Abstract
Deep reinforcement learning is considered an effective technology in quantum optimization and can provide strategies for optimal control of complex quantum systems. More precise measurements require simulation control at multiple experimental stages. Based on this, we improved a multi-objective deep reinforcement learning method in mathematical convex optimization theory for multi-process quantum optimal control optimization. By setting the single-process quantum control optimization result as a multi-objective optimization truncation threshold and reward function transfer strategy, we finally gave a global optimal solution that considers multiple influencing factors, rather than a local optimal solution that only targets a certain error. This method achieved excellent computational results on superconducting qubits. Optimum control of multi-process quantum computing can be achieved only by regulating the microwave pulse parameters of superconducting qubits, and such a set of global parameter values and control strategies are given.