AIrTonomy: An Experimental Infrastructure for Testing Next-Generation Autonomous Aerial Vehicles

Autonomous aerial vehicles (AAVs) have the potential to transform the future of air mobility. However, a major challenge remains: Despite advances in research on artificial intelligence (AI), data-driven machine learning (ML), and significant investment in AAV research, AAVs are not yet fully autonomous. They cannot engage in behavior that allows them to safely perform complex real-world tasks in interaction with other agents and their physical environments, particularly under challenging conditions. In this paper, we present an innovative nation-wide infrastructure called AIrTonomy that will allow researchers and industry to resolve this failure. AIrTonomy is an open and remotely accessible research experimentation infrastructure that tightly integrates a uniquely instrumented proving ground for physical flights with a cyberinfrastructure for supporting virtually simulated, as well as mixed-reality flight experiments using continuously updated high-fidelity digital twins. AIrTonomy will support scientists and engineers in a new converging field of aerial autonomy research to verify and validate the safety of AI/ML for autonomous aerial systems so that they can perform complex real-world tasks under constraints and close the Sim2Real gap. The physical infrastructure of AIrTonomy, located at Purdue University, integrates four facilities, a unique fleet, and advanced communication and sensing infrastructures into a 15-mile triangular corridor, supporting experiments where AAVs fly in real urban airspace and perform complex tasks jointly with remotely located humans. Once operational, AIrTonomy will support more than 240 physical and 12,000 simulated virtual flights (~30 min duration) monthly, with a projected user base of 1,300+ researchers over its lifetime.