Maze Solver and Autonomous Agent
A* search and a PyTorch CNN solving grid-based mazes, comparing classical planning against a learned policy.
- Python
- PyTorch
- A* Search
- CNN
An AI agent that navigates grid-based mazes two ways: A* search with heuristic path planning, and a convolutional neural network trained in PyTorch to navigate the same environments. Building both made the trade-off concrete — where state-space search and heuristic design win, and where a learned policy generalises better. The search half is the same shortest-path problem a routing protocol solves, minus the part where the graph changes underneath you.