AgentDiag: An agent-assisted diagnostic framework for board-level functional failures

Diagnosing functional failures in complicated electronic boards is a challenging task, wherein debug technicians try to identify defective components by analyzing some syndromes obtained from the application of diagnostic tests. The diagnosis effectiveness and efficiency rely heavily on the quality of the in-house developed diagnostic tests and the debug technicians´ knowledge and experience, which, however, have no guarantees nowadays. To tackle this problem, we propose a novel agent-assisted diagnostic framework for board-level functional failures, namely AgentDiag, which facilitates to evaluate the quality of the diagnostic tests and bridge the knowledge gap between the diagnostic programmers who write diagnostic tests and the debug technicians who conduct in-field diagnosis with a lightweight model of the boards and tests. Experimental results on a real industrial board and an OpenRISC design demonstrate the effectiveness of the proposed solution.