Introduction

ARORA stands for A Realistic Open environment for Rapid Agent training. ARORA is a geospecific simulator for training agents using Reinforcment Learning ARORA is developed to train agents to complete navigational tasks in a large-scale city environment with the complexities of physics-based movement, vehicle modelling, and a continuous action space.

The goal of ARORA is to provide an open-source platform where scientists can explore AI tasks that simulate agents and the real world with high fidelity. ARORA is built on top of the Unity game engine, allows for headless training (i.e., without any graphical interface), and provides a flexible application programming interface (API) with several sensors.