Flapping Wing Micro Aerial Vehicles (FWMAVs) have gained significant interest due to their agility, maneuverability and aerodynamic efficiency, making them promising candidates for applications in constrained and dynamic environments. However, the highly nonlinear dynamics and complex aerodynamics of flapping-wing flight pose considerable challenges for flight control design. In this thesis, the digital twin of a physical FWMAV prototype currently under development was constructed, and a reinforcement learning (RL) framework was employed to develop flight control algorithms. The simulation interface integrates the MuJoCo physics engine with a custom-made Gymnasium environment to accurately reproduce the nonlinear flight characteristics of flapping motion. Within this framework, the Proximal Policy Optimization (PPO) algorithm was applied to learn control policies capable of producing smooth and stable flapping motions.