Vehicle Black Box Data Persistence Using AI Collision Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle event data recorders, or black boxes, are limited in their ability to capture relevant data leading up to a collision due to their static configuration and constraints related to space and power, often failing to record significant data outside their recording window.
Innovation Solution
A machine learning-based method is implemented to predict potential collisions, allowing for intelligent data persistence by copying data from a cyclic buffer to long-term storage when a collision is predicted, using a trained ML model that classifies event data and stores it in a persistent black box system with its own controller and long-term storage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If a black box continuously records data in a cyclic buffer, then data is captured within the recording window, but data occurring outside the recording window is overwritten and lost
Solution Approach 1:
The system performs preliminary actions by predicting potential collision events using machine learning models before they occur. When a collision is predicted, the system proactively persists data to long-term storage in advance, ensuring data is preserved before the actual collision happens and before the cyclic buffer would overwrite it.
Solution Approach 2:
The system creates copies of data from the cyclic buffer and stores them in long-term storage. This copying mechanism allows the system to preserve data indefinitely without affecting the original cyclic buffer operation, enabling retrieval of historical data that would otherwise be overwritten.
2Loss of information
If all possible data is recorded while a vehicle is in motion, then complete data coverage is achieved, but significant power and storage space are required which are not feasible
Solution Approach 1:
The system extracts and persists only the data that is relevant to potential collision events, as identified by the machine learning model. This selective extraction approach avoids recording all possible data, thereby reducing storage space and power requirements while still capturing critical information when needed.
Solution Approach 2:
The system dynamically changes the recording parameters based on predicted events. During normal operation, only limited cyclic buffer recording is performed. When a collision is predicted, the system changes parameters to persist data to long-term storage, optimizing the balance between data coverage and resource consumption.
3Adaptability or versatility
If a black box is statically configured to record for a fixed duration, then device complexity is reduced, but the device fails to adapt to varying operational conditions and collision risks
Solution Approach 1:
The system transitions from a static configuration to a dynamic one by incorporating machine learning models that continuously analyze sensor data and predict collision risks. This allows the recording duration and data persistence behavior to adapt dynamically based on real-time operational conditions and predicted events.
Solution Approach 2:
The system implements feedback loops where sensor data is continuously fed into machine learning models, which generate predictions that in turn control the data persistence behavior. This feedback mechanism enables the system to adapt its operation based on real-time conditions while maintaining manageable complexity through structured control flows.
Data Source
AI summary
The disclosed embodiments are directed to improving the persistence of pre-accident data in vehicles. In one embodiment a method is disclosed comprising receiving events broadcast over a vehicle bus; classifying the events using a machine learning model, the classifying comprising indicating that a collision is imminent; and copying data from a cyclic buffer of a black box device into a long-term storage device in response to the classifying.