FanoutNet: A Neuralized PCB Fanout Automation Method Using Deep Reinforcement Learning

In modern electronic manufacturing processes, multi-layer Printed Circuit Board (PCB) routing requires connecting more than hundreds of nets with perplexing topology under complex constraints and highly limited resources, so that takes intense effort time human engineers. PCB fanout as a pre-design has been proved to be an ideal technique reduce the complexity by pre-allocating resources pre-routing. However, current design heavily relies on experience engineers, there is no existing solution for automation in industry, which limits quality automation. To address problem, we propose neuralized method deep reinforcement learning. best our knowledge, are first literature fanout. We combine Convolution Neural Network (CNN) attention-based network train policy model value model. The models learn representations layout netlist make decisions evaluations place employ Proximal Policy Optimization (PPO) update parameters models. addition, apply router improve routing. Extensive experimental results real-world industrial benchmarks demonstrate approach achieves 100% routability all cases improves wire length average 6.8%, makes significant improvement compared state-of-the-art methods.