Artificial Intelligence for Multi Agent Systems

Multi-agent systems can solve problems that are difficult or impossible for an individual agent to tackle, and the coordination and collaboration between autonomous agents is the key to bring such systems to their full potential. In this project, a set of autonomous vehicles are coordinated to collaborate on a variety of problems, in a simulated UNITY environment with obstacles to avoid.

---

---

Starcraft AI bot

• Behaviour trees and back chaining for controlling a starcraft AI system.
• Installation: https://github.com/davechurchill/ualbertabot/wiki/Installation-Instructions https://www.youtube.com/watch?v=Bw2IDR5gt7c

Scouting

4_scouting.mp4

Starcraft Full Game

CHERnh76v4.mp4

---

Drone Soccer

3_goal.mp4

---

Coordination

• We use Minimum Spanning Tree (MST) to perform Vacuum Cleaner Planning
• Minimum Set Cover to solve Indoor UGV search.
• V-shape formation for formation sweep
• Dijkstra algorithm for planning shooter coordination of three vehicles.
• Collision Avoidance

2_formation.mp4

3_semaphore.mp4

---

Authors:

• Path Planning (Marco Schouten)
• Vehicle Routing Problem (Marco Schouten, Xuecong Liu)
• Indoor UGV search (Marco Schouten, Xuecong Liu)
• Collision Avoidance (Marco Schouten, Justin Salér)
• Drone Soccer (Marco Schouten, Justin Salér)
• Starcraft (Marco Schouten, Justin Salér)