The New York University Tandon School of Engineering is the second oldest private engineering school in the United States. Since its establishment in 1854, Tandon has played a pivotal role in areas such as microwave and radar development, industrial engineering, and the United States space program. In 1989, Professor of Engineering Farshad Khorrami established the Agile Robotics and Perception Lab, launching Tandon’s reputation as a leader in the development of aerial robotics. Today, the lab is run by Professor Giuseppe Loianno, who has been making advancements in aerial robotic technology that will shape the future of the drone industry.

The majority of drones on the market are small (2-10 lbs) to medium (10-20 lbs) sized devices. This includes hobby and racing drones as well as drones like the DJI Matrice series. The popularity of drones in these size categories is due in part to the fact that they are just the right size to support whatever you need. They are compact, light, and agile, making them easy to use, yet large enough to host all the necessary software and hardware to ensure safety and effectiveness. Professor Loianno is interested in scaling down the size of drones to make them more cost-effective while maintaining superior functionality.

To achieve this, Professor Loianno and his students need to develop algorithms that will allow micro drones, which are too small to support the tools typically found on larger drones, such as cameras and advanced navigational systems. ” The main mission of the lab is to create agile autonomous machines that can navigate all by themselves using only on-board sensors in unstructured, and dynamically changing environments and without relying on external infrastructure, such as GPS or motion capture systems,” states the Agile Robotics and Perception Lab’s home page. “The machines need to be active, they should collaborate with humans and between each other and they need to navigate in the unknown environment extracting the best knowledge from it.”

In a paper titled “Learning to Fly in Seconds,” which was first submitted in 2023 and then resubmitted with revisions in 2024, Professor Loianno and his students Jonas Eschmann and Dario Albani reveal how they taught micro drones to navigate without some of the aforementioned tools. The paper explains how Reinforced Learning (RL), akin to a progressive internal lesson plan, has significant potential for the control of autonomous drones. However, they note that one of the drawbacks of RL is that “the simulation-to-reality transfer often brings a hard-to-bridge reality gap. Moreover, RL is commonly plagued by prohibitively long training times.”

Professor Loianno and his team found a way to work around the slow progression of RL while still utilizing its success rate. “In this work, we propose a novel asymmetric actor-critic-based architecture coupled with a highly reliable RL-based training paradigm for end-to-end quadrotor control,” the authors state in the paper. “We show how curriculum learning and a highly optimized simulator enhance sample complexity and lead to fast training times.” Remarkably, Professor Loianno and the team were able to program an inexpensive, off-the-shelf, bare-bones micro drone to fly as well as a drone equipped with a camera and visual avoidance sensor. The drone was able to learn how to operate and navigate within 18 seconds using data uploaded from an off-the-shelf Apple MacBook Pro.

The implications of this research are vast. The system can be applied to larger drones, providing them with further redundancy. This could enable the safe integration of swarms of autonomous drones in shared airspace. For Professor Loianno, he envisions a future where small, inexpensive drones are readily deployable to assist in emergencies, scientific research, mapping, and much more. “Drones will go wherever humans are denied access,” Professor Loianno once said, “and anyone who has studied in the field will be in a very strong position to find employment or entrepreneurial opportunities.” The work Professor Loianno is orchestrating at the Agile Robotics and Perception Lab not only honors Tandon’s history of engineering excellence, but, as Jelena Kovačević, dean of NYU Tandon, said, “Giuseppe’s work to improve the capabilities of autonomous robots is laying the foundation for material advances in safety, security, urban living, and more. In the process, he’s contributing to Tandon’s evolution as a major hub of robotics research.”

Previous Drone News: