Reinforcement Learning Applied to the Optimization of Power Delivery Networks with Multiple Voltage Domains

This paper proposes a reinforcement learning (RL)-based method to optimize power plane design with multiple voltage domains. The proposed method enables the simultaneous selection of the plane for expansion and its expansion direction, determining the size and shape of multiple power planes. The results show that the proposed method provides various optimized power plane designs that satisfy the target impedance and ensure the reference plane for the input/output (I/O) interface.