All Projects
Academic
LaserTank AI Solver
AI solvers for the LaserTank puzzle game using A* search, MDP planning, and reinforcement learning algorithms.
AI Search A* Algorithm Reinforcement Learning MDP Python
View Source
Overview
Implemented multiple AI solving algorithms for LaserTank, a puzzle game where a tank must navigate to a flag using movement, rotation, and laser interactions with map elements — finding optimal solutions in as few moves as possible.
Algorithms Implemented
- Uniform Cost Search & A* Search: Graph-based search with custom heuristics for optimal pathfinding
- Value Iteration & Policy Iteration: MDP-based planning for stochastic environments
- Q-Learning & SARSA: Model-free reinforcement learning for environments with unknown dynamics
Problem Formulation
The game was formulated as a search problem with:
- State space: tank position, orientation, and map configuration
- Action space: move forward, turn left, turn right, shoot laser
- Uniform cost of 1 per step
- Goal: reach the flag in minimum moves while avoiding game-over conditions
Key Learnings
- Comparative analysis of search, planning, and learning approaches to the same problem
- Heuristic design and admissibility proofs for A* optimality
- Convergence properties of value iteration vs. policy iteration
- Exploration-exploitation trade-offs in Q-learning and SARSA