Quantum Reinforcement Learning for Process Synthesis

A recent study highlights quantum reinforcement learning (RL) as a promising approach for addressing process synthesis challenges. Researchers have established a generalized framework that redefines process synthesis as a Markov decision process, incorporating quantum-enhanced RL algorithms to boost scalability. Previous quantum RL methods faced limitations due to qubit demands that increased significantly with problem complexity. This research addresses that issue by implementing state encoding algorithms, which separate qubit requirements from the size of the problem. A classical RL-based strategy was used as a reference to assess quantum algorithms under the same training conditions. All algorithms were tested on a flowsheet synthesis problem with varying unit counts to evaluate their performance and scalability, with results indicating that each method successfully identified the optimal flowsheet configuration.