Abstract:In recent years, there has been extensive research on the routing problem of printed circuit boards (PCBs). Due to the increasing number of pins, high pin density, and unique physical constraints, manual PCB routing has become a time-consuming task to achieve design convergence. Previous work decomposed the problem into escape routing and area routing, focusing on these problems separately. However, there was always a gap between these two problems, requiring significant human effort for iterative algorithm adjustments. Furthermore, previous area routing work mainly focused on routing between ball grid array (BGA) packages in escape routing. However, in practice, many components are not in the form of BGA packages, such as passive devices, decoupling capacitors, and through-hole pin arrays. Therefore, it is necessary to study a unified routing approach. The current unified routing approach adopts the A* algorithm, but there is still room for improvement in routing speed. This paper proposes a new algorithm called Unet-Astar, which accelerates the routing efficiency by employing deep learning algorithms in a simulated environment. Additionally, a Deeper Unet is proposed for generating recommended regions for the routing algorithm. The new network structure can provide more contextual information, thereby improving routing efficiency. Experimental results demonstrate the effectiveness and efficiency of the proposed algorithm. Specifically, for all given test cases, our router achieves approximately a 70% improvement in runtime speed compared to the old router. Another major contribution of this work is the development of a routing problem set generator, which can generate parameterized routing problem sets with different sizes and constraints. This enables the evaluation of different routing algorithms and the generation of training datasets for future data-driven routing methods. All the code has been open-sourced and can be found at https://github.com/Firesuiry/Unet-Astar-For-PCB-Routing.

NeuroEscape: Ordered Escape Routing Via Monte-Carlo Tree Search and Neural Network

Circuit Routing Using Monte Carlo Tree Search and Deep Neural Networks

Ordered Escape Routing with Consideration of Differential Pair and Blockage.

GoodFloorplan: Graph Convolutional Network and Reinforcement Learning-Based Floorplanning

Ordered Escape routing for grid pin array based on Min-cost Multi-commodity Flow

Mixed-Crossing-Avoided Escape Routing of Mixed-Pattern Signals on Staggered-Pin-Array PCBs.

A Constraint-Driven Compact Model with Partition Strategy for Ordered Escape Routing

Escape: an optimization method based on crowd evacuation behaviors

MORE-Router+: Multilayer Multi-capacity ORdered Escape Routing Via Bus-oriented Layer Assignment

System partitioning on MCM using a new neural network model

MC-MCF: A Multi-Capacity Model for Ordered Escape Routing

A Fast Escape Router for Large-Scale Staggered-Pin-Array PCBs

CDCtree: Novel Obstacle-Avoiding Routing Tree Construction Based on Current Driven Circuit Model

Unet-Astar: A Deep Learning-Based Fast Routing Algorithm for Unified PCB Routing

Learning to Search Feasible and Infeasible Regions of Routing Problems with Flexible Neural k-Opt

Applying Neural Monte Carlo Tree Search to Unsignalized Multi-intersection Scheduling for Autonomous Vehicles

URBER: Ultrafast Rule-Based Escape Routing Method for Large-Scale Sample Delivery Biochips

Standard Cell Routing with Reinforcement Learning and Genetic Algorithm in Advanced Technology Nodes

MCMCF-Router: Multi-capacity Ordered Escape Routing Algorithms for Grid/Staggered Pin Array

Automatic Routability Predictor Development Using Neural Architecture Search

Routing optimization with Monte Carlo Tree Search-based multi-agent reinforcement learning