Drone Swarms: A Transformational Technology
Back in 2016, hundreds of Intel drones flew over the skies at Disney World, choreographed to music as part of a holiday show titled “Starbright Holidays.” Intel’s drones also performed behind Lady Gaga during the Super Bowl in 2017. And earlier this year, China celebrated Chinese New Year with a massive drone show. Although these displays are impressive and are certainly a technological achievement, they are prescripted and do not demonstrate the transformational potential of swarms of autonomous drones that can respond to, and act on, their environment.
Any quick search of recent literature will spotlight applications already being tested, along with many other envisioned future applications that are far more demanding than drone shows. Reconnaissance missions in areas of conflict or natural disasters, delivery of emergency supplies and everyday packages, fire suppression, search and rescue missions, space exploration, weather monitoring and forecasting, patrolling borders and wildlife preserves, and mapping on land and underwater are just some of the many examples of applications for this incredible technology.
The potential for deployment of a revolutionary set of battlefield tactics is driving the US military to prioritize development of this critical technology as a means to reduce risk to our warfighters. Obvious contributions are drone swarms that monitor the threat environment and defend military assets against enemy attacks. Swarms of drones, each relatively low in cost and collectively difficult to thwart, could replace expensive systems and platforms that are vulnerable to attack in various missions, such as those involving intelligence, surveillance, and reconnaissance (ISR). In other settings, large numbers of drones could overwhelm an adversary’s ability to attack our military personnel, allowing them to complete their mission unimpeded.
The Defense Advanced Research Projects Agency (DARPA) describes one such application: “DARPA’s OFFensive Swarm-Enabled Tactics (OFFSET) program envisions future small-unit infantry forces using swarms comprising upwards of 250 unmanned aircraft systems (UASs) and/or unmanned ground systems (UGSs) to accomplish diverse missions in complex urban environments. By leveraging and combining emerging technologies in swarm autonomy and human-swarm teaming, the program seeks to enable rapid development and deployment of breakthrough capabilities.”
What makes a swarm of drones superior to past technology is the fact that the drones communicate with one another to adjust the behavior of the entire swarm in response to real-time information collected from sensors on the drones in the swarm. As new information is gathered, the swarm can maneuver or otherwise adjust its behavior.
Response decisions can be programmed into the swarm’s control system or a single operator can manage the swarm as an entity. To make the swarm less vulnerable to disruption, all the members of the swarm can be programmed to take on “leadership roles,” or swarm-level behavior can be designed to emerge from the independent, minimally coordinated contributions of each drone, foregoing a traditional leadership role.
Drone swarms are, in every sense, a transformational technology. They can replace humans in dangerous or hostile environments. In sufficient numbers, they can collect information from multiple locations and directions, integrating it to form insights not otherwise available. That information can inform decision makers who cannot enter the environment nor capture information from multiple angles and perspectives. Drones can operate in environments—space, underwater, fires, battlefields—that are otherwise impossible or too risky for humans or alternative technologies. And drones can leverage advances in many technologies, such as smart sensors, applying them in otherwise impossible applications.
Successful widespread implementation of drone swarm technology will require continued advances in a number of areas. Progress has already occurred to the point that testing is underway by many organizations, but there is significant room for further advances. The following capabilities will eventually define the success of drone swarm technology and the applications to which it can contribute.
Autonomous operations are critical to the success of swarms because coordinated movement and action is their fundamental contribution. Intel’s entertainment programs mentioned earlier have coordinated up to 500 individual drones, and reports suggest that China has successfully tested swarms of 2,000 drones, about the number used in their New Year’s celebration. However, these applications depend on pre-established choreography that specifies the swarm’s formations. The DARPA OFFSET program involves autonomous swarms of 250 or more drones to enable real-time information-based decisions about how to best achieve the objectives by the swarm, independent of direct guidance by operators.
Although autonomy is essential, at the same time the swarm must be managed. The tacticians in command of a swarm must be able to understand the swarm’s assignment, progress, and health without being directly responsible for the activity of each vehicle in the swarm. This is not only relevant in the design phase, but is also a crucial element of the control environment in many applications. Not all of the information that the swarm collects can be anticipated, and in some applications, the interface between the swarm and the controller becomes essential.
Human-swarm interfaces that support interactions with swarms must be efficient and effective and recognize that some applications in conflict and disaster settings require near-instantaneous decision making. An important challenge is the design of both the interface through which data is collected and disseminated among the swarm’s members, and the processing system that translates that data into meaningful and relevant decision-support information.
There is clearly a balancing act between the need for autonomy and the importance of management by an operator. Swarms must be able to perform well-defined behaviors without central controllers, and in some cases should be capable of completing a mission on their own. In other situations, it is essential that unexpected information be recognized and trigger an appropriate reaction in a timely fashion from a human operator.
Safety is another factor critical to many applications. The most significant concern with not having an operator directly piloting each drone is the apparent loss of positive control over the vehicles. This factor is relevant not just because of the environments in which drone swarms might be operating (e.g., battlefields, disaster settings), but also because there is the potential for things to go wrong.
Individual drones could fail because of mechanical issues, environmental factors, or counterattacks. Communications within the swarm could be disrupted if a “leader drone” simply malfunctions or an adversary takes aggressive hostile action (e.g., jamming, etc.). Addressing swarm safety and resiliency is often application specific, requiring insights into situational challenges beyond those that are omnipresent in standard theaters of operation. Creating control systems to ensure safety via real-time responses to threats is a critical challenge for any application.
In summary, swarms of drones have enormous potential across a broad spectrum of applications. Although implementation by the military is already a priority, significant opportunities also exist in the civilian domain.
Advancements in the field have been swift, with the recent introduction of prototype systems designed for applications that could hardly have been imagined a decade ago. Progress will continue in the future, with key challenges involving the balancing act between autonomy and control and the need to develop situation-specific assurances that the swarm can be managed safely, avoiding problems due to either malfunction or hostile action. The likelihood is great that these challenges can be met and that swarms of drones will become a widespread phenomenon.
This article was written by Dr. Spencer Lynn, Senior Scientist; David Koelle, Principal Software Engineer; and Rich Wronski, Vice President of the Sensing, Perception, and Applied Robotics Division; at Charles River Analytics (Cambridge, MA). For more information, visit here .