Search & Reinforcement Learning Experiments

What needed to be solved

Compare an algorithm that plans with a known heuristic against an agent that improves through repeated reward-driven interaction.

How the project addressed it

Implemented A* with explicit frontier and cost handling, then built a Q-learning loop around states, actions, rewards, and policy updates.

My contribution

Implemented the algorithms, designed test scenarios, adjusted parameters, debugged the programs, and interpreted the results.

• Implemented A* pathfinding using path cost and an admissible heuristic.
• Updated a Q-table using rewards and estimates of future value.
• Used compact environments that make algorithm behavior easy to inspect.

What the work demonstrated

Demonstrated the practical difference between deterministic search and learned decision policies in small, inspectable environments.