Drone Company Posts Paper to “ArXiv” For Feedback On their Swarm System of Drones


When physicist, Paul Ginsparg saw that the methods he and his colleagues were using to share information were inadequate, he decided to develop an online archive to store shareable data. In 2001, Ginsarg became a professor at Cornell University where he further developed the repository calling it arXiv (pronounced “archive”). Today arXiv has greatly influenced how the scientific community has embraced an open source approach to scientific research. On arXiv, researchers can share their studies before publication as a way to get feedback from peers. Many scientists believe that through sharing information this way, they are bettering the scientific process by getting input from the global scientific community.

On November 9, 2020, Xin Zhou, Jiangchao, Zhu, Hongyu Zhou, Chao Xu, and Fei Gao used arXiv to release their findings on autonomous drone swarms in a cluttered environment. As the demand for drones increases, enabling them to navigate safely is of great importance. In the near future, it is very likely that swarms of drones will be used in urban environments to make deliveries, monitor traffic, or assist police departments. Military operations have already shown an interest in using drone swarms in a wide range of missions. The problem is to find a way for a swarm of drones to safely interpret an unknown environment without colliding with each other or obstacles.

For a single drone operated within line of sight of a pilot, avoiding a collision is not so much a concern. But for a swarm to work, the drones need to be flown autonomously and in sync. The team from Cornell had to find a way for drones to navigate through a new environment using only onboard sensing. The team was particularly interested in applying sense and avoidance for drones in situations with limited wireless connectivity to prove their model could be used in any situation. This would mean that a swarm of drones could be deployed in rural areas such as a densely packed forest.

The paper is titled EGO-Swarm: A Fully Autonomous and Decentralized Quadrotor Swarm System in Cluttered Environments. EGO stands for ESDF-free Gradient-based lOcal planner, which, as described by the Cornell team, is a means for reducing computation time in drone processing. The paper goes on to state that the EGO-Swarm “lays a solid baseline for onboard local planning of a single quadrotor in unknown environments. The expansion consists of two parts, topological planning, and reciprocal collision avoidance.”

Each drone has its own mapping and collision avoidance sensors. As it flies through an environment, the drone creates a real time map of its surroundings. The map gives the drone the information needed to plot its own navigational path. The map, in combination with avoidance sensors, keeps the drone from crashing with a stationary obstacle or a moving one like another drone. Unlike other proposed drone swarm concepts, these drones do not need to communicate with each other for success.

Because each drone operates independently, they keep themselves on the correct path based on what is in front of them. If a wireless connection is severed, the drones will continue to fly safely because of their onboard sensing capabilities. Other swarm models are based on the idea that the drones work as a comprehensive unit, relaying information to each other to give situational awareness. For example, if a military mission requires the inspection of an unknown region, a swarm of drones could be sent in to gather intel. However, many such situations would be in places with little to no internet connection. A swarm whose drones do not need to communicate together would be needed.

Along with the paper, the team also released a video of the drones in action. Though they did not go into a densely packed forest, the drones smoothly navigated through a clustered copse of trees on campus. In the video, you can see 3 drones simultaneously flying around and through the trees. The drones cross paths but are always able to avoid crashing with each other and the trees. The video also shows a simulation of the same flight illuminating the paths the drones navigate through.

In the video the team released, only 3 drones are used, though they say they have tested the model with upwards of 10 drones at a time. The experiments were run in lab simulations and real world scenarios to validate the findings. The team says they still need to work on the system’s ability to predict and sense unreliable moving objects. This is why posting the research to a source like arXiv is of great importance, it will allow them to get valuable feedback from others in the robotics community. As the authors state in the paper, “we hope to offer the robotics community a complete autonomous navigation solution for quadrotor swarms in complex environments.”


ABOUT US: DroneVideos.com is a Nationwide Media Company specializing in custom Drone Videos for real estate, commercial, farms, construction, golf courses, roof inspections and more. All of our Drone Operators are fully licensed and insured. When you purchase a Drone Video Package from us, you will receive a video professionally edited, color corrected and presented to you on an SEO-Friendly webpage that you can easily share online and on Social Media with a click of a button. Click here to get started.
Start Your Order
We Offer a Variety of Drone Video Packages
to Fit Your Needs and Budget