Test Generation for Hybrid Systems Using Clustering and Learning Techniques

—In this paper, we propose a test generation method that employs clustering and learning techniques to reduce test generation time in hybrid systems. While learning-oriented test generation is a well-studied problem for digital systems, there are limited efforts for utilizing learning during generation of directed tests for hybrid systems. This paper makes two important contributions: i) it develops an efficient technique to cluster a set of functional scenarios that are expected to have similar test generation trajectory, and ii) it employs efficient learning mechanism such that beneficial information is shared during test generation of similar functional scenarios in a cluster. Our experimental results using two popular hybrid systems demonstrate that our approach can significantly (up to 3.8 times, 2.9 times on average) reduce the overall test generation time. I. LEARNING-ORIENTED TEST GENERATION Validation of hybrid systems needs to consider the complexity of both continuous ...

• Type: Journal
• Year: 2016
• Where: VLSID
• Keywords: VLSI, VLSID 2016