Vehicle Sensor Anomaly Detection Using Mahalanobis Distance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting anomalies in multidimensional time-series data from vehicle sensors are computationally inefficient and often lag behind real-time data, making them impractical for use in autonomous or semi-autonomous vehicle systems.
Innovation Solution
A computationally efficient technique using a weighted moving mean and inverse covariance matrix to detect anomalies, with a Kalman filter adjustment that prevents faulty sensor data from affecting the vehicle's operation, allowing for real-time anomaly detection and improved sensor data fusion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing anomaly detection methods are used on vehicle computers, then anomaly detection capability is provided, but computational efficiency is poor and real-time operation is compromised
Solution Approach 1:
The patent segments the anomaly detection process into distinct computational stages: calculating weighted moving means, computing covariance matrices, determining Mahalanobis distances, and comparing against thresholds. This segmentation allows the vehicle computer to process sensor data in manageable computational steps, improving efficiency while maintaining detection capability.
Solution Approach 2:
The patent performs preliminary calculations of weighted moving means and covariance matrices from historical sensor data before evaluating current sensor readings. By pre-computing these statistical parameters, the system reduces the computational burden during real-time anomaly detection, enabling faster processing on vehicle computers.
2Reliability
If existing anomaly detection methods are used, then anomaly detection is performed, but the detection lags behind real-time data
Solution Approach 1:
The patent implements continuous anomaly detection by constantly updating weighted moving means and covariance matrices as new sensor data arrives, rather than processing data in batches. This continuous computation ensures that anomaly detection keeps pace with real-time sensor data streams, minimizing detection lag for autonomous vehicle operations.
3Measurement precision
If complex anomaly detection algorithms are implemented, then detection accuracy improves, but device complexity and implementation difficulty increase
Solution Approach 1:
The patent replaces complex mechanical or computational filtering methods with a statistical approach using weighted moving means, covariance matrices, and Mahalanobis distances. This substitution provides a mathematically elegant solution that achieves high detection accuracy through straightforward calculations, reducing algorithmic complexity while maintaining precision.
Data Source
AI summary
A computer includes a processor and a memory storing instructions executable by the processor to receive a time series of vectors from a sensor, determine a weighted moving mean of the vectors, determine an inverse covariance matrix of the vectors, receive a current vector from the sensor, determine a squared Mahalanobis distance between the current vector and the weighted moving mean, and output an indicator of an anomaly with the sensor in response to the squared Mahalanobis distance exceeding a threshold. The squared Mahalanobis distance is determined by using the inverse covariance matrix.