Automated Datasheet Extraction and DRC Suite Adapted for Reverse Engineering

Today, Air Force engineers manually review schematics in order to reverse engineer Circuit Card Assemblies (CCAs) when understanding and maintaining legacy avionic systems. To improve the accuracy of these designs and accelerate design review, Mobilio Co., leveraging subcontracts with Diode Inc. and JITX Inc., proposes to use artificial intelligence techniques to develop an enhanced Design Rule Checker (DRC) that verifies the functionality of a reverse-engineered schematic of a CCA. Diode Inc. will assist Mobilio in constructing a large, validated, labeled dataset of 100,000 example schematics for training. Mobilio Co. will use its existing Kipo AI system to ingest datasheets for electronic components and add detailed machine-readable data on each component to the dataset. Mobilio will build on JITX’s advanced DRC framework, training on the schematic dataset to arrive at an optimal set of design rules. Mobilio will improve the resulting AI-enabled DRC with feedback from Air Force users, and will integrate it into a software package that securely integrates with user-specified schematic generation tools and produces a report of design violations from an input schematic. The resulting AI-enabled DRC tool will prevent design errors and improve the accuracy of reverse-engineered designs of avionic systems.This proposal leverages cutting-edge capabilities from multiple industry leaders. Mobilio Co. has previously developed software, Kipo AI, that extracts technical data from datasheets for electronic components that are part of the CCA. Kipo AI surpasses the capabilities of similar tools: it can extract detailed data like electrical specifications, reference designs, required circuitry, terminal details, and more, and works on even the most complex component types, like CPUs and processors. The depth of this data gives Mobilio’s approach unique capabilities that other tools lack: for instance, checking for required circuitry like a pull-down resistor specified in the datasheet. Second, Diode Inc. has special strengths in curating, synthesizing, and augmenting electronic design data for the purpose of developing DRCs. Diode Inc has internally developed one of the largest high-quality datasets of schematics, and will help construct a training dataset that is large, detailed, diverse, and high-quality, essential for effective AI model training. Finally, JITX has the industry’s most flexible and capable DRC framework, built on a custom programming language used to define circuits. Mobilio will build upon this framework, leveraging its flexibility to iteratively derive novel design rules from the training data. Other DRC frameworks lack this flexibility, limiting the scope of potential design rules the AI training method can yield. JITX is used by industry leaders such as Northrop Grumman and Lockhed Martin for avionic system design. The resulting AI-enabled DRC would catch a broader range of design rule violations and reduce design errors.