Technical Digest

With the emergence of fifth-generation (5G)-cellular and Internet of Things technologies, alongside legacy wireless systems, the radio frequency (RF) spectrum is becoming heavily congested,creating availability and throughput challenges for wireless service providers as well asour military forces. In addition to the congestion, our forces face threats from advanced jammingand cyberattack. This competition for use and control of the RF spectrum is one of thecritical challenges facing the nation. Exemplifying its focus on making critical contributionsto critical challenges, the Johns Hopkins University Applied Physics Laboratory (APL) is pleasedto present this issue of the Johns Hopkins APL Technical Digest. The issue describes the Colosseum—the first-of-its-kind wireless communication test bed that APL developed for the DefenseAdvanced Research Projects Agency (DARPA) Spectrum Collaboration Challenge (SC2). This testbed enables research into artificial intelligence and machine learning for networked systems—systems that autonomously negotiate use of the RF spectrum in real time based on instantaneoususer demand, available spectrum, and environmental conditions. Through this effort, APL hasprovided a new capability enabling research beyond constrained RF allocations, ultimately contributingto improved resilience of networked systems on the battlefield.

As fifth-generation (5G) cellular technology emerges, it is apparent that the radio frequency (RF)spectrum is constrained by the ever-growing demands of application bandwidth and the numberof devices vying for that bandwidth. Databases and procedures for managing the spectrum havebecome very complex and do not scale to meet today’s on-demand spectrum requirements. Inpursuit of novel methods to overcome these limitations, the Defense Advanced Research ProjectsAgency (DARPA) launched the Spectrum Collaboration Challenge (SC2) in 2016. The goal of thechallenge, which culminated with the third and final competition in the fall of 2019, was to inspireparticipants to research, develop, and systematically test artificial intelligence algorithms across anetwork of radios to find the future paradigm for ensuring that the RF spectrum can support thebandwidths that next-generation applications will require. In support of SC2, the Johns HopkinsUniversity Applied Physics Laboratory (APL) designed, developed, and hosted the Colosseum, thefirst-of-its-kind wireless research test bed in which competitors tested their algorithms and conductedtheir experiments in competition events. In addition to introducing SC2 and its goals, thisarticle briefly describes the test bed architecture and the challenge events.

Development and management of the Colosseum, the wireless communications research test bed forthe Defense Advanced Research Projects Agency (DARPA) Spectrum Collaboration Challenge (SC2),was a complex undertaking. With its world-class expertise in communication systems and experiencein information technology infrastructure, the Johns Hopkins University Applied Physics Laboratory(APL) was well positioned to design, host, and maintain the Colosseum on its Laurel, Maryland,campus from 2016 to 2019. The effort required close coordination among members of the APL teamand between APL and DARPA. To effectively and efficiently manage the design and maintenance ofthe Colosseum, APL applied tested project management tools and techniques, a development andoperations approach, and an agile framework. This article focuses on the early planning and initialdevelopment efforts and documents the project management attributes, including the compositionof the APL team as well as the software tools, that contributed to the success of the effort.

The Johns Hopkins University Applied Physics Laboratory (APL) developed a complex test bedof software and hardware called the Colosseum to support the Defense Advanced ResearchProjects Agency (DARPA) Spectrum Collaboration Challenge (SC2). Following a developmentand operations (DevOps) approach was critical to the team’s ability to design and build theColosseum. Such an approach enhances collaboration between operations and developmentteams and takes advantage of technology, particularly automation tools. Tasks for the DevOpsteam included developing software codebases, deploying system configurations, and monitoringhardware system status such as power levels, system temperature, fans, and system uptime.The team accomplished these tasks by following a DevOps approach and using a variety of toolsets. This article describes the processes and tools the team used to design, build, and maintainthe Colosseum.

A key component of success in the Defense Advanced Research Projects Agency (DARPA) SpectrumCollaboration Challenge (SC2) was ensuring that each competitor had fair access to thelimited physical resources available in the competition. The Johns Hopkins University AppliedPhysics Laboratory (APL) designed and developed a custom Resource Manager as part of the Colosseum,the wireless research test bed at the foundation of the SC2 competition. By allocatingresources through a token system, the Resource Manager ensured that competitors had fair andequal access to resources in the Colosseum. The Resource Manager also provided mechanismsfor automated experiment handling and orchestration that increased the scheduling efficiencyof the resources and gave competitors equal access to all 128 nodes in the Colosseum. From 2016to 2019, the Resource Manager maintained continuous availability of Colosseum resources thatenabled international competitors to develop new artificial intelligence algorithms for radio frequency(RF) spectrum management.

One of the major constructs of the Defense Advanced Research Projects Agency (DARPA) SpectrumCollaboration Challenge (SC2) framework was the standard radio node (SRN). The JohnsHopkins University Applied Physics Laboratory (APL) designed the Colosseum, the massive wirelesstest bed behind SC2, and the SRN within it. The SRN provided SC2 competitors a software-definedradio (SDR) as well as compute and storage node resources so that they could develop, test, anddemonstrate collaborative intelligent radio network (CIRN) solutions. The SRN was designed todynamically allocate and de-allocate competitors’ container images while providing them completeaccess to and control of physically attached SDR and network resources. The SRN ensuredthe competition’s security, integrity, and fairness and isolated each competitor’s files and software.This article discusses the SRN’s architecture and its supporting and commanding systems, includinglocally managed services and processes.

The application of artificial intelligence and machine learning promises to usher in a new paradigmfor emerging wireless communication systems. The goal of the Defense Advanced ResearchProjects Agency (DARPA) Spectrum Collaboration Challenge (SC2) was to push this new paradigmforward. However, legacy radio systems already in place, such as radars for weather monitoring,receivers for spectrum monitoring, and battlefield jammers, will remain in use for a long time.Therefore, intelligent radios must operate around and adapt to these legacy systems to avoid interferingwith them. In support of DARPA’s SC2, the Johns Hopkins University Applied Physics Laboratory(APL) designed and built a wireless research test bed, referred to as the Colosseum, whereSC2 competitors could test and develop solutions to enable this new communications paradigm.A critical component of the Colosseum was its legacy radio emulators, referred to as Colosseumincumbents, that represented today’s systems. These incumbents emulated the radio frequency(RF) behavior of existing real-world radio systems, serving as RF obstacles that SC2 competitorshad to detect and work around while simultaneously administrating their own communicationsfor maximum data throughput efficiency.

The Defense Advanced Research Projects Agency (DARPA) Spectrum Collaboration Challenge(SC2) required competitors to develop shared spectrum solutions for next-generation communicationsystems. To enable competitors to test their designs and DARPA to measure and evaluatetheir utility, the Johns Hopkins University Applied Physics Laboratory (APL) designed and built awireless research test bed called the Colosseum. One of its components, the Traffic GenerationSystem, enabled on-demand generation and logging of Internet Protocol (IP) version 4 (IPv4) trafficin the Colosseum. The Traffic Generation System simulated a set of network applications runningsimultaneously on a group of peer nodes, such as a video conferencing application connectingfour participants. The Traffic Generation System provided a continuous and unpredictablestream of traffic so that competitors could be measured against a maximum expected trafficflow transmitted through their radios with no possibility of gaining an unfair advantage. IP trafficprovides good evaluation metrics because IP packets can be counted, and statistics such as datathroughput, latency, jitter, and loss can be calculated. This article discusses the software, hardware,and networking design of the Traffic Generation System.

The Johns Hopkins University Applied Physics Laboratory (APL) designed and built a wireless communicationsresearch test bed, called the Colosseum, for the Defense Advanced Research ProjectsAgency (DARPA) Spectrum Collaboration Challenge (SC2). SC2 aimed to motivate research intoautonomous wireless communication systems to uncover a new paradigm for managing theoversubscribed radio frequency (RF) spectrum. This article describes the Colosseum’s RF EmulationSystem, which mimicked real-world phenomenon such as propagation delay, Doppler shift, andpower attenuation between 128 two-channel radios, or 65,536 wireless communications channels.The RF Emulation System emulated isolated virtual environments across multiple concurrentexperiments, enabling challenge competitors to research, develop, and test next-generation artificialintelligence solutions for wireless network systems.