AI-Based Constrained Random Verification for Design Under Test
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Constrained random verification (CRV) for complex designs, such as memory management units with FIFOs, faces challenges in generating specific stimuli to hit corner cases efficiently, requiring significant human effort and laborious processes to achieve adequate coverage.
Innovation Solution
An AI-based CRV method that employs a two-stage framework, using a transformer-based algorithm to select a limited constraint range and a reinforcement learning-based algorithm to generate stimuli, automating the process without human expert guidance to increase the hit rate of corner cases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If constrained random verification is used for complex designs, then verification coverage can be achieved, but the hit rate of corner cases decreases and human effort increases
Solution Approach 1:
The patent transforms the verification process by changing parameters from completely random stimulus generation to AI-guided constrained random verification. The AI model learns from historical verification data and adjusts constraint parameters dynamically to target corner cases, thereby improving the hit rate while maintaining verification coverage.
Solution Approach 2:
The patent replaces manual expert guidance (mechanical human effort) with an AI-based automated system. The AI model substitutes the mechanical process of human engineers analyzing and generating corner case stimuli, automatically learning patterns from verification data and generating targeted test cases without continuous human intervention.
2Reliability
If constrained random verification is used for complex designs, then verification coverage can be achieved, but the required human effort and labor increase
Solution Approach 1:
The patent implements self-service verification by enabling the AI model to autonomously analyze verification results, identify uncovered corner cases, and generate targeted stimuli without human intervention. The system serves itself by automatically learning from feedback and improving verification coverage independently, reducing the need for human engineers to manually analyze and create test cases.
Solution Approach 2:
The patent replaces manual expert guidance (mechanical human effort) with an AI-based automated system. The AI model substitutes the mechanical process of human engineers analyzing and generating corner case stimuli, automatically learning patterns from verification data and generating targeted test cases without continuous human intervention.
3Reliability
If traditional CRV is used, then verification process can be completed, but time-to-market increases due to laborious processes
Solution Approach 1:
The patent applies preliminary action by having the AI model pre-learn from historical verification data and identify potential corner cases before formal verification begins. The system performs preliminary analysis of design patterns and constraint relationships, preparing targeted test cases in advance, which accelerates the verification process and reduces time-to-market while maintaining completeness.
Data Source
AI summary
An artificial intelligence (AI)-based constrained random verification (CRV) method for a design under test (DUT) includes: receiving a series of constraints; obtaining a limited constraint range according to the series of constraints; generating a series of stimuli according to the limited constraint range; and verifying the DUT by the series of stimuli; wherein at least one of the step of obtaining the limited constraint range according to the series of constraints and the step of generating the series of stimuli according to the limited constraint range employs an AI algorithm.