Value Iteration Policies for a Warehouse Robot (MDP)

Instead of planning one path, the robot gets a complete policy: the best action from every cell in the warehouse, so it can wake up anywhere and still retrieve and deliver the box. What I implemented: • Value iteration over a warehouse with uneven floor costs (0–95 per cell) and 8-directional movement • Separate policies for reaching the box and for delivering it • Handling stochastic motion (Part B): the robot may drift diagonally or sideways from its intended direction, and the policy accounts for those probabilities Course project for CS 7638 Robotics: AI Techniques (Georgia Tech OMSCS). The simulator was provided by the course; the policy code is mine. Code available on request.