This project implements a full robotics pipeline combining SLAM (Simultaneous Localization and Mapping) and motion planning for a simulated drone tasked with finding and navigating to a treasure in a forested environment. Developed as part of the Georgia Tech course CS7638: Artificial Intelligence for Robotics.
The assignment was split into two parts:
- Part A: Used GraphSLAM to estimate the drone’s position and landmark locations using noisy control inputs and range measurements.
- Part B: Designed a motion planner with steering and distance clipping, obstacle detection, and recovery behaviors to guide the drone safely to the treasure.
- GraphSLAM with pose and landmark graph updates
- Handling of action noise and sensor uncertainty
- Greedy planner with steering angle constraints and collision detection
- Obstacle avoidance and alternative path recovery strategies
- How to integrate localization and planning under uncertainty
- The tradeoffs between precision and computational simplicity
- How to implement recovery logic in constrained planners
This project was completed for academic credit and adheres to Georgia Tech’s honor code. Source code is not publicly shared to comply with course policy.