Astralis: A High-Fidelity Simulator for Heterogeneous Robot and Human-Robot Teaming

Abstract

We introduce Astralis simulator, a high-fidelity robot simulation platform for the development of multi-robot and human-robot coordination algorithms which can be seamlessly translated to real-world environments. The simulator provides novel features of dynamically initializing the virtual environment with real-world 3D point cloud data and a uniformly random arrangement of static and dynamic obstacles in the environment. This allows the user to generate several variants of a base scenario for strategic planning and algorithm validation. The simulator can receive high-level command inputs to control a team of Unmanned Aerial Vehicles (UAVs), Unmanned Ground Vehicles (UGVs), and human avatars. The simulated robot models are built with high fidelity control and navigation capabilities which can be readily deployed on real robot platforms. We use Astralis simulator to analyze human-robot coordination algorithms for tracking, following and leading targets in a search and rescue mission. The algorithm is validated using a UAV and a UGV in simulation and on physical platforms. Our simulator provides comparable results to the real-world experiments in terms of the executed trajectories by the robots.