N241-056 TITLE: Autonomous, Mission-based Traffic Engineering

OUSD (R&E) CRITICAL TECHNOLOGY AREA(S): Integrated Network Systems-of-Systems; Trusted AI and Autonomy

OBJECTIVE: Develop autonomous, mission-based traffic prioritization mechanisms/techniques and orchestration to ensure network priorities are aligned with the Commander’s Intent – especially during contested operations when there will not be sufficient network resources to satisfy all of the operational needs.

DESCRIPTION: Tactical networks typically contain an admixture of critical, essential, and non-essential traffic. Criticality of the traffic depends on the types of missions that are currently being executed, mission phase, etc. and is, consequently, highly dynamic. Reference (1) outlines command and control (C2) constructs for the U.S. Navy and Army. These constructs have different warfare commanders, functional commanders, and coordinators – with different network applications / priorities – that must be synchronized to achieve operational objectives. The Navy seeks technical solutions for (1) identifying different types of traffic, (2) associating each traffic type with specific platforms as well as functions/missions, (3) enabling the Commander to prioritize these functions/missions, and (4) translating the Commander’s prioritization into network policies that can be implemented across the network to ensure, to the extent possible, end-to-end delivery of mission-critical traffic.

There are two main technical challenges that must be solved:

(1). Reliable mapping of Commander’s intent and mission objectives into structured data forms that can be combined with policy representations and reasoned with by machines. Supervised Natural Language Processing (NLP) training requires more exemplars than may be available given the highly dynamic nature of Commander’s intent or mission objectives. Machine representations need to precisely capture the original human meaning within the intent as well. Reference 2 provides an overview of the application of artificial intelligence in different areas of the private sector.

(2). How to combine machine representations of policies? Ontologies can be used to capture schema but live tactical feeds for situation awareness, which may be ad hoc, are also critical.

PHASE I: Develop a framework/approach to address the challenges outlined above. Prepare a report documenting the proposed framework/approach along with any preliminary results or data that help demonstrate the viability of the proposed approach. Include in the report a proposed set of benchmarks for assessing performance of the framework and a clear articulation on how the framework is viable with incomplete training data sets. The latter is important because warfare commanders are individuals who have different preferences for receiving and displaying information to support a decision.

PHASE II: Implement the proposed framework using representative data sets for different functions or missions. Demonstrate how the framework correctly interprets intent and then translates that intent into traffic engineering policies. Show how this intent is met both with dynamic network demand and changing circumstances (e.g., priorities change due to a triggering event). Prepare a report documenting the framework implemented, how it was tested, the resulting performance, and recommendations or lessons learned for future implementations.

PHASE III DUAL USE APPLICATIONS: Integration and transition into ADNS is the objective of Phase III. The commercial sector has historically relied on fixed, terrestrial networks and can either easily procure more bandwidth to alleviate congestion or add redundancy. For truly mission critical traffic, the commercial sector builds dedicated networks with dedicated resources to guarantee performance. However, the push towards 5G deployment and increasing need for real-time control systems for autonomous vehicles, automated manufacturing, smart city concepts, etc. are pushing the commercial sector to develop prioritization mechanisms and how to orchestrate their employment across the network to ensure end-to-end delivery of mission-critical traffic. A potential commercial transition option that can be explored is to integrate the algorithms developed into the zero-touch network management solutions developed for 4G/5G mobile services by ZTouch.

REFERENCES:

1. Ziegenfuss, MAJ Mark Paul. Employing the US Navy’s Composite Warfare Commander Construct to a US Army Corps to conduct Command and Control in Multi-Domain Operations, Naval War College, 18 August 2021. https://apps.dtic.mil/sti/trecms/pdf/AD1152828.pdf
2. Blagec, Kathrin; Barbosa-Silva, Adriano; Ott, Simon and Samwald, Matthias. A curated, ontology-based, largescale knowledge graph of artificial intelligence tasks and benchmarks, Nature Scientific Data, 17 June 2022. https://www.nature.com/articles/s41597-022-01435-x

KEYWORDS: Command, Control, Network, Prioritization, Mission, Automated, Dynamic, artificial intelligence/machine learning, AI/ML

---