Reinforcement Learning Based PCB Routing Using Dueling Double Deep Q Network

With the development of technology, the circuits on printed circuit boards(PCB) become more and more complex. Traditional PCB routing methods often rely on experiences and heuristic rules, which are inefficient and difficult to guarantee the global optimal solution in complex design. This paper introduces a multi-agent reinforcement learning algorithm based on local observation, which is designed to optimize the auto routing process of PCB by dueling double deep Q network(D3QN). This method combines the advantages of deep reinforcement learning and double Q learning, and can learn and optimize strategies more effectively in high-dimensional state space. The experimental results show that the routing method in this paper shows high adaptability and flexibility in dealing with routing tasks, which not only improves the routing efficiency, but also improves the routing rate.