This work presents a comprehensive investigation of AI-powered robotic manipulation and grasping, emphasizing reinforcement learning techniques integrated with virtual simulation. The da Vinci Research Kit (dVRK) platform is employed in a simulated environment to develop and evaluate advanced grasping policies. This paper discusses recent developments in robotic manipulation and describes our approach to constructing the simulation, focusing on how the Unified Robot Description Format (URDF) and PyBullet can be leveraged for precise modeling and training. Experimental results illustrate the promise of combining reinforcement learning with virtual simulations for surgical robotics, highlighting robust and adaptive grasping strategies.