When Technology Mimics Life
Have you noticed that over the years, firework shows have become more and more elaborate? During Fourth of July, concerts and sporting events, it's becoming more common to see extravagant artwork of the American flag or team logos light up the sky. These brilliant light shows at night are not a result of the traditional fireworks we are used to, but hundreds of autonomous drones programmed with intricate choreography. The concept of autonomous drones operating as one group to complete a task is also known as drone swarms.
The Vijay Kumar Lab at the University of Pennsylvania is exploring swarms as a new use case for autonomous drones based on biomimicry. Biomimicry is a concept in which the design of structures, materials and systems are modeled after nature and natural systems. Common examples include airplane wing design being modeled after birds, and wetsuit material being inspired by beavers.
Inspired by the swarming behavior found in nature, the Vijay Kumar Lab’s new research paper, Swarm of Inexpensive Heterogeneous Micro Aerial Vehicles, aims to determine if a large number of autonomously functioning vehicles can be reliably deployed in the form of a swarm to carry out a desired task. With ModalAI’s highly-integrated, optimized size, weight, power and cost (SWAP-C), VOXL companion computer onboard, the Kumar Lab’s open swarm framework and reference designs will enable drone developers and researchers to innovate and achieve a faster time to market.
Swarms for Many Use Cases
While individual drones can be useful, a swarm of autonomous drones can fulfill many use cases. The researchers at the Kumar Lab intend to apply these swarms of micro aerial vehicles frameworks to specific applications like crop monitoring. Swarms of drones can fly over fields of crops and use their image sensors to identify progress of growth or any anomalies in the field, saving farmers time and money.
In addition to agriculture inspection, swarms can be used for search and rescue in dense or urban environments, where a fleet of drones can help first responders widen the field of view of their search. Other use cases include indoor or outdoor asset inspection, natural disaster recovery, or night time drone shows.
Scaling Drone Swarms with the Vijay Kumar Lab
Drone swarms aren’t a new concept, in fact, Vijay Kumar and his team at the University of Pennsylvania have been researching how to scale swarms of autonomous drones for the past 17 years. In this new research venture, the Kumar Lab set out to build an algorithm and reference design that demonstrates collaborative communication of heterogeneous drone swarms. Traditionally, drone swarms are homogenous, meaning that every drone in the group is the same model and equipped with the same computing power and sensing. These swarms are typically easier to set up and run because each drone has the same technical architecture, and is programmed to run the same task. Heterogeneous swarms, however, are a more powerful yet complicated venture. Drones in a heterogeneous swarm are different, and can carry different computing and sensing power, meaning that the total capabilities from a heterogeneous swarm are greater than that of a homogeneous swarm. Where 100 identical drones in a homogeneous swarm would operate in unison, drones in heterogeneous swarms can each be programmed to perform a different task, thus optimizing the swarm’s efficiency.
Vijay Kumar Lab's fleet of autonomous, heterogenous micro aerial vehicles
In addition to widening the scope of the potential use cases of swarms, heterogeneous swarms operate with a stronger resiliency across communication and sensors. The Kumar Lab specifically used both WiFi and Zigbee mesh networks for communication to ensure that the several different sensors aboard the three drone models would have a robust network to communicate with.
The Kumar Lab project serves as a proof of concept and open source reference design for other research institutions or drone developers to replicate and improve. One important requirement for the Kumar Lab’s swarm project is that the autonomous drones in the swarm be micro, or SWAP-C optimized, meaning they are small in form factor, and low cost. The Kumar Lab was able to accomplish just that along with the advancement of computer vision techniques. The heterogeneous drones in the swarm from this project do not require any external infrastructure to operate autonomously. Traditional approaches to operating drone swarms rely on (RTK) GPS or external infrastructure such as motion capture or stationary UWB nodes to help pinpoint their location. The Kumar Lab has developed an algorithm for the swarms to operate without the external infrastructure.
VOXL Enables SWAP-Optimized Drone Swarms
Due to the nature of this project and requirements that the swarm hardware be SWAP-C optimized, the researchers at the Vijay Kumar Lab turned to ModalAI for our open, SWAP-optimized, Blue UAS Framework companion computer, VOXL. UPenn project lead Dinesh D. Thakur states
“we’ve partnered with ModalAI in the past for our various projects and continue to work with them because their U.S.-made autopilots like the VOXL deliver high-performance, reliability and operate on a developer friendly, open platform. VOXL is the only credit-card sized, cellphone grade, highly integrated PCB that is available in the market for full autonomy. The size is critical for us to make a 10cm fully autonomous MAV. No other vendor currently provides this”
VOXL powers three drones for the Kumar Lab Swarm Project: the Starling, Dragon DDK, and Dragonfly 230.
The Starling’s original design, created as a collaborative effort of ModalAI and Kumar Lab teams, is the star of this project. Weighing in at about 185g, the Starling is one of the smallest micro aerial vehicles that is enabled by the high computing power from VOXL, electronic speed controllers (ESCs), and a variety of sensing and communication modules. The goal of this platform was to combine the necessary mechanical and electronic components in a compact design that works well and is also easy to manufacture and tweak. Numerous factors were considered, such as safety, size, rigidity, vibration reduction, and ease of use. Smaller vehicles are more suitable for use in indoor test facilities or robotics labs, where many algorithms are initially developed and tested. Therefore, the Starling platform was designed to have a small footprint and weight while still providing the flexibility of a research platform.
VOXL Starling Reference Drone Coming Soon
The Kumar Lab research paper on Swarm of Inexpensive Heterogeneous Micro Aerial Vehicles will be presented at the 17th International Symposium on Experimental Robotics in Malta on November 15–18, 2021. The DOI for the book is 10.1007/978-3-030-71151-1_37 and is available to purchase here. The software related to this research is available here.
The Starling drone reference design and ModalAI VOXL SDK will be available to the public soon. Be sure to sign up for our newsletter to be the first to hear when the Starling is released.
Help Us Shape The Future of Swarming Micro UAVs
One of our core beliefs at ModalAI is that transparency fosters innovation. The team at ModalAI is excited to introduce the new compact, VOXL-powered, Starling drone reference design. As we are finalizing this release, we would like to make sure that we have taken into account the design considerations which are important to our customers. We are asking for your feedback, which will help us make final tweaks to the design in order to maximize its range of applications. To be a part of this collaborative innovation, share your thoughts in our survey below.