Terrain and mobility
DBG targets contested, mobile, terrain-constrained missions where line-of-sight changes, blocked paths, and shifting topology can break assumptions quickly.
← Back to DBG home
Project Microsite
Dynamic Belief Games
Training predictive intelligent networking agents for contested mobile ad hoc networks.
DBG trains Predictive Intelligent Networking (PIN) agents to help mobile ad hoc networks adapt proactively to mobility, terrain, and interference. It combines adversarial scenario generation, a 3D digital twin, and RF-grounded validation to test and improve decisions before deployment.
3D digital twin for training and testing
RF-grounded calibration from physical measurements
Mission-constrained route and uptime evaluation
The Challenge
Why current mobile networks fail under real mission conditions
DBG is motivated by mobile, contested, terrain-constrained missions where reactive heuristics fail to keep pace with rapid link changes and adversarial pressure.
Adversarial pressure
The networking problem is shaped by jamming, spoofing, spectral contention, and uncertain operational state rather than a stable, fully observed environment.
Operational constraints
The system must respect relay budgets, limited power, mission routing plans, and the practical requirement to adapt without changing waveform firmware.
How DBG Works
Separate the framework, the agents, and the platform
Dynamic Belief Games is the training and decision framework. PIN agents are the learned networking agents. DBG Gym is the digital twin used to train, test, and compare policies under controlled but realistic variation.
Sense
Agents form uncertainty-aware beliefs from heterogeneous observations about terrain, structure, mobility, spectrum, cyber conditions, and mission-relevant objects.
Train
Dynamic Belief Games generates adversarial training scenarios inside a controlled digital twin so agents can learn under realistic, high-variance conditions.
Adapt
PIN agents learn policies for topology, routing, queueing, prioritization, and decision support while explicitly managing downside risk.
DBG Gym
A digital twin for training, testing, and visualizing network behavior
DBG Gym creates realistic terrain, materials, mobility, and traffic conditions while exposing the system through multiple views and scenario controls that support repeatable experimentation.
- Realistic terrain, materials, mobility, and traffic conditions
- Control Tower, observer, and soldier-style visual modes
- Scenario editing, repeatable experiments, and scalable training runs
Validation
Ground the digital twin in real-world RF behavior
The project is not positioned as simulation only. The validation layer is intended to compare digital-twin assumptions against field measurements and testbed-informed radio behavior.
- Virtual scenes are anchored to real radio behavior rather than treated as purely synthetic environments.
- The validation layer compares digital-twin assumptions against physical measurements and testbed-informed calibration.
- This makes DBG more credible as a training and testing framework for deployment-oriented networking decisions.
Demonstrated Capability
Demo 1 — Platoon Coverage vs Uptime
DBG Gym's first demonstration compares two platoon routes through a single urban mission envelope. Both routes share the same start and end sectors, the same mission legs, and the same formation-spacing rules; only the route geometry differs. Coverage, outage, and path loss are measured per node and per route.
Setup
- Routes follow ATP 3-21.8 planning, motivated by NG-NRMM cost maps.
- Both paths share start and end sectors and mission legs, with no node pair closer than 50 ft.
- Corridors follow building geometry; open-field zones outside the cover band are excluded.
Outcome
Coordinated routes — synced between buildings — sustained higher uptime than staggered, dispersed routes. Staggered routes showed a large coverage dip around a central obstruction; coordinated routes avoided it. DBG predicts and suggests such routes to maximize uptime under the same mission envelope.
Resources
Resources and contacts
The project archive holds the earlier project-page text while the microsite is being developed. For research opportunities or technical conversations, reach the project leads directly.
Project archive
Reference the earlier project-page text while the new microsite is being developed.
Open archiveRyan Farell
Lead research scientist and main technical contact for current DBG work.
Email RyanChandrajit Bajaj
Principal investigator for the project and broader research direction.
Email ChandrajitTeam and Contact
Leadership, contributors, and project continuity
Leadership
- Chandrajit Bajaj, Principal Investigator
- Ryan Farell, Co-Principal Investigator
Contributors
- Andrew Farell
- Logan Kronforst
- Brian Kim
- Callihan Bertley
- Luke McLennan
Administration
- Catherine Andersson, Administrator
Funding and opportunities
- Funded by Army – AFC UTDD (C5ISR); DOD Award W911NF-24-C-0006.
- Phase II runs 06 Apr 2025 through 08 Jan 2027.
- For research opportunities, contact Ryan Farell or Chandrajit Bajaj directly.
Internal Workstreams
Convenience-gated collaborator pages for engineering progress and visual systems
These internal pages are intended for active collaborators reviewing implementation, visualization, validation, and current engineering progress. They are hidden behind a lightweight password prompt for convenience only, not strong access control.
Locked internal page
Systems & Protocols
Radio simulation, PIN control overlays, SDR and AR integration, cyber-autonomy baselines, and cross-repo engineering work.
Enter protected pageLocked internal page
DBG Gym & Visualization
3D simulation, digital-twin views, terrain/material controls, and interactive scenario visualization.
Enter protected page