approach 1:
The system merges multiple assessment functions into an integrated driver monitoring system. The camera system, steering wheel sensors, and driver state detection are combined into a unified evaluation framework that assesses gaze, manipulation, and driver state simultaneously, reducing overall system complexity while maintaining comprehensive safety assessment
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method for controlling autonomous driving in an autonomous vehicle includes detecting a situation in which autonomous driving is impossible while the vehicle operates in an autonomous driving mode, outputting a control-right handover request warning alarm and then activating a minimal risk maneuver driving mode, determining a human driver gaze validity based on the detected situation, determining a human driver intervention validity upon determination that the human driver gaze is valid, and determining control-right handover of the autonomous vehicle based on the human driver intervention validity. Thus, the control-right may be reliably transferred from a system to a human driver.