AirSim is an open source tool developed by Microsoft that can be used to simulate the flights of drones. Built on Unreal Engine, the simulator is a powerful tool to understand the working of autonomous vehicles and is the right platform for experimentation with Artificial Intelligence, Deep Learning and Reinforcement Learning algorithms.
AirSim has been developed as an Unreal plugin that enables the flight simulations to take place in any of the Unreal environments provided by the tool. Some of the pre-built environments to work on are: Neighborhood, Landscape Mountains, Coastline, etc.
In this project, we have implemented our Reinforcement Learning Algorithm in the Neighborhood environment that consists of a varied set of obstacles that the drone may encounter such as trees, houses, electric poles and so on. AirSim proves to be extremely useful when attempting to simulate flights and the performances of autonomous vehicles - it provides a framework of learning that is as realistic as possible. This helps in avoiding the task of learning by these vehicles in the real world - it is very difficult to gather large amounts of data without leading to these systems crashing, which ultimately deems expensive.
With the help of the number of APIs present, AirSim eases the process of gathering and retrieving of data that is crucial for the purpose of training and learning of drones and other vehicles.
For further information, refer to the github repository [3].