Driver Interaction Control for Automated Takeover Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current driver assistance systems struggle to reliably distinguish between intentional and unintentional take-over intentions of a driver during automated driving, leading to potential unsafe deactivations of the automated driving system.
Innovation Solution
A method and device that classify driver interactions using a skeleton model and sensor data to generate control signals, allowing for precise differentiation between intentional and unintentional take-over intentions, thereby enhancing road safety by ensuring reliable and safe handovers between automated driving and driver control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If driver assistance systems use simple sensor detection of pedal or steering wheel movements to detect take-over intention, then the system response is fast and simple, but the system cannot reliably distinguish between intentional and unintentional driver actions
Solution Approach 1:
The patent combines multiple data sources including skeleton model data from cameras, sensor data from pedals and steering wheels, and interaction signals to create a comprehensive classification system. This merging of multiple detection methods enables reliable distinction between intentional and unintentional driver actions while maintaining system responsiveness.
Solution Approach 2:
The patent introduces a skeleton model as an intermediary representation that bridges the gap between raw sensor data and driver intention classification. The skeleton model processes camera data to create a standardized representation of driver posture and movement, which then can be combined with other sensor inputs for accurate intention detection.
2Reliability
If the system deactivates automated driving upon detecting any driver interaction, then the system responds quickly to potential driver needs, but unintentional interactions cause unnecessary deactivations
Solution Approach 1:
The patent performs preliminary classification of driver interactions using the skeleton model and predefined movement patterns before final system deactivation decisions. By pre-processing and pre-classifying interactions, the system can quickly determine whether an interaction is intentional or unintentional, reducing the time needed for final decision making while maintaining high accuracy.
3Measurement precision
If the system uses detailed skeleton model and multiple sensor data for classification, then the distinction between intentional and unintentional interactions becomes accurate, but the data processing requirement increases
Solution Approach 1:
The patent segments the driver interaction analysis into distinct components: skeleton model generation from camera data, sensor data processing from pedals and steering wheels, and interaction signal processing. Each segment is processed independently and then integrated for final classification, which reduces the computational burden compared to processing all data simultaneously while maintaining high precision.
Data Source
AI summary
A method for controlling at least one driver interaction system. The method includes reading-in data regarding at least one body part of a driver of the vehicle and an interaction signal which represents a detected interaction of the driver, including at least one operating unit for operating a driving function of the vehicle. The method also includes carrying-out a classification of the interaction of the driver as purposeful or incidental by utilizing the data. Moreover, the method includes generating a control signal for controlling at least one driver interaction system as a function of a result of the classification.