Home / Technology / AI & IT / DARPA OFFSET developing offensive swarm tactics with hundreds of drones and ground robots for diverse missions in complex urban environments

DARPA OFFSET developing offensive swarm tactics with hundreds of drones and ground robots for diverse missions in complex urban environments

Artificial intelligence is being employed in a plethora of defense applications including novel weaponry development, command and control of military operations, logistics and maintenance optimization, and force training and sustainment. AI also enhances of autonomy of unmanned Air, Ground, and Underwater vehicles.


It is also assisting in new concepts of operations like vehicle swarms in which multiple unmanned vehicles autonomously collaborate to achieve a task. This could confer a warfighting advantage by overwhelming adversary defensive systems. Swarming is where a military unit engages an adversary from all directions simultaneously, either with fire or in force, according to a report by the RAND National Defense Research Institute.


With more than half of the world’s population living in cities today – a portion that is projected by the United Nations to rise to two-thirds by 2050 – the US military is preparing for potential battles in metropolitan areas. Urban canyons—with their high vertical structures, tight spaces, and limited lines of sight—constrain military communications, mobility, and tactics in the best of times. These challenges become even more daunting when U.S. forces are in areas they do not control—where they can’t rely on supply chains, infrastructure, and previous knowledge of local conditions and potential threats.


Unmanned air vehicles (UAVs) and unmanned ground vehicles (UGVs) have long proven beneficial in such difficult urban environs, performing missions such as aerial reconnaissance and building clearance. But their value to ground troops could be vastly amplified if troops could control scores or even hundreds—“swarms”—of these robotic units at the same time.


The prime bottleneck to achieving this goal is not the robotic vehicles themselves, which are becoming increasingly capable and affordable. Rather, U.S. military forces currently lack the technologies to manage and interact with such swarms and the means to quickly develop and share swarm tactics suitable for application in diverse, evolving urban situations, says DARPA.


DARPA  launched its new OFFensive Swarm-Enabled Tactics (OFFSET) program in December 2016 with the goal of getting small-forces infantry to use “swarms of 250 unmanned aircraft systems (UASs) or small unmanned ground systems (UGSs) or more to accomplish diverse missions in complex urban environments.”  DARPA’s OFFSET program seeks to advance the integration of modern swarm tactics and leverage emerging technologies in swarm autonomy and human-swarm teaming. Cognitive autonomy has the potential to transform all defense and security systems,” explains Vern Boyle, vice president, advanced technologies, Northrop Grumman Mission Systems. “OFFSET will explore a variety of applications in relevant mission scenarios.” “We are applying cutting-edge technologies in robotics, robot autonomy, machine learning and swarm control to ultimately enhance our contributions to the warfighter.”


DARPA OFFSET program is Seeking ideas for Drone Swarms to Assist Troops

To help overcome these challenges and dramatically increase the effectiveness of small-unit combat forces operating in urban environments, DARPA has launched its  OFFensive Swarm-Enabled Tactics (OFFSET) program. OFFSET seeks to develop and demonstrate 100+ operationally relevant swarm tactics that could be used by groups of unmanned air and/or ground systems numbering more than 100 robots.


These swarm tactics for large teams of unmanned assets would help improve force protection, firepower, precision effects, and intelligence, surveillance, and reconnaissance (ISR) capabilities. OFFSET plans to offer frequent opportunities for engagement with anticipated end users in the U.S. Army and U.S. Marine Corps and would share successfully tested swarm tactics with them on a rolling basis.


“With the technologies and tactics to be developed under OFFSET, we anticipate achieving a deeper understanding of how large numbers of increasingly autonomous air and ground robots can be leveraged to benefit urban warfighters,” said Timothy Chung, DARPA program manager. “We aim to provide the tools to quickly generate swarm tactics, evaluate those swarm tactics for effectiveness, and integrate the best swarm tactics into field operations. If we’re successful, this work could also bring entirely new scalable, dynamic capabilities to the battlefield, such as distributed perception, robust and resilient communications, dispersed computing and analytics, and adaptive collective behaviors.”




Program Vision

An unprecedented confluence of advanced technologies is unleashing transformative capabilities, particularly in the context of distributed, large-scale, autonomous systems. The increasing availability of such systems has renewed and reinvigorated the exploration and acceleration of swarm capabilities.


The program intends to leverage rapidly emerging immersive and intuitive interactive technologies such as augmented and virtual reality, and voice, gesture, and touch-based technologies to create a novel command interface with immersive situational awareness and decision presentation capabilities.


These technological advances have converged to give rise to five key areas for swarm systems.

  • Swarm autonomy encompasses adaptive, complex, collective behaviors for intelligent movement, decisions, and interactions with the environment.
  • Human-swarm teaming highlights the need and ability to interact, influence, and infer swarm system behaviors.
  • Swarm perception involves large-scale, distributed and dispersed sensing, fusion, and distillation of information.
  • Swarm networking pertains to adaptive, resilient, fragmented sharing and storage of distributed information.
  •  Swarm logistics addresses both hardware and software deployment, support, and maintenance of large-scale systems.


The goal of the OFFensive Swarm-Enabled Tactics (OFFSET) program is to advance and accelerate elements of these enabling swarm technologies, focusing on the swarm autonomy and human-swarm teaming components of swarm system capabilities. OFFSET places specific emphasis on operationally relevant swarm tactics as the basis for leap-ahead advances in swarm capabilities. The Defense Advanced Research Projects Agency will be funding research into the design, development, and demonstration of novel frameworks to enhance how humans interact with autonomous swarms.


Swarm tactics specifically capitalize on the unique attributes of swarm systems to effect tactical advantage. Swarm size (the number of elements in the swarm) has often been the prevailing (albeit limiting) consideration when identifying advantages of swarm technologies. The underlying premise of the OFFSET program is that the potential of swarm systems has yet to be fully explored and realized. Swarm tactics that fully leverage the richness of swarm systems beyond swarm size, to include agent and collective complexity, heterogeneity in the swarm’s composition, and collaborative interactions between humans and the swarm, lead to disruptive swarm capability advantages.


The key benefit of designating “swarm tactics” as the focal point of OFFSET is to more clearly and intuitively capture commander’s intent by representing autonomous swarm capabilities in a tactical lingua franca. A common vocabulary would anticipate, if not accelerate, rapid advances in autonomy, artificial intelligence, platform designs, and perception and embedded component technologies. The swarm tactics-centric developments envisioned by OFFSET also aim to reduce the cognitive impedance mismatch that currently limits interactions with existing multi-robot system technologies.


The warfighter’s needs are directly dictated by the swarm mission, characterized by high-level operational objectives to be achieved by combined employment of swarm systems and associated human teammates. In contrast, swarm algorithms form the foundation of simple functions or “skills” that the swarm is capable of executing collectively.


As the central focus of OFFSET, swarm tactics catalyze operationally relevant swarm system capabilities via the composition of swarm primitives, conducted in sequence and/or concurrently, to address tactical objectives in support of the mission. While these functions (e.g., maneuver forward, measure signal strength, sense obstacles) may not individually offer actionable operational value, the combination of these swarm algorithms into swarm primitives, or collective behaviors, can represent more integrated swarm capabilities. For example, such swarm primitives may encode behaviors to locate points of interest in an area, identify ingress points to a building, or define and secure a perimeter.


Although a swarm system could dramatically enhance human execution of existing tactics for, e.g., isolating and clearing a building, suppressing enemy fires, or maintaining flank security, OFFSET intends to inspire the generation of innovative swarm tactics that disruptively create new opportunities for future swarm system capabilities.


One of the many potential operating environments where future swarm capabilities may demonstrate game-changing impact is in urban operations in dense urban environments. Particular challenges include the increasingly vertical, occluded, and channelized contexts in which small-unit ground forces must maneuver, defend, and engage both dynamic environments and adversaries. Such impediments are exacerbated by the fact that such operations must often be conducted in areas where knowledge, access, and/or control of factors like infrastructure, supply chains, local conditions, and potential threats are severely limited. The very nature of these dense urban areas calls for advances in distributed and dispersed unmanned system capabilities organic to company-level and below ground units of the future.


OFFSET envisions the ideation of novel swarm tactics via two complementary avenues, both offering relevant paths towards near-term and future swarm tactics-based capabilities:

  • Implement swarm tactics comprising only physically realizable components, that is, utilize only existing and/or maturing sensors, datatypes, embedded computing and network resources, as well as experimentally demonstrated swarm algorithms or swarm primitives.
  • Design swarm tactics using synthetic swarm components, that is, leverage access to data only available within the game environment to fabricate and/or enable futuristic or emerging swarm capabilities.
  • The former approach captures and challenges existing capabilities to lead to potential discovery of new swarm tactics. Examples of existing state-of-the-art technologies might include LIDAR, electro-optical, infrared vision sensors; occupancy grid-based decomposition of 3D environments; collision-free navigation and multi-robot path-planning algorithms; cooperative simultaneous localization and mapping (SLAM) solutions.

Alternatively, the latter approach illuminates and potentially identifies those components which may dramatically impact future swarm capabilities. Examples of notional synthetic technologies that could be encoded and simulated purely within the virtual game environment might include fictitious sensors able to localize and identify all personnel; futuristic mobility and endurance capabilities for air or ground platforms; or advanced networking technologies for highly resilient communications.

Both approaches aim to inform the complex, multi-faceted considerations of swarm systems with the goal of iteratively and interactively deriving insights into swarm technology gaps as well as identifying potential swarm capabilities that give rise to strategic surprise.

To accomplish these goals, OFFSET seeks to develop an active swarm tactics development ecosystem and supporting open systems architecture, including:

  • An advanced human-swarm interface to enable users to monitor and direct potentially hundreds of unmanned platforms simultaneously in real time. The program intends to leverage rapidly emerging immersive and intuitive interactive technologies (augmented and virtual reality, voice-, gesture-, and touch-based) to create a novel command interface with immersive situational awareness and decision presentation capabilities. The interface would also incorporate a swarm interaction grammar, similar in concept to playbooks coaches in soccer, basketball, and other games prepare with pre-made plays combined with “freestyle” design tools that allow dynamic action and reaction based on real-time conditions in the field.
  • A real-time, networked virtual environment that would support a physics-based, swarm tactics game. In the game, players would use the interface to rapidly explore, evolve, and evaluate swarm tactics to see which would potentially work best on various unmanned platforms in the real world. Users could submit swarm tactics and track their performance from test rounds on a leaderboard, as well as dynamically interact with other users.
  • A community-driven swarm tactics exchange. This curated, limited access program portal would house apps to help participants design swarm tactics and combine collective behaviors, as well as swarm algorithms. It would provide these key ingredients to an extensible architecture for end-user-generated swarm tactics and help create a lasting community to innovate and cultivate the most effective tactics, with the potential to integrate third-party tactics and playtesters in the future.
  • OFFSET aims to demonstrate its technologies through frequent live experiments with various unmanned air and ground platforms. Every six months, operational vignettes of progressively increasing complexity would challenge both the swarm architecture and the developed swarm tactics across numerous technological and operational test variables, such as swarm size, proportion of air and ground platforms, and mission duration. Users would employ the swarm interface to test the best of the virtual tactics in the real world, and interactively supply their unmanned platforms with near-real-time tactics updates using automated deployment technologies.

“We’re interested in developing practical swarm systems in an agile way, so future operators will have the tools they need to outsmart and outperform urban adversaries,” Chung said.


If successful, OFFSET will produce an advanced swarm system, comprising a demonstrated swarm software architecture with implementation of swarm tactics and advanced swarm interfaces; a physical swarm system testbed for substantive experimentation and operationalization; and a robust developer and user community for enduring engagement in the advancement of swarm system capabilities. The insights derived from the OFFSET program will inform not only the technological trade spaces of swarm systems design, but also and more broadly, the scalability of manned-unmanned teaming constructs, test and evaluation for autonomous systems, and open system architectures for distributed, networked capabilities.


DARPA Awards

On October 12, 2017 DARPA announced plans to  solicit ideas for a drone swarm system that could  overcome the challenges of Urban canyons—with their high vertical structures, tight spaces, and limited lines of sight and dramatically increase the effectiveness of small-unit combat forces operating in urban environments.


The swarm sprints aim to encourage rapid innovation and continuous incorporation of breakthrough technologies. Each of the five core sprints emphasizes one of the key OFFSET thrust areas – swarm tactics, swarm autonomy, human-swarm teaming, virtual environment, and physical testbed – to ultimately enable cross-cutting breakthroughs in swarm systems capabilities.


DARPA  has awarded contracts to the following organizations for the second Swarm Sprint: Carnegie Mellon University Corenova Technologies, Inc., Cornell University, Heron Systems Inc., Michigan Technological University, Siemens Corporation, Corporate Technology,  University of Colorado, Boulder, and University of North Carolina, Charlotte.


DARPA  selected Northrop Grumman Corporation as a Phase 1 Swarm Systems Integrator for the Agency’s OFFensive Swarm-Enabled Tactics (OFFSET) program. Northrop Grumman  launched its first open architecture test bed as part of the program, for participants to create and test their own swarm-based tactics on the platform.

This cohort of sprinters for the second swarm will leverage existing technologies or develop new ones leading to enhancements in swarm autonomy, highlighting enriched foundations in swarm primitives and/or swarm algorithms. These contributions to the overarching OFFSET swarm systems architectures will offer additional building blocks for novel capabilities that spotlight the potentially disruptive advantages of autonomous swarms for complex urban operations.


The Defense Advanced Research Projects Agenc (DARPA) in Oct 2018, solicited proposals for its third swarm sprint, which will focus on the topics of human-swarm teaming and swarm tactics through its OFFensive Swarm-Enabled Tactics (OFFSET) program. The first of two topic areas covered in the third sprint is human-swarm teaming, which refers to the design, development, and demonstration of novel frameworks to enhance how humans interact with autonomous swarms. This topic area recognizes and seeks to address the complexity of the swarm systems themselves, as well as the cognitive, physical, and contextual needs of human teammates or tacticians when conducting urban operations.


The third swarm sprint also seeks to increase the collection of swarm tactics found in the OFFSET swarm tactics exchange, a repository of tools and tactics – or approaches – sprinters can employ. Selected swarm sprinters will design and implement additional sophisticated swarm tactics employing a heterogeneous swarm of air and ground robots, and addressing the mission context of “conducting an urban raid” within four square city blocks over a mission duration of one-to-two hours.

The fourth swarm sprint consists of two topics areas: developing synthetic technologies in the OFFSET virtual environments and identifying applications of artificial intelligence (AI) to discover and learn novel swarm tactics. For the first topic area, proposers will seek to develop and implement synthetic capabilities in simulation, representing potential future technologies, such as distributed “see-through-wall” sensors, passive swarm communications, or enhanced sensor/computing arrays, to enable and demonstrate novel swarm tactics. Proposed technologies could be near-term advances that are being prototyped in laboratories or far-term ideas that are primarily conceptual but physically grounded.


The second topic area represents an ad hoc sprint for leveraging artificial Intelligence for accelerating swarm tactics design, for which sprinters will discover, learn, and harden novel swarm tactics by applying artificial intelligence frameworks via enhancements of the OFFSET virtual environments.


DARPA has awarded contracts to the following organizations to focus on either enhancements to the OFFSET virtual environments or applications of artificial intelligence (AI) in the fourth swarm sprint:

Thrust Area: Virtual Environment to Johns Hopkins University Applied Physics Lab, Michigan Technological University

Thrust Area: Applications of AI to Carnegie Mellon University, Charles River Analytics, Inc., Research Foundation for SUNY; University at Buffalo, Siemens, Soar Technology, Inc., Spatial Integrated Systems


The OFFSET program’s fifth swarm sprint, focused on two topics: operational swarm tactics in support of urban missions and innovative technologies to prototype and integrate into physical swarm system testbeds. Swarm Sprinters in either topic in the fifth swarm sprint will integrate their technologies into one or both of the OFFSET swarm systems architectures, and will have the chance to further develop and demonstrate their innovative solutions in relevant field tests. OFFSET swarm sprints aim to encourage rapid innovation and continuous incorporation of breakthrough technologies. Each core sprint emphasizes one of the key OFFSET thrust areas. Previous calls have sought submissions on the topics of swarm tactics, swarm autonomy, human-swarm teaming, and virtual environment.


The Defense Advanced Research Projects Agency has selected eight industry and academic entities for the fifth OFFensive Swarm-Enabled Tactics tech sprint focused on conducting human-machine teaming activities in urban environments. Physical testbed activities cover hardware modifications and the integration of capabilities such as sensors, navigation systems, aerodynamic elements and mobility features. Participants under this category include Johns Hopkins University Applied Physics Laboratory, Michigan Technological University, HDT Expeditionary Systems, Sentien Robotics and Texas A&M University. Swarm tactics participants will work to design and implement concepts for enacting combat strategies in dense urban terrain.


Examples of potential new technologies for integration into the swarm system testbeds include, but are not limited to, swarm sensors, swarm fielding technologies, swarm communications approaches, modular platforms, and mechanisms for swarm manipulation. “We’re looking for imaginative proposals for component technologies that, when brought together with our OFFSET physical testbeds, can lead to technological breakthroughs for future swarm capabilities. says Timothy Chung, program manager in DARPA’s Tactical Technology Office (TTO). “The ability to enhance commercial-off-the-shelf (COTS) air and ground platforms can lead to new modes of swarm operations and disruptive swarm tactics for warfighters.


The fifth swarm sprint also seeks to increase the collection of swarm tactics found in the OFFSET swarm tactics exchange, a repository of tools and tactics sprinters can employ. Selected swarm sprinters will design and implement additional sophisticated swarm tactics employing a heterogeneous swarm of up to 250 air and ground robots, and addressing the mission context of “seize key urban terrain” within eight square city blocks over a mission duration of four-to-six hours.


Maryland researchers awarded DARPA grant to develop robotic swarm strategies

A team of four University of Maryland researchers has been awarded a nine-month, $646K cooperative agreement from the Defense Advanced Research Projects Agency (DARPA) for “Robust Semi-Autonomous Swarm Tactics for Situational Awareness in Uncertain Environments.” Assistant Professor Huan Xu (AE/ISR) is the principal investigator.


The grant is among the first “core swarm sprints” awarded in DARPA’s OFFensive Swarm-Enabled Tactics (OFFSET) program. In addition to the University of Maryland, the cohort includes Carnegie Mellon University, Lockheed Martin Advanced Technology Laboratories, SoarTech, Inc., and Charles River Analytics, Inc.


The OFFSET program is funding “swarm” autonomy and human-swarm teaming technology research to improve the capabilities of groups of aerial and ground robots that work together to help humans accomplish tasks such as surveillance in complex and changing urban settings. In current drone surveillance situations, robot teams are limited in number, use “one-size-fits-all” surveillance strategies, and assume the environment is static. However, since environments are complex and change frequently, better surveillance strategies are needed.


The Maryland researchers will combine multiple existing autonomous surveillance strategies into a single framework, enabling simultaneous overlapping deployment in different subsets of the environment. They will produce a set of semi-autonomous aerial swarm tactics that can be used simultaneously in different parts of the environment and will account for uncertainties, sensor errors and dynamic obstacles. The goal is to develop and evaluate new swarm simulation algorithms that can automatically adapt to the complexity of urban environments.


Humans that use such a system would be able to fine-tune three different semi-autonomous surveillance strategies, which would lead to improved surveillance capabilities in urban environments. The researchers will develop their strategies during the first six months of the project. In the final three months these strategies will be tested on robotic platforms.



Charles River Analytics Inc. receives DARPA grant to develop AI based swarm tactics

Charles River Analytics Inc., developer of intelligent systems solutions, received funding from the Defense Advanced Research Projects Agency (DARPA) under the OFFensive Swarm-Enabled Tactics (OFFSET) program to develop Meta-Reinforcement Learning Innovation for Robust Swarm Tactics (MERLIN-RST). MERLIN applies a meta-reinforcement learning approach to discover and learn novel swarm tactics.


Through biology-inspired algorithms and deep machine learning, tactics in the OFFSET program are intended to help swarms achieve and adapt to mission objectives. According to the company, human programmers cannot support the evolution of these tactics quickly or accurately enough due to numerous variants in adversary strategy and evolving environments that make manual development a challenge. Under the MERLIN effort, researchers will augment OFFSET simulators to model swarm performance in a wide range of challenging urban environments as part of the OFFSET Program’s Fourth Swarm Sprint.


DARPA to test whether single user can control 200 drones, reported in Oct 2021

The Defense Advanced Research Projects Agency is gearing up for a major exercise in which it plans to demonstrate a single user controlling a swarm of about 200 unmanned systems in an urban environment. The November capstone event at Fort Campbell, Kentucky, is part of DARPA’s OFFensive Swarm-Enabled Tactics (OFFSET) program, which envisions smaller units able to mass up to 250 small aerial and ground unmanned systems in urban areas.


One of the major integrators on the effort, Northrop Grumman, will seek to test its technology that allows a single user via a tablet to control 200 vehicles. Raytheon Technologies is another key player in the exercise. Northrop’s open-architecture system, dubbed Rapid Integration Swarm Ecosystem, allows an operator to draw a sketch on a tablet, which tells the unmanned systems to perform a specific task. Cherry described this as the “swarm grammar.” There could be preplanned or user-defined swarming grammar, and each activity might mean something different based on that swarm grammar.


For example, she said, the user could circle a building on the tablet, telling the unmanned systems to converge on the structure. Or the user could draw a question mark over a target, which denotes a specific task based on predetermined swarm grammar, she added. Right now, the program is primarily focused on assigning intelligence-gathering tasks to the swarm, but Cherry also described using the unmanned systems to overwhelm an adversary. “Honestly, the sky is the limit here. On OFFSET, we’ve really tried to be as broad as we can,” she said.


The service has shown interest, and Cherry said the company believes its technology could easily integrate with other Army efforts, such as the Integrated Visual Augmentation System, which would provide soldiers with augmented reality and situational awareness data in a heads-up display — with the next notional step being passing swarm data down to soldiers through that system. In the interim, the company wants to mature its technology by getting it in the hands of soldiers at a real-life exercise and receiving important feedback.


There are plans for Northrop to test its technology at the Army Expeditionary Warrior Experiment early next year, and the company has asked to participate in Project Convergence 2022, the service’s annual campaign of learning.



References and resources also include:










About Rajesh Uppal

Check Also

DARPA FoundSci: Revolutionizing Scientific Discovery with Autonomous Scientists

Introduction: The Defense Advanced Research Projects Agency (DARPA) is renowned for its cutting-edge research initiatives …

error: Content is protected !!