Foggy- and Proximity-Aware Reinforcement Learning for Analog and Mixed-Signal Circuit Placement

As artificial intelligence (AI) continues to transform various fields, leveraging these advancements to overcome the challenges for achieving high performance and fine precision in electronic design automation (EDA) becomes increasingly crucial. Analog circuit layout placement, a complex and time-intensive task, often suffers from inefficiencies in traditional design approaches due to intensive optimization needed. In this paper, we introduce an advanced reinforcement learning (RL) method to automate the analog circuit placement process with improved accuracy and scalability. Our approach leverages the proximal policy optimization (PPO) framework to optimize layout placement objectives such as area, wire length, and layout effects during fabrication, including foggy and proximity effects. By utilizing a topological representation, we further reduce the complexity of the state space, enhancing optimization efficiency. Our experimental results demonstrate that the proposed method reduces foggy and proximity effect variations by up to 98.9% and 95.8% respectively and decreases wire length up to 34.6% compared to other RL-based and popular analytical methods, while maintaining competitive performance in terms of area.