Artificial Intelligence in Electronic Design Automation: A Survey on Logic Synthesis and Netlist Optimization

Abstract

Modern digital integrated circuits contain very large numbers of logic gates, and the conventional heuristics used in logic synthesis and netlist optimization struggle to explore the resulting design space efficiently. Artificial Intelligence (AI) and Machine Learning (ML) have therefore been investigated as complementary tools for these tasks. This survey examines the role of AI in logic synthesis and netlist optimization, covering technology-independent restructuring of And-Inverter Graphs (AIGs), selection of synthesis flows, technology mapping, early prediction of power, performance, and area (PPA), generative hardware design, and verification. Related studies on reinforcement learning, graph neural networks, and generative models show how each technique supports a different stage of the synthesis flow. The survey also discusses the benefits of AI, including faster design exploration, improved multi-objective optimization, and early quality-of-results (QoR) estimation, together with remaining challenges involving training data, computational cost, tool integration, and formal verification. Overall, AI is best viewed as a complement to established synthesis tools rather than a replacement for them.