How AI is Revolutionizing Close-Combat Operations: Part I

How AI is Revolutionizing Close-Combat Operations: By Rob Albritton, Senior Director/AI Practice Lead, Ted Hallum, Senior Defense ML Engineer, and John Bellamy, Principal ML Engineer

Author: Rob Albritton, Senior Director, AI Center of Excellence

Data. We are inundated with it as individuals, so imagine the amount of data produced, ingested and processed by an organization as large as the Department of Defense (DoD). Now picture the warfighter who relies on that data but doesn’t always have access. That has been the reality. However, thanks to technology powered by artificial intelligence (AI), warfighters are increasingly in a better position to meet the mission of protecting our nation through the use of data.

Root of the problem

On any given day, the U.S. military collects massive amounts of disparate data at the tactical edge where users operate in austere environments with sparse communications connectivity and limited storage availability. According to Special Operations Command (SOCOM) leadership, special operations forces capture more than 127 terabytes of digital material each year, the equivalent of 44,000 feature-length movies or 50 NFL seasons worth of video (every team, not just one).

Clearly, data collection is growing. But the ability to manage that data has not grown alongside it. Part of the issue is the reliance of processing technologies on the cloud, especially at the tactical level. To conduct rapid analytics on tactical data in bandwidth constrained or disadvantaged environments, soldiers, Marines, and special operators cannot rely on machine learning or deep learning (ML/DL) algorithms that reside in the cloud and require massive amounts of bandwidth and compute capability.

Small tools, big AI solutions

To solve the problems of limited connectivity, storage and accessibility, learning models and geospatial analytics must be pushed as close as possible to where the mission occurs, at the tactical edge. In some cases, this is accomplished through edge computing, the capturing and processing of data close to the source. Low power computational hardware and algorithm compression tools make it possible for industry, academia, and government to provide the tactical edge with AI-enabled analytics that were once available only to those with access to big-iron computational resources.

Here are a few examples of how edge computing powered by AI is helping the military meet missions. The U.S. Army is purchasing approximately 13,000 Black Hornet nano unmanned aerial surveillance (UAS) systems, one for every infantry squad. Close-combat units deploying the Army’s Integrated Visual Augmentation System (IVAS) and Army Engineer Research and Development Center Geospatial Research Lab (ERDC GRL) developed full motion video to 3D will soon produce nearly 1.5 terabytes of photogrammetric point cloud, 3D tile, and obj model data per day. Additionally, every soldier, Marine, and special operator now carries with them Nett Warrior, TAK, or other end-user devices capable of collecting video and still imagery. This represents tremendous amounts of data being collected by individuals and small groups with handheld devices.    

Consider that just 20 years ago, ASCI Red was the world’s first 1 Teraflop supercomputer requiring 2,500 square feet of space. Today, the NVIDIA Jetson AGX Xavier outperforms ASCI Red and fits in the palm of your hand. This type of small form factor computing technology means soldiers, Marines, and special operators can push ML and geospatial analytics processing to the tactical edge, without reliance on robust network connectivity to send large datasets and ML models to cloud-based compute resources. Aided Target Recognition, Rapid Target Acquisition, full motion video (FMV) to 3D model generation, and other mission-critical, AI-enabled analytics can now be accomplished in tactical vehicles, at outposts, or anywhere else the mission would be enhanced by availability of high-performance computing.

Myriad technologies and models combine to make these systems work, revolutionizing close combat operations. High-performance edge computing processors, model containerization, and open source intelligent data federation tools make up part of the equation. Data federation, (the combining of autonomous data stores to form one large data store) is key to the success of these efforts. Federation of tactical data empowers the Intelligence Community, commanders, decision makers, and other units to discover and leverage tactical data they normally would not be able to access. This enables enhanced mission execution and decision making based on more real-time data collection and analysis.

Using tactical data and AI-driven analytics processed by small form factor high performance compute solutions is allowing strategic mission achievement in tactical environments like never before.

 Octo’s research

In collaboration with academia and the Federal Government, Octo is leveraging our innovation lab, oLabs, to explore technology methods and models that leap technology hurdles and facilitate squad-level AI-enabled geospatial analysis, more efficient methods of storage, and federated ingest, dissemination, and discovery of tactical data at the enterprise.

Collaborative research efforts are currently underway in oLabs, the University of Missouri, the Virginia Tech Discovery Analytics Center, U.S. Army Cross Functional Teams, and Joint Services. The Department of Defense and Intelligence Community can enable hyper aware close combat forces and deliver intelligent geospatial and AI-enabled processing to the tactical point of need by embedding geospatial and ML applications on low power edge compute devices.

In Part II of this series, we’ll look at how Octo’s solution, CXSearch, is contributing to the evolution of AI-powered solutions that provide actionable intelligence at the tactical edge.