AICore Simulation MVP

This repository contains the minimum viable simulation framework (MVP) for the AICore PhD project.

The goal of this stage is not performance or realism, but to establish a running, observable, and extensible architecture for multi-UAV swarm research.

---

Project Scope (Current Stage)

Layer 0 – Simulation & Visualization

• Gazebo Classic with a minimal custom world
• Multi-UAV spawning (10 UAVs)
• /model_states publishing via gazebo_ros_state
• TF and static TF support
• RViz2 visualization with:
  • UAV markers
  • Obstacle markers
  • Grid-based cell markers

Layer 1 – Decision Logic (Minimal but Functional)

• Thompson Sampling–based cell selection
• Independent Beta(α, β) model per (UAV, grid cell)
• Reward defined by local neighborhood statistics
• Learning state visualized in RViz via MarkerArray

At this stage, learning output is visualized, not yet coupled to physical actuation.

---

What This Repository Demonstrates

• A fully executable ROS 2 system
• Clear separation between:
  • Simulation (Layer 0)
  • Decision-making (Layer 1)
• An architecture ready to be extended toward:
  • Communication-aware coordination
  • Real UAV actuation
  • Federated or distributed learning layers

---

How to Run (Quick Start)

source /opt/ros/foxy/setup.bash
source install/setup.bash
ros2 launch aicore_layer0 layer0_min.launch.py