Exploration and Artifact Detection

ETH RSS Competition

Collaborator(s): Oliver Hausdörfer, André Gomes, Maria Lopes, Stefano Bassino, Erya Guo

(Slides)

Overview

This project was developed as part of the ETH Robotics Summer School competition, where our team utilized the SuperMegaBot to autonomously explore an unknown environment and detect artifacts. Our approach combined state-of-the-art SLAM, exploration algorithms, and deep learning-based object detection to navigate and identify key objects in the scene.

Sensors on the SuperMegaBot

Methodology

Our approach involved tuning and integrating the following software components:

  • LiDAR SLAM & SE Closed-Loop → For real-time localization and mapping
  • TARE Planner → To efficiently explore the environment
  • FALCO Planner → As a local motion planner
  • YOLOv5 → For real-time object detection
  • DBSCAN → For post-processing the position of the detected objects
Graphics showing the methodology of how the SLAM and exploration planner works

Results

Below is a visualization of the point cloud map generated by the robot, along with detected objects such as an umbrella, clock, and stop sign.

Pointcloud of the explored environment

Thanks to our optimized strategy, we secured first place in the competition 🎉.