Vehicle AI Anomaly Detection Using Accident-Based Diagnostic Patterns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing advanced driver assistance systems (ADAS) and autonomous driving systems face challenges in ensuring the correct and safe operation of artificial intelligence algorithms due to hardware failures or anomalies, which can lead to critical accidents, and current methods are inefficient and costly, often requiring hardware redundancy or extensive computational resources.
Innovation Solution
A vehicle-based AI diagnostic system that uses a computing device to communicate with a network to receive diagnostic patterns, execute AI algorithms, and detect hardware anomalies by matching future driving conditions with past accident scenarios, prioritizing high-risk misclassifications, without relying on additional hardware, thus improving safety and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If hardware redundancy or extensive computational resources are used to ensure AI algorithm safety, then system reliability improves, but device complexity and cost increase
Solution Approach 1:
The AI algorithm performs self-diagnostics by executing diagnostic patterns and comparing outputs against expected results. The system monitors its own hardware health without requiring external monitoring systems or redundant hardware, allowing the AI to detect and report hardware failures autonomously
Solution Approach 2:
Instead of using redundant hardware to verify AI algorithm correctness, the system creates virtual copies of diagnostic scenarios by generating diagnostic patterns from historical accident data. These patterns simulate failure conditions that can be tested against the running AI algorithm to detect hardware anomalies
2Reliability
If traditional hardware redundancy methods are used to detect AI hardware failures, then detection reliability improves, but manufacturing cost increases
Solution Approach 1:
The system creates virtual diagnostic patterns that copy the essential characteristics of historical accident conditions without requiring physical redundant hardware. These synthesized diagnostic patterns are generated from past accident data and used to test the AI algorithm under simulated failure scenarios
Solution Approach 2:
The system changes the operational parameters of the AI algorithm by inputting diagnostic patterns with specific characteristics (e.g., edge cases, boundary conditions) that are likely to reveal hardware failures. By varying input parameters rather than hardware configuration, the system achieves failure detection without additional manufacturing cost
3Measurement precision
If comprehensive AI algorithm verification is performed, then measurement precision of hardware failures improves, but computational resources required increase
Solution Approach 1:
The system extracts only the essential diagnostic patterns from historical accident data that are most likely to reveal hardware failures. By selecting and focusing on critical diagnostic cases rather than verifying all possible scenarios, the system achieves high detection accuracy with reduced computational overhead
Solution Approach 2:
The system performs partial verification by testing the AI algorithm against a curated set of diagnostic patterns rather than exhaustive verification of all possible inputs. This partial action focuses computational resources on the most critical failure modes, achieving sufficient detection accuracy without the full computational cost of complete verification
Data Source
AI summary
In some examples, a vehicle is able to communicate with a computing device over a network. The vehicle includes a processor configured to determine a future driving condition for the vehicle, and further determine that the future driving condition corresponds to a past accident condition. The processor receives, from the computing device, verification data for at least one diagnostic pattern corresponding to the past accident condition. The processor executes an artificial intelligence (AI) algorithm using the verification data corresponding to the at least one diagnostic pattern. Based on determining that an output of the AI algorithm does not correspond to an expected output for the verification data, the processor performs at least one action.