Driver HMI Cue Optimization Using Simulated Human Behavior
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems lack integrated simulation methods for human behavior and interaction with HMI systems, particularly in assessing driver state and safety criticality, and fail to effectively communicate cues to drivers in automated driving scenarios, leading to issues in personalization and data generation for advanced driving systems.
Innovation Solution
A computer-implemented method simulating human behavior through a model comprising perception, cognition, and physiological layers, generating synthetic training data for HMI systems, and using machine learning to optimize cue generation for effective communication with drivers, incorporating mental workload and state vector analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a machine learning system is used to assess driver state and generate training data, then the personalization and effectiveness of HMI communication is improved, but the complexity of data generation and system integration increases
Solution Approach 1:
The system is divided into distinct functional modules: a driver state detection system that monitors physiological and behavioral parameters, a simulation environment that generates virtual driving scenarios, and a machine learning training system that processes detection data. This segmentation allows each module to be developed and optimized independently while maintaining overall system personalization capabilities.
Solution Approach 2:
The system performs preliminary data generation through simulation environments before actual deployment. Virtual driving scenarios are pre-generated with various driver states and HMI interactions, creating a comprehensive training dataset in advance. This preliminary action reduces the complexity of real-time data collection and system integration during actual operation.
2Manufacturing precision
If simulation methods are integrated for human behavior and HMI interaction, then the quality of training data for advanced driving systems is improved, but the difficulty of detecting and measuring driver states increases
Solution Approach 1:
A simulation environment acts as an intermediary between the driver state detection system and the machine learning training system. The simulation generates virtual driver responses and HMI interactions based on detected states, translating complex physiological measurements into standardized training data formats. This intermediary layer simplifies the detection and measurement processes while maintaining high training data quality.
Solution Approach 2:
The system creates copies of real driving scenarios in a virtual simulation environment. Instead of directly measuring complex driver states in real-time, the system replicates driving situations and driver responses in simulation, generating training data that captures nuanced driver states without the complexity of direct real-world measurement.
3Productivity
If cues are optimized based on driver state and mental workload, then the effectiveness of driver-vehicle communication is improved, but the device complexity for state monitoring and cue generation increases
Solution Approach 1:
The cue generation system is designed to be dynamic and adaptive, automatically adjusting cue characteristics based on real-time driver state detection. The system monitors multiple driver parameters and dynamically modifies HMI cue intensity, timing, and modality to match current driver workload and attention levels, maximizing communication effectiveness without requiring overly complex manual configuration.
Solution Approach 2:
The system implements closed-loop feedback where driver state detection results directly influence cue generation parameters. Detected driver states feed into the machine learning model, which predicts optimal cue strategies, and the effectiveness of these cues is continuously evaluated based on subsequent driver responses. This feedback mechanism optimizes communication effectiveness while keeping the system architecture manageable through iterative learning.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a computer-implemented method for simulating human behavior in the context of interactions between a human and an HMI system (11), a method for training, testing or validating a machine learning system for an HMI system (11) and an HMI system (11), and can be used in the field of automated driving. The method for training a machine learning system to be deployed in an HMI system (11) for optimizing the communication from the HMI system (11) to the human by using training data comprising synthetic training data generated by using the computer-implemented method for simulating human behavior. The method comprises the steps: - Given an output state vector (230) corresponding to a desired physiological response of the human for a given input state vector (203, 231), the optimal cue (104) to be generated by the HMI system (11) is identified during the training. -Therefore, a reward (209) representing the conformity of the human's physiological response with a desired physiological response is calculated. - The desired physiological response is depending on the task the HMI system (11) shall accomplish through generating and outputting the optimal cue (104). Exemplary tasks comprise: • Assisted or Automated Driving: HMI systems for drivers • Aviation: HMI systems for pilots and co-pilots • Rail transport: HMI systems for train drivers • Control of machines: HMI systems for machine users.