Bio-inspired artificial intelligence-reinforcement learning, artificial neural network and bio-inspired methods- is used in this paper in order to design an algorithm capable of learning to flight autonomously.In this paper, we introduced and explained about our algorithm called FE-TRPO (Flight Enhanced TRPO) and tested it in Gazebo simulator by an AscTec Pelican simulated model where the multi-rotor shows the capability of learning to fly by learning from its own actions and the rewards receiving from the environment in an autonomous way, where the only controller is the reinforcement learning based algorithm and no conventional controllers such as PID, MPC, etc is used. Finally, our algorithm learns to fly using only the data generated from IMU. As a result, our algorithm can be used by a multi-rotor at any unknown environment where no guiding system such as Vicon exist. A video clip of our work can be viewed in: https://www.youtube.com/watch?v=Jj42A5q-YA.