The control and stability of drones is a challenging problem. There is need for a more dynamic and robust control that the drone can use to adjust itself to an unknown environment directly. This paper presents a framework for using reinforcement learning to control altitude of a drone. We use PID to stabilize $x$ and $y$ axis of the drone. The drone is trained using Q-learning of Reinforcement Learning in a simulated environment. The trained model is then tested in the real world. Furthermore, a comparative analysis of reinforcement learning and PID algorithm is presented. Finally, an application of way-point navigation from one given point to other in an environment filled with obstacles at different points formulated as a 3-dimensional grid is presented using Q-learning of Reinforcement Learning.