Test Point Selection Using Deep Graph Convolutional Networks and Advantage Actor Critic (A2C) Reinforcement Learning
Shaoqi Wei, Kohei Shiotani, Senling Wang, Hiroshi Kai, Yoshinobu Higami, Hiroshi Takahashi, Gang Wang · 2023
Identifying optimal test points to maximize fault coverage is crucial for improving field tests of large-scale integrated circuits (LSIs). In this paper, we introduce Deep-TPs-Explorer, a method that utilizes deep graph-convolutional neural networks (GCNs) to identify a more effective set of test points, thereby enhancing the random testability of logic circuits. For efficient training of the GCN, we employ the Advantage Actor-Critic (A2C) reinforcement learning algorithm. The effectiveness of our proposed method is validated using the ISCAS89 and ITC99 benchmark circuits.