In-Vehicle ML Anomaly Detection for Constrained Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle health diagnostics systems face challenges with high computing resource requirements and costly wireless data transmission for advanced analytics, necessitating a cost-effective and resource-efficient anomaly detection method.
Innovation Solution
Implementing an in-vehicle machine learning-based anomaly detection system that identifies potential abnormalities and triggers data transfer to a cloud infrastructure only when necessary, using a counter-based approach to manage data transmission and leveraging cloud diagnostics for detailed analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If continuous data transfer to cloud infrastructure is implemented for vehicle diagnostics, then diagnostic accuracy is improved, but wireless data transmission costs and computing resource requirements increase
Solution Approach 1:
The system performs preliminary anomaly detection using machine learning models directly in the vehicle before triggering data transfer to the cloud. This preliminary filtering action identifies only abnormal conditions that require cloud-based analysis, preventing unnecessary continuous data transmission while maintaining diagnostic accuracy for actual anomalies
Solution Approach 2:
The system implements local anomaly detection capabilities using onboard machine learning models that analyze sensor data in real-time. This local processing quality allows the vehicle to autonomously determine when cloud communication is necessary, reducing overall data transmission requirements while maintaining high diagnostic precision for abnormal conditions
2Speed
If advanced analytics are performed onboard for real-time diagnostics, then response time is improved, but computing resource allocation increases
Solution Approach 1:
The diagnostic system is segmented into two parts: lightweight machine learning anomaly detection models that run onboard for real-time monitoring, and more computationally intensive cloud-based analytics for detailed analysis. This segmentation allows rapid local response to anomalies while avoiding continuous heavy computing resource allocation
Solution Approach 2:
The system applies partial action by implementing only the essential anomaly detection functionality onboard using simplified machine learning models, rather than implementing complete advanced analytics locally. This partial onboard capability provides real-time anomaly identification with minimal computing resources, while full analytics are performed selectively in the cloud
Data Source
AI summary
A method for signal anomaly detection includes receiving at least one signal and determining, using at least one machine learning anomaly detection model, whether a value associated with the at least one signal is within a range of an expected value. The method also includes, in response to a determination that the value associated with the at least one signal is not within the range of the expected value, incrementing a counter, and, in response to a determination that a value of the counter is greater than or equal to a threshold value, identifying, based on the at least one signal, signal anomaly information. The method also includes communicating the signal anomaly information to a remote computing device, receiving, from the remote computing device, diagnostics information responsive to the signal anomaly information, and, in response to receiving the diagnostics information, initiating at least one corrective action procedure.