Vehicle Event Detection for Selective High-Resolution Data Capture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern vehicles face challenges in storing and uploading vast amounts of sensor data due to limited on-board storage space and wireless connection bandwidth, leading to suboptimal data collection for training machine-learning models, especially for anomalous events that require richer data sets.
Innovation Solution
The vehicle system preprocesses data through object identification, compression, and edge computing to selectively store and upload richer data sets for anomalous events while reducing storage and bandwidth demands, using machine-learning models and rule-based algorithms to detect and classify such events in real-time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If all sensor data is stored in on-board storage, then complete data is available for analysis, but storage space is quickly exhausted
Solution Approach 1:
The system extracts only the necessary data elements for training machine-learning models from the complete sensor data set. By identifying and removing redundant information, the system retains only the critical data needed for model training, thereby reducing storage requirements while maintaining data effectiveness.
Solution Approach 2:
The data storage system is segmented into multiple components: on-board storage for critical data, cloud storage for archival data, and edge computing resources for processing. This segmentation allows the system to distribute data across different storage tiers, optimizing the balance between data availability and storage capacity.
2Quantity of substance
If all sensor data is uploaded in real-time, then complete data is available for processing, but bandwidth is quickly exhausted
Solution Approach 1:
Data preprocessing and filtering are performed in advance on the vehicle before upload. By preparing and selecting only the most relevant data for transmission, the system reduces the total data volume that needs to be uploaded, thereby optimizing bandwidth utilization without sacrificing the quality of data available for model training.
Solution Approach 2:
Edge computing devices serve as intermediaries between the vehicle's sensors and the cloud processing system. These intermediaries perform local data processing, filtering, and preliminary analysis, reducing the burden on bandwidth by transmitting only processed and essential data to the cloud.
3Quantity of substance
If data is preprocessed to reduce size, then storage and bandwidth requirements are reduced, but data quality for training may deteriorate
Solution Approach 1:
Different levels of data processing and quality are applied to different types of data based on their specific requirements. Critical data that requires high fidelity for training is retained in full resolution, while less critical data undergoes more aggressive compression. This localized quality approach ensures that data quality is optimized according to the specific needs of each data type.
4Reliability
If more data is collected for anomalous events, then model training robustness is improved, but storage and bandwidth constraints are exceeded
Solution Approach 1:
The system dynamically adjusts data collection and retention strategies based on detected event types. When anomalous events are detected, the system automatically increases data retention and collection for those specific events, while maintaining reduced storage for normal operations. This dynamic adaptation allows the system to prioritize storage resources for the most valuable training data without permanently exceeding storage constraints.
Data Source
AI summary
In one embodiment, a computing system accesses contextual data associated with a vehicle operated by a human driver. The contextual data is captured using one or more sensors associated with the vehicle. The system determines one or more predicted vehicle operations by processing the contextual data based at least on information associated with pre-recorded contextual data associated with a number of vehicles. The system detects one or more vehicle operations made by the human driver. The system determines that an event of interest is associated with the contextual data based on a comparison of the one or more vehicle operations made by the human driver and the one or more predicted vehicle operations. The system causes high-resolution contextual data associated with the event of interest to be stored.