Post-Event Vehicle Occupant Injury Prediction from Simulation Metadata
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing statistical and machine learning models struggle to accurately predict vehicle occupant injuries due to limited availability of actual event data, particularly for scenarios involving second and third row occupants, out-of-position occupants, and new vehicle structures like electric and autonomous vehicles, which hampers timely and effective response by first responders and hospital personnel.
Innovation Solution
A simulation-based vehicle event metadata system that includes a vehicle event database, processor, and memory, which generates and utilizes simulation-based data to predict occupant injuries by applying vehicle event pulses to a vehicle simulation model, collecting injury data, and adjusting predictions based on occupant frailty parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If actual vehicle event data is used for injury prediction, then model accuracy improves for common scenarios, but prediction reliability deteriorates for data-limited scenarios (second and third row occupants, out of position occupants, new vehicle structures)
Solution Approach 1:
The system performs preliminary actions by conducting virtual simulations across diverse vehicle architectures and occupant scenarios before actual events occur. The simulation framework pre-generates injury prediction models for various vehicle types (including new electric and autonomous vehicles), occupant positions (second row, third row, out of position), and event conditions, making these predictions available immediately when needed without relying on limited historical data for each specific scenario.
Solution Approach 2:
The system creates virtual copies of actual vehicle events through high-fidelity simulations. Instead of relying solely on limited real-world event data, the system generates synthetic event data by copying and replicating various vehicle architectures, occupant configurations, and crash scenarios in a virtual environment. These simulated copies provide sufficient training data to improve prediction reliability for scenarios where actual event data is scarce.
2Reliability
If simulation-based metadata is used to supplement data-limited scenarios, then prediction reliability improves, but system complexity increases
Solution Approach 1:
The system introduces simulation-based metadata as an intermediary layer between limited actual event data and injury prediction models. This intermediary simulated data bridges the gap by providing additional training examples for scenarios with scarce real-world data, while the hybrid approach (combining actual and simulated metadata) maintains model interpretability and reduces the complexity burden compared to pure simulation approaches.
Solution Approach 2:
The simulation framework is designed to be universal across multiple vehicle architectures, occupant types, and event scenarios simultaneously. By creating a multi-functional simulation system that can handle various vehicle types (traditional, electric, autonomous), occupant positions (seated, out of position, second row, third row), and crash conditions, the system achieves broad reliability improvement without proportionally increasing complexity for each specific scenario.
3Adaptability or versatility
If comprehensive vehicle event simulations are performed across multiple conditions, then injury prediction coverage improves, but computational resources required increase
Solution Approach 1:
The system performs comprehensive simulations in advance to pre-compute injury predictions for various vehicle architectures, occupant positions, and event conditions. By conducting these computationally intensive simulations beforehand and storing the results as pre-computed metadata, the system achieves broad prediction coverage while reducing real-time computational resource requirements when actual events occur.
Solution Approach 2:
The simulation system focuses computational resources on generating high-quality, detailed simulation data for specific local scenarios where actual event data is most scarce (second row occupants, third row occupants, out of position occupants, and new vehicle structures). Rather than uniformly distributing computational effort across all scenarios, the system concentrates resources on these critical data-limited areas, achieving improved coverage where it is most needed while managing overall computational load.
Data Source
AI summary
A vehicle event database includes simulation-based vehicle event metadata associated with a vehicle architecture. The metadata is generated by determining vehicle event pulses, applying the vehicle event pulses to a vehicle simulation model, collecting injury data associated with body regions of a simulated occupant within the vehicle simulation model, and predicting a probability of injury to the body regions of an actual occupant of a vehicle having the vehicle architecture based on the injury data. A vehicle event notification is received from a vehicle. Vehicle data and occupant data are received. The simulation-based vehicle event metadata associated with the vehicle architecture of the vehicle is retrieved from the vehicle event database. An injury to a first body region of the body regions of the actual occupant is predicted based on the vehicle data, the occupant data, and the simulation-based vehicle event metadata using an injury prediction model.