Vehicle Event Detection Using Dynamic Sensor Resolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for identifying vehicle and driver events face challenges with complex sensor data, requiring impractical computational power on-vehicle or inefficient data transmission to backend servers, especially when event classification is uncertain.
Innovation Solution
An event detection and reporting system that subsamples full resolution sensor data to produce reduced resolution data for on-vehicle classification, transmitting only reduced resolution data with high confidence classifications and enhanced resolution data when confidence is low, thereby minimizing computational demands and data transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full resolution sensor data is transmitted to backend server for event classification, then classification accuracy is improved, but data transmission overhead and computational costs increase
Solution Approach 1:
The system changes the resolution parameter of sensor data based on classification confidence. When confidence is high, reduced resolution data is transmitted; when confidence is low, full resolution data is transmitted. This dynamic parameter adjustment optimizes the balance between transmission efficiency and classification accuracy.
Solution Approach 2:
The system applies partial action by transmitting only the necessary amount of data resolution. Instead of always transmitting full resolution data, it transmits reduced resolution data when sufficient for classification, and only transmits full resolution data when necessary for uncertain classifications.
2Measurement precision
If on-vehicle processing uses high computational power to classify complex sensor data, then event classification accuracy is improved, but device complexity and cost increase
Solution Approach 1:
The system introduces an intermediary confidence assessment mechanism between sensor data capture and final classification. The processor first assesses classification confidence using reduced resolution data, then decides whether full resolution processing is needed, avoiding the need for always-high computational power on-vehicle.
Solution Approach 2:
The classification process is segmented into two stages: initial classification using reduced resolution data for high-confidence events, and secondary classification using full resolution data for low-confidence events. This segmentation allows simpler on-vehicle processing with fallback to more complex processing only when needed.
3Productivity
If reduced resolution sensor data is used for event classification, then data transmission is minimized, but classification confidence may be insufficient
Solution Approach 1:
The system implements feedback by transmitting both the classification result and the associated confidence level to the backend server. This allows the server to identify low-confidence classifications and request full resolution data for re-evaluation, creating a feedback loop that maintains reliability while optimizing transmission efficiency.
Data Source
AI summary
A method for transmitting event data from an event detection and reporting system in a vehicle to a backend server includes storing full resolution sensor data in a vehicle buffer memory and subsampling the full resolution sensor data to produce reduced resolution of the sensor data. A vehicle event is detected based on an analysis of the reduced resolution sensor data, and the vehicle event is classified based on the reduced resolution sensor data. A determination is made as to whether the vehicle event is classified as a particular type of event with a minimum degree of confidence. The method then includes transmitting an indication that the vehicle event has not been classified with the minimum degree of confidence when the vehicle event has not been classified with the minimum degree of confidence, and transmitting reduced resolution sensor data for the detected vehicle event when the vehicle event is classified with at least the minimum degree of confidence.