About

AutoCkt is an advanced automation framework designed for analog circuit optimization, leveraging deep reinforcement learning (DRL) techniques. The primary goal of this project is to streamline the traditionally manual and iterative process of circuit design by using AI to optimize critical design parameters (such as transistor sizes, capacitors, and power consumption) for a given set of performance specifications.

The framework integrates seamlessly with popular simulation tools like NGSPICE and Spectre, enabling real-time parameter adjustments and performance evaluation. Through the use of reinforcement learning algorithms like Proximal Policy Optimization (PPO), AutoCkt can autonomously explore design spaces, significantly reducing optimization times and improving overall circuit performance. This is particularly useful in high-complexity designs, where traditional methods might be inefficient or too slow.

The system’s flexibility allows it to adapt to various circuit topologies, including operational amplifiers and other analog circuits, while also supporting advanced features like post-layout parasitic extraction and transfer learning.

Paper based: 📄AutoCkt: Deep Reinforcement Learning of Analog Circuit Designs