Corner Case Verification Sequences for Hardware Design Testing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current hardware design verification methods, particularly those incorporating AI, face challenges in efficiently identifying and addressing corner cases due to the need for extensive human effort and time, despite theoretical advancements in AI methodologies like RL agents and deep learning.

Innovation Solution

Employing Generative AI and RL agents to automate the identification and simulation of corner case scenarios, reducing the reliance on human resources by using computational methods to define and implement test cases.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional verification methods with human effort are used, then verification accuracy can be maintained, but time consumption and labor resources increase significantly

Engineering Contradiction:

Improveverification accuracyVSAvoidtime consumption

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system performs preliminary analysis by generating candidate corner case scenarios using Generative AI before actual verification execution. This pre-generation of test scenarios allows the system to prepare comprehensive test cases in advance, reducing the time needed during actual verification while maintaining thoroughness through AI-generated corner case identification.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces an intermediary AI-based verification system that acts as a bridge between traditional verification methods and automated testing. The Generative AI and RL agents serve as intermediaries that automatically generate and execute test scenarios, reducing human intervention while maintaining verification quality through intelligent algorithm-driven corner case identification.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Difficulty of detecting and measuring

If AI techniques are integrated into verification processes, then corner case identification capability improves, but system complexity and training requirements increase

Engineering Contradiction:

Improvecorner case identification capabilityVSAvoidsystem complexity

Core Design Contradiction:

Difficulty of detecting and measuringVSDevice complexity

Solution Approach 1:

The verification system is segmented into distinct functional modules: Generative AI for corner case generation, RL agents for scenario fulfillment, and traditional verification components. This segmentation allows each module to specialize in specific tasks, improving corner case detection while managing overall system complexity through modular architecture that can be implemented and maintained independently.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The AI-based verification system is designed with multi-functionality, where the Generative AI can generate various types of corner case scenarios across different design domains, and RL agents can adapt to different verification targets. This universal approach reduces the need for domain-specific customizations, managing complexity while maintaining broad applicability and enhanced detection capability.

Inventive Principle:

Principle #6Universality (Multi-functionality)

3Measurement precision

If comprehensive datasets are collected for AI training, then model prediction accuracy improves, but data acquisition time and resources increase

Engineering Contradiction:

Improvemodel prediction accuracyVSAvoiddata acquisition time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary data generation by using Generative AI to create synthetic corner case scenarios and associated datasets before formal verification. This pre-generation of training data through AI synthesis reduces the need for extensive manual data collection from physical measurements and historical sources, achieving comprehensive training datasets more efficiently while maintaining model prediction accuracy.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentEP4703957A1Method for verification of a target design, method for training corner case investigation module, verification device, test device, computer program product and computer-readable storage medium

Publication Date: 2026.03.04 SIEMENS AG

Figure 1~2
Figure 3

AI summary

The invention relates to a method for verification of a target design (12), the method comprising the following steps performed by a verification device (32). Receiving a corner case scenario (28) of the target design (12), the corner case scenario (28) comprising a corner case state of the target design (12) satisfying corner case conditions causing a corner case in the target design (12); Investigating by a sequence investigation module (30) of the verification device (32), a set of sequences (34) comprising one or more sequences of steps (36), for transferring the target design (12) from a start state of the target design (12) to the corner case state of the target design (12); and generating control data (40) for a test device (42) to control the target design (12) to perform the operation steps of at least one sequence of steps (36) of the set of sequences (34).