SwarmMATE™: Swarming Drones for Real-world Missions

John A. Sauter SWARMMATE™: SWARMING DRONES FOR REAL-WORLD MISSIONS

The idea of unmanned drones flying together in expansive swarms may at first sound chaotic and unpredictable – yet with the right approach, these machines can enable their users to carry out coordinated and highly advanced missions. Through their research, John Sauter and colleagues at SoarTech have developed SwarmMATE™ – a transformative software that enables drones to mimic the behaviours of swarming animals found in nature. Their architecture now paves the way for real-world applications – from tracking forest fires and storms to aiding rescue missions.

Complex Swarms deeply intricate financial ecosystem. top-down planning approaches,’ says In each of these scenarios, there is no Sauter. ‘But those approaches simply From the seas to the skies, swarms of central authority telling each individual don’t scale when you are talking about insects, birds and fish can be found exactly what to do; instead, complexity tens, hundreds, or thousands for across the natural world. The reasons emerges from sets of simple and unmanned systems.’ why these animals collect into swarms intuitive rules, which enable individuals can be highly varied – but in every case, to collectively achieve impressive feats. Within such an intricate web of properties emerge in these expansive interacting machines, system designers groups that are vastly more complex Recently, these behaviours have seen must determine how individual drones than any individual animal could increasing interest across a wide variety should best contribute to the overall conceive on its own. As John Sauter of fields of research. In their work, goals of a mission, while also assessing explains, this has led to some highly Sauter and his colleagues at SoarTech the performance of the overall system. sophisticated feats of engineering. study one particularly fascinating area: When operating in environments with swarms of unmanned aerial drones, unpredictable terrains, obstacles, and ‘Natural systems have devised means to which work together to carry out weather conditions, it can be extremely coordinate the activities of many simple complex tasks. challenging for designers to come agents for accomplishing complex up with plans for each drone that tasks,’ he says. ‘Termites build large Similarities in Design ensure the whole swarm achieves the complex mounds with floors, chambers, objectives of the mission. arches, and cooling all without the aid In recent years, technological advances of a single architect. Ants can find and have led to cutting-edge designs of While studying this problem in the build shortest paths from distant food compact, lightweight, and inexpensive mid-80s, Dr Van Parunak, a colleague of sources to their nests without using a drones; capable of communicating Sauter, turned to natural systems for a navigation app.’ with each other even as their positions possible answer. There, he found many continually change. In principle, these examples of complex swarm behaviours Although we may not realise it, similar characteristics are ideal for using even though individually, the members phenomena are also present in our own swarms of drones to carry out complex were not doing anything very society. For example, in the markets that tasks. However, such an ambitious goal sophisticated. Through their analysis drive our economy, the simple acts of is faced with numerous challenges. of swarms found in nature and human buyers and sellers aiming to maximise ‘Researchers have tried to control society, Sauter’s team developed their own profits each contribute to a these large systems using traditional advanced algorithms that mimic the

WWW.SCIENTIA.GLOBAL Real-world Missions

By fitting drones with scientific instruments such as sensors and cameras, they can become highly sophisticated tools in the coordination of advanced, potentially dangerous missions. Such a system now displays clear potential for immense societal benefits. ‘We aim to coordinate robotic swarms for performing missions that are useful for human society,’ describes Sauter. ‘We can scale from just a few robots to massive swarms comprising thousands of robots using the same algorithms, so we can solve both the small and really large problems we face.’

These capabilities have already been explored for military applications – but ability of groups of animals or insects to Thirdly, SwarmMATE™ has tools to increasingly, the SoarTech team have overcome these challenges, producing measure and verify that the swarm come to examine the immense societal a result much more powerful than the behaviours are consistent with the benefits they could soon unlock. ‘A sum of its individual members. goals of their mission, to a high degree common theme among these missions of confidence. Finally, SwarmMATE™ is the need to cover large, irregularly ‘By studying these systems, we’ve includes an execution framework that shaped, dynamically changing identified some common design enables a highly distributed swarm to areas,’ Sauter continues. ‘Swarms principles that determine how they communicate and coordinate using are an ideal way to organise a team organise and coordinate their activities,’ the swarming algorithms developed by of vehicles to distribute the tasking, Sauter illustrates. ‘We can now create SoarTech. and dynamically adapt to changing artificial societies in our computers, with mission requirements.’ Among the many many agents operating under simple Together, these measures enable applications this enables is the tracking rules; and train them to solve some engineers to develop high-performance of forest fires and severe weather; really hard problems.’ With over 20 years drone swarms that are robust against informing search and rescue operations; of experience conducting these tests obstacles, and which can adapt to and monitoring agricultural crops. and studies, the researchers at SoarTech dynamic, unpredictable conditions in are now well equipped for making the the field. By following the algorithms Mapping Forest Fires widespread societal use of swarming built into the software, drones can both drones a step closer to reality. cover larger areas within regions that Forest fires have grabbed many would be hard to reach from the ground, headlines in recent years – unleashing Introducing: SwarmMATE™ while performing more coordinated widespread devastation both on natural manoeuvres to maximise the amount of environments, and the communities The team’s efforts have now culminated information they gain. living within them. In the face of this in the launch of their tool – Swarming threat, there is a pressing need for Mobile Autonomous Teaming ‘Through our research, we have firefighting operations to monitor and Environment (SwarmMATE™) – first developed a truly revolutionary map the spread of fires in real time, to introduced in 2017. This advanced software architecture that encapsulates determine where and how quickly the framework addresses four core the best of what I’ve learned about fire is moving and where best to deploy challenges associated with drone swarm intelligence and control,’ Sauter limited firefighting resources. As Sauter swarm design and deployment: firstly, describes. ‘SwarmMATE™ is a platform describes, SwarmMATE™ algorithms can it breaks down the overall goals of that coordinates the behaviours of coordinate drone swarms to perform a mission into sets of smaller tasks groups of robots performing a range of these functions. that can be performed by a single missions.’ With these measures in place, drone or team of drones. Secondly, Sauter’s team could next focus their ‘We developed swarming algorithms the software determines the sizes of attention on applying drone swarms in to control a swarm of air vehicles with swarms to fit the needs of the mission. specific scenarios. thermal sensors to search a large,

WWW.SCIENTIA.GLOBAL mountainous, forested area for forest fires,’ he says. ‘This deep into valleys where it is difficult for manned aircraft to involved finding the hot spots, patrolling around the perimeters reach.’ Equipped with tools such as infrared sensors, radar of fires, mapping their extents, and then continuing to track and and radios, these swarms can search remote areas for signs update maps of their perimeters as they spread.’ By consulting of human life, while also providing cellular connectivity where these maps, authorities would be far better equipped to direct there isn’t otherwise any service available – allowing victims to firefighting teams to combat blazes as efficiently as possible. communicate their positions.

Tracking Severe Weather Monitoring Agricultural Crops

Through further studies, Sauter’s team has shown how swarms As farmers around the world aim to improve the sustainability can also be used to monitor extreme weather events – which of their agricultural practices, while maintaining high yields to can take place over expansive areas and ranges in altitude. meet growing demands for food, they are increasingly seeking Their techniques enable far more sophisticated measurements to minimise their use of chemicals and resources. Using than those made by traditional instruments such as weather drone swarms fitted with cameras capturing a wide region of balloons and dropsondes dropped from manned aircraft, which the electromagnetic spectrum, they can scan large areas of have extremely limited monitoring ranges and operation times. farmland in a short space of time; identifying any areas where crops are suffering from factors such as insect-related disease When equipped with weather sensors, swarming drones can or water stress. smartly organise themselves into positions that give them the best possible meteorological data. ‘Data needs to be As a result, these farmers can pinpoint exactly where additional collected on wind speeds, direction, temperatures, pressure water or chemical applications are needed. ‘A swarm of drones and humidity, to better predict the severity and direction a can then deliver just the right kinds of chemical only when storm,’ Sauter illustrates. ‘The swarm has the intelligence to and where it is needed, instead of the current practices of adapt to the changing weather patterns as it moves across a broadcasting these chemicals everywhere as a preventative region, ensuring continuous coverage and tracking.’ In turn, this measure,’ Sauter explains. This would enable farmers to greatly will enable meteorologists to generate far better forecasts of reduce the environmental impact of their operations – both how storms will progress over time, giving communities time to reducing water usage in their local regions, and making fields prepare. more friendly to plants and animals that pose no threat to crops. Informing Search and Rescue Diverse Societal Benefits Fast, thorough search techniques are particularly crucial for saving lives in search and rescue operations. Whether searching While intelligent drone swarms hold the potential to for hikers lost in the wilderness, or victims caught in disaster transform operations in each of these areas, the full extent zones, these operations will often involve scanning expansive, of their capabilities extend far further still – ranging from remote, and possibly dangerous regions with difficult terrain. law enforcement and drug interception, to logistics and These are extremely difficult to cover extensively using manned traffic monitoring. In the future, Sauter and his colleagues at aircraft – but are far less of an issue for intelligent drone SoarTech will continue their efforts to expand the reach of swarms. their SwarmMATE™ software in numerous aspects of society. Ultimately, the team hopes that the benefits enabled by ‘An intelligent swarm of vehicles can cover a much broader swarming drones could soon become a key feature of our area in a shorter time than manned aircraft and for much less everyday lives. cost,’ says Sauter. ‘They can fly closer to the ground and peer

WWW.SCIENTIA.GLOBAL Meet the researcher John A. Sauter Soar Technology, Inc. Ann Arbor, MI USA

John Sauter has over three decades of experience in research FUNDING and development. He has a particular interest in using artificial and swarm intelligence for the control of complex systems DARPA – including systems of multiple vehicles, and the scheduling AFRL and control of distributed logistics. Sauter has served as NAVAIR Principal Investigator on numerous projects at DARPA, where US Army he pioneered the development of one of the first swarming algorithms enabling fully distributed control over multiple FURTHER READING unmanned vehicles. He has worked at Soar Technology, Inc (SoarTech) since 2013, where he is Director of Autonomy. Sauter JA Sauter, K Bixler, S Kitchen, R Chase, RF emitter localization is also an Instructor for the Unmanned Vehicle University, with robotic swarms, In Unmanned Systems Technology XXII, where he designed and teaches the graduate-level course, 2020, April (Vol. 11425, p. 114250D), International Society for Autonomous Unmanned Systems. Optics and Photonics.

JA Sauter, K Bixler, Design of unmanned swarm tactics for an CONTACT urban mission, In Unmanned Systems Technology XXI, 2019, May (Vol. 11021, p. 110210K), International Society for Optics E: [email protected] and Photonics. W: https://soartech.com/ JA Sauter, K Bixler, SwarmMATE™: a swarm engineering and KEY COLLABORATORS verification environment, In Unmanned Systems Technology XXI, 2019, May (Vol. 11021, p. 110210N), International Society for Dr VanParunak, ABC Research Optics and Photonics. Kellen Bixler, SoarTech

WWW.SCIENTIA.GLOBAL