The US Army Research Laboratory (ARL) is counting on supercomputing and large-scale analytics to provide the competitive edge it needs to maintain its position as the nation’s premier laboratory for land forces. As laid out in the recently-released Technical Implementation Plan for 2015-2019, the ARL sees advanced computing as fundamental to its mission to “discover, innovate, and transition science and technology to ensure dominant strategic land power.”

The laboratory’s first comprehensive plan to guide its portfolio into 2030 and beyond includes seven major thrusts: computational sciences; materials research; sciences-for-maneuver; information sciences; sciences-for-lethality and protection; human sciences; and assessment and analysis. Among these, there are 24 program initiatives outlined in the implementation plan that target the realization of vastly improved Army capabilities both for the near-term (FY15-FY19), mid-term (FY20-FY25) and longterm (FY26-FY30) trajectory.

The ARL report speaks to specific challenges, including the oft-refrain of government-funded labs: the need to make scarce resources go farther.

“Emerging trends suggest that the future Army’s operational environment will likely be dominated by decreasing domestic budgets and reduced force structure; increased velocity and momentum of human interaction and events; potential for adversarial capability overmatch; proliferation of weapons of mass destruction; spread of advanced cyberspace and counter-space capabilities among our adversaries; and increased likelihood of operations among populations, in cities, and in complex terrain.

“Within the context of this highly non-linear and complex operational environment, the Army – America’s principal land force – must shape the security environment; set the theater of operations; efficiently project national power; effectively execute combined arms maneuver in the air, land, maritime, space, and cyberspace domains; initiate and maintain wide area security; conduct cyberspace operations in the land domain; and integrate special operations across the Army’s mission set. These core competencies, the Army’s strengths and essential contributions to the Joint Force of the deep future, will strongly rely on S&T developments.”

The primary mode of attacking this diverse challenge set is through the deployment of computational science and the applications of advanced computing technologies. The ARL’s computational sciences campaign sets out a three-fold mission to (1) harness the potential of computational sciences and emerging high-performance computers (HPC) to maintain the superiority of Army material systems through predictive modeling and simulation technologies; (2) facilitate information dominance, distributed maneuver operations, and human sciences through computational data intensive sciences; and (3) significantly increase and tailor advanced computing architectures and computing sciences technologies on the forefront to enable land power dominance.

Tactical High Performance Computing, the first element of this plan, comes with a bold goal: provide 100 petaflops of computing power to improve mission effectiveness and mitigate risk in hostile environments.

The ARL aims to overcome the power and performance limitations associated with standing up a 100-petaflops supercomputer by using a distributed computing approach. The system will be an aggregation of deployed devices and mobile HPC platforms operating at the tactical edge, so-called “tactical cloudlets.” In addition to enabling real-time processing for the benefit of mounted soldiers, the amassed computing power will provide intelligence analysts with real-time data analytics.

Advanced computing research to support this real-time distributed processing system falls into four categories:

• Efficient use of emerging architectures to facilitate this level of computing capacity straddling both fixed and deployed devices.
• Provisioning systems within a distributed computing architecture in a way that limits network hops.
• Dynamic binary translation to limit software re-writes and promote optimization in a runtime environment to achieve maximum performance.
• Power- and architecture-aware computing for enhanced intelligence of provisioning systems to increase awareness of computing capacity and mission appropriateness

“The critical, mobile ad hoc networks that will form the connections in tactical cloudlets to the large-scale databases and complex applications that will be performed by these resources make this research uniquely military and Army in nature,” the ARL plan states. “Numerous applications are envisioned for this system in the future and include artificial intelligence aids for decision making, processing large-scale datasets (text, video), and navigation systems for autonomous vehicles (HPC-enabled autonomous vehicles providing on-demand processing).”

The roadmap for meeting these goals emphasizes several emerging computing paradigms, including quantum networks, bio-computing, heterogeneous, quantum annealing, and neuro-synaptic computing architectures. For example, a far-term goal (FY26-30) is: “Incorporation of novel processing paradigms and hardware (Quantum, neuro-synaptic, and bio-computing) as part of a broader distributed computing solution for Soldiers.” In the near-term (FY15-19), the ARL will use neuro-synaptic and quantum annealing emulation architectures and small scale systems to support algorithm research.

The second core program thrust is very large-scale data analytics. The ARL plans to mine its vast data stores, generated from battlefield networks, sensors, experiments, observations, and numerical simulations, to provide information to commanders across a range of military operations. Core program goals include information supremacy, vastly improved situational awareness, as well as accelerated soldier training trough live and virtual data analytics. As with the HPC program, this effort relies on exploiting emerging and next generation hierarchical computing architectures. The ARL sees extraordinary potential for scientific advance inherent in large-scale complex data, but these benefits hinge on the development of scalable computational methods and massively parallel hierarchical computing architectures.

Technical areas for this program include scalable algorithms, scalable model order reduction, uncertainty quantification methods, real-time experimental data processing, and methods for live-virtual simulations. Cognition and quantum-based computational approaches are listed as far-term goals.

To support its HPC and big data efforts, the ARL is putting together small interdisciplinary teams comprised of computer scientists, computer engineers, mathematicians and other experts to ensure that skill sets are closely aligned with technical goals.