CG-XMARL v4.1 Technical Specification
Abstract
CG-XMARL v4.1 is a military-grade multi-agent reinforcement learning system for autonomous UAV formation control integrated with DCS World 2.9. The system implements a 3-UAV wedge formation pursuit control architecture with advanced consensus-based virtual structure (CBVS) guidance, hierarchical finite state machines, and a planned HGAT-MAPPO-WM (Hierarchical Graph Attention Transformer with Multi-Agent Proximal Policy Optimization and World Model) neural architecture.
Fields of Study & Subjects
Reinforcement Learning in Robotics
Distributed Control Multi-Agent Systems
Aerospace Engineering and Control Systems
Publication details
- Publication date
- 5 February 2026
- Document type
- Preprint