01 / Context
Purpose
Every move leaves an occupied trail and reduces the space available to both players. The agent explores possible continuations to choose a move under uncertainty about its opponent’s next decisions.
02 / Implementation
Implementation
Represent the board
Track occupied cells and player positions, check collisions and copy game states for simulation. Candidate moves turn left, continue straight or turn right.
Search possible continuations
Use upper-confidence-style selection, expand candidate directions, run randomized rollouts and backpropagate visit/win counts. Simulations use a configured wall-clock budget.
Connect the agent
Receive game-state messages through a Qt/UDP interface, update the internal board and send a directional movement command to the game platform.
03 / Evidence & interpretation
Evidence
The repository contains board simulation, an MCTS-style search agent, and a Qt/UDP game interface. It documents an implemented algorithmic exploration; tournament results, measured win rates, and standardized agent comparisons are not provided.
Limitations
- The implementation is a learning project using MCTS-style search. It has no measured claim of optimality, benchmark superiority, or latency guarantees.
- The bundled Advanced Programming course game platform is separate upstream material. My contribution is the search agent and its interface.
04 / Contribution
Contribution & attribution
Independent project by Seyyed Mohammad Hamidi: MCTS-style search agent and Qt/UDP game interface. The bundled course game platform is separate upstream material.
05 / Code & methods
Source & documentation
View TRON MCTS on GitHub (opens in a new tab)Source reviewed at commit c72cd2811695, 28 October 2026. Documentation links are pinned to this snapshot.