Vehicle Takeover Request System Using User Activity Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for transitioning from autonomous driving to manual driving in vehicles do not effectively account for individual user activities and environmental factors, leading to suboptimal alertness and reaction time assessments, which can result in inadequate takeover requests.
Innovation Solution
A method and system that classify user activities using cameras and sensors, employing convolutional neural networks to analyze human body and hand key points, and a self-learning neural network to determine the most effective takeover request stimulus based on the user's current activity and environment, ensuring timely and appropriate alerts for regaining control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a fixed-time warning system is used to alert users to regain control, then the system operation is simple, but the effectiveness of takeover requests is reduced because it does not account for individual user activities and reaction times
Solution Approach 1:
The system continuously monitors user activity through cameras and sensors, classifying users into activity groups based on real-time behavior. This feedback loop allows the system to adapt warning timing and stimulus types (visual, auditory, haptic) to individual user states, significantly improving takeover effectiveness while managing complexity through automated classification algorithms
Solution Approach 2:
The system performs preliminary classification of user activity groups and pre-determines optimal warning strategies before takeover situations arise. By analyzing user behavior patterns in advance and preparing customized alert sequences, the system ensures immediate effective response when takeover is needed, rather than reacting to generic time-based thresholds
2Measurement precision
If the system monitors detailed user activity and classifies activity groups, then the accuracy of takeover timing is improved, but the device complexity and computational requirements increase
Solution Approach 1:
The system segments user activities into distinct activity groups (e.g., reading, watching video, using phone) with characteristic reaction time ranges. By categorizing continuous user behavior into discrete segments, the system achieves precise reaction time measurement without requiring complex continuous analysis, simplifying the monitoring architecture while maintaining high accuracy
Solution Approach 2:
The system uses camera and sensor data to create simplified digital representations (key points) of user posture and activity state rather than processing full video streams. This copying approach extracts essential features for activity classification, reducing computational complexity while preserving measurement precision for reaction time assessment
3Adaptability or versatility
If the system uses multiple stimulus types (visual, auditory, haptic) for takeover requests, then the adaptability to different user activities is improved, but the device complexity increases
Solution Approach 1:
The system applies different stimulus types (visual, auditory, haptic) tailored to specific activity groups. For example, visual alerts may be optimized for users engaged in auditory activities, while haptic feedback may be more effective for users watching videos. This localized customization of alert qualities improves adaptability without requiring all stimulus types to be equally complex across all scenarios
Data Source
AI summary
Embodiments provide a method and system for monitoring user activity in a vehicle. A first sensor in the vehicle acquiring current activity data regarding the user. A second sensor in the vehicle can acquire environment data regarding an environment surrounding the user. A first set of classifiers can generate key point data regarding the user, which indicates body hand/eye movement points of the user. A second set of classifiers can assign the current activity to the activity group. A storage can store response information regarding a user, which includes mappings between activity groups and corresponding response time duration for the user to regain control of the vehicle. The system can generate a take-over request for the user to regain control of the vehicle based on the mapping and current activity.