Vehicle Neural Network Error Tainting for Continuous Operation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Vehicle neural network processors face unique and frequent errors due to their application-specific nature, which can lead to safety concerns during autonomous operation, as they may produce corrupted results or hang, causing potential accidents and requiring efficient error handling mechanisms.
Innovation Solution
A system and method for handling errors in neural network processors, including an error detector to identify data errors and a neural network controller that signals pending results as corrupt without terminating the neural network execution, allowing for continuous operation and minimizing disruption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the neural network processor terminates execution upon detecting an error, then safety is improved by preventing corrupted results, but system productivity deteriorates due to interruption of continuous operation
Solution Approach 1:
The patent segments the error handling process by distinguishing between different types of errors (data errors vs. other errors) and applying different responses. Data errors trigger only a taint signal to skip corrupted results, while other errors may terminate execution. This segmentation allows the system to maintain productivity for recoverable errors while ensuring safety for critical errors.
Solution Approach 2:
The patent implements a skipping mechanism where, upon detecting a data error, the system signals that pending results are tainted and skips those specific corrupted results without terminating the entire neural network execution. This allows the processor to rush through or bypass the corrupted data portion while continuing overall operation, thereby maintaining productivity while preserving safety.
2Productivity
If the neural network processor skips tainted results upon detecting a data error, then system productivity is maintained through continuous operation, but reliability deteriorates by potentially propagating corrupted data
Solution Approach 1:
The patent introduces an intermediary mechanism - the taint signal - that acts as a mediator between error detection and result processing. When a data error is detected, the taint signal is generated to mark pending results as corrupted. This intermediary allows the system to identify and handle tainted data appropriately without completely halting execution, thus balancing productivity and reliability.
3Reliability
If comprehensive error detection mechanisms are implemented in the neural network processor, then reliability is improved by detecting more errors, but device complexity increases due to additional error detection and handling components
Solution Approach 1:
The patent extracts the error detection and handling logic as a separate, dedicated component within the neural network processor. By taking out the error detection functionality from the main processing path and implementing it as a distinct error detector module, the system achieves comprehensive error detection while managing complexity through modular design. This separation allows error handling to be addressed independently without complicating the core neural network execution path.
Data Source
AI summary
A system for handling errors in a neural network includes a neural network processor for executing a neural network associated with use of a vehicle. The neural network processor includes an error detector configured to detect a data error associated with execution of the neural network and a neural network controller configured to receive a report of the data error from the error detector. In response to receiving the report, the neural network controller is further configured to signal that a pending result of the neural network is tainted without terminating execution of the neural network.