Multi-Agent Reinforcement Learning for Inverse Design in Photonic Integrated Circuits
This work addresses the need for more adaptive optimization algorithms in photonic integrated circuits, which are crucial for optical computing, though it appears incremental as it applies existing RL techniques to a specific domain.
The paper tackles the problem of inverse design for photonic integrated circuits, which traditionally suffers from local minima in gradient-based optimization, by introducing multi-agent reinforcement learning algorithms that outperform previous state-of-the-art methods in two- and three-dimensional design tasks with only a few thousand environment samples.
Inverse design of photonic integrated circuits (PICs) has traditionally relied on gradientbased optimization. However, this approach is prone to end up in local minima, which results in suboptimal design functionality. As interest in PICs increases due to their potential for addressing modern hardware demands through optical computing, more adaptive optimization algorithms are needed. We present a reinforcement learning (RL) environment as well as multi-agent RL algorithms for the design of PICs. By discretizing the design space into a grid, we formulate the design task as an optimization problem with thousands of binary variables. We consider multiple two- and three-dimensional design tasks that represent PIC components for an optical computing system. By decomposing the design space into thousands of individual agents, our algorithms are able to optimize designs with only a few thousand environment samples. They outperform previous state-of-the-art gradient-based optimization in both twoand three-dimensional design tasks. Our work may also serve as a benchmark for further exploration of sample-efficient RL for inverse design in photonics.