GalaxyRVR
Overhead and onboard vision, proximity sensing, route planning, simulation, and control for a small differential drive mobile robot.
Drift Hackathon · 9 November 2025
Problem and Context
Arena robots need to combine position, obstacle, target, and motion data that may arrive at different rates or disagree. At the Drift Hackathon, we explored overhead vision, onboard sensing, and simulation as inputs to navigation.
User and System Flow
- Calibrate the overhead camera using arena markers and a perspective transform.
- Detect targets and estimate the robot position in world coordinates.
- Generate waypoints while accounting for mapped or observed obstacles.
- Combine camera, odometry, ultrasonic, and infrared observations.
- Translate navigation output into motor commands, visualize progress, and replan as needed.
Capabilities
The navigation prototype combines camera processing, target detection, robot localization, dead reckoning during camera delays, waypoint planning, line following, obstacle avoidance, differential drive control, path visualization, and simulation. A separate image model puzzle utility sits outside the navigation loop.
Architecture
Python and OpenCV process USB webcam and ESP32-CAM inputs. Arduino firmware and a Python client handle serial communication, motors, ultrasonic distance, infrared obstacles, servo movement, and battery information. Navigation modules turn observations into paths and motor commands; simulation exercises parts of the logic without hardware.
Status and Credits
The navigation builds on Drift's hackathon starter code.