Vehicle Anomaly Detection Using Cell-Based Map Evaluation Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional anomaly detection devices for vehicle-related anomalies suffer from decreased detection accuracy due to the use of evaluation models created from travel data across wide regions, leading to false negatives in detecting anomalies like cyberattacks.
Innovation Solution
An anomaly detection device that utilizes a grid-imposed map to store evaluation models for each cell, calculating an anomaly level based on vehicle information and positional relationships with neighboring cells to improve detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If evaluation models are created from travel data across wide regions, then the coverage area is increased, but the anomaly detection accuracy deteriorates
Solution Approach 1:
The patent divides the map into multiple cells and creates separate evaluation models for each cell based on local travel data. This segmentation approach allows the system to maintain small, location-specific data regions that improve detection accuracy while collectively covering the entire map area through the ensemble of multiple cell-based models.
Solution Approach 2:
The patent implements local quality by creating evaluation models that are specific to each cell's local environment rather than using a single global model. Each evaluation model is trained on travel data from its specific cell, capturing local driving patterns and anomalies, thereby improving detection accuracy for location-specific cyberattacks while the collection of all cell models provides comprehensive coverage.
2Quantity of substance
If evaluation models are created from travel data across wide regions, then the data quantity is increased, but the anomaly detection accuracy deteriorates
Solution Approach 1:
The patent segments the large quantity of travel data into smaller subsets, each associated with a specific cell on the map. By creating evaluation models based on these segmented data subsets rather than using all data uniformly, the system maintains sufficient data quantity for model training while improving detection accuracy through location-specific patterns.
Solution Approach 2:
The patent introduces a spatial dimension by organizing data according to geographic cells on a map. This transforms the data structure from a flat, undifferentiated collection to a spatially-organized hierarchy where data quantity is distributed across multiple dimensional cells, enabling both sufficient data availability and improved detection precision through spatial context.
Data Source
AI summary
An anomaly detection device includes: an obtainer that obtains vehicle information related to the status of a vehicle and including location data indicating the location of the vehicle; a model storage that stores, for each of a plurality of cells of a grid imposed on a map, an evaluation model for evaluating the vehicle information of the vehicle located at the cell; and a determiner that calculates, based on the vehicle information and evaluation models each being the evaluation model and corresponding to evaluation cells including a first cell including the location of the vehicle indicated in the location data and one or more second cells each having a predetermined positional relationship with the first cell, an anomaly level indicating a degree of anomaly of the vehicle information, determines, based on the anomaly level, whether the vehicle information is anomalous, and outputs a determination result.