Distributed Analytics in Tactical Networks
Funding Agency:
- Department of Defense
The aim of this Special Notice under the DEVCOM Army Research Laboratory (ARL) Broad Agency
Announcement (BAA) (W911NF‐23‐S‐0001) as amended, under Grants.gov Opportunity W911NF‐23‐S‐
0001‐SPECIALNOTICE‐Distributed Analytics in Tactical Networks, is to solicit proposals related to the ARL
BAA Topic, “Networking Structure, Dynamics, and Protocols”. The future battlefield will be increasingly
complex (dynamics, scale, heterogeneity, peer adversaries) and resource constrained (bandwidth,
storage, memory, computing power). This complex network must support multiple conflicting demands
and must support distributed analytics at the edge. The Army is interested in collaborating with an
awardee on research efforts to further develop adaptive, goal‐driven, semantically‐aware, distributed
analytics for situational understanding at the tactical edge. Proposals are expected to focus on one of the
research goals and to identify possible linkages to the other research goals.
Unless stated here, all terms and conditions in the ARL BAA W911NF‐23‐S‐0001 as currently amended are
in effect for this special notice.
Research Goals:
1. Distributed in‐network reasoning with adaptive analytics at the edge. Novel frameworks and
algorithms that enable analytics at the edge under limited resource availability (machine learning
(ML) services, compute, memory, storage, and comms) through intelligent information
management techniques. Suggested research topics for consideration are:
a. Adaptive networked information management, including fusion, reshaping, and
deduplication/compression for resource‐adaptive analytics at the point of need.
b. Semantic information and knowledge representation of applications and analytics
performance of multimodal applications in distributed and resource constrained compute
environments.
c. Active system querying and resource reallocation for intelligent information gathering and
environment monitoring.
2. Agile AI at the Edge. Analytics at the edge must adapt not only to varying requirements (such as
accuracy and latency) but also to resource availability (ML services, compute, memory, storage,
and comms) and to the adversarial environment. Suggested research topics for consideration
are:
a. Movable and adaptable AI algorithms at the point of need with resource‐constrained
compute and bandwidth‐limited networks.
b. AI/ML model training and fine‐tuning/adaptation at the edge, including joint approaches to
move data and models.
c. AI/ML models that automatically adapt to adversarial attacks, providing accuracy‐resilience
tradeoffs.
3. Cross‐layer multimodal networking and communications. Future tactical networks will comprise
a multitude of diverse communication technologies and interfaces spanning the electromagnetic
spectrum. Novel approaches are required to control extremely heterogeneous networks to
ensure resilience in highly dynamic tactical scenarios, adapting to meet survivability,
throughput, and latency challenges. Suggested research topics for consideration are:
a. Multi‐flow multi‐modal network optimization with multiple and time‐varying metrics:
distributed optimization and quantification of accuracy‐overhead tradeoffs.
b. Multi‐agent network control (e.g., monitoring, traffic engineering, routing, load balancing,
etc.) strategies that take into account heterogeneity of communication modalities, diverse
network flow objectives and priorities.
4. Network Monitoring for sensing temporal & spatial network dominance / identification of
windows of opportunity (WoO): WoO will be dynamic and possibly short‐lived, requiring fast
accurate inferencing, which will entail efficient collection of metrics/attributes from different
domains (“network state”) that evolve at different time and spatial scales, and are stochastic.
Suggested research topics for consideration are:
a. Strategies to gather relevant information to infer network state and identify windows of
network overmatch or vulnerability (connectivity, coverage, or capacity). Quantification of
underlying accuracy‐timeliness‐overhead tradeoffs.
b. Network state prediction including duration/location of windows of
overmatch/vulnerability, while accounting for incomplete information and uncertainty
caused by environment and system dynamics.
Proposals are expected to focus on one of the four research goals listed above and to identify possible
linkages to the other research goals.
The expected award range is from $500,000 to $700,000 per award per year for 2 years.
20 March 2024
Joshua Wells
Contract Specialist Phone 9195410817
Grant Specialist Email