Seeking signs of life, NASA has been probing the moon for ice for years. What if they could do this at scale and more efficiently?
In collaboration with the Robotics Club we will be developing a computer vision algorithm using cameras and LiDAR so a lunar surface rover (or a fleet of such rovers) could search for ice.
- On the machine learning side, we are working on building software to help the rover detect and avoid obstacles, locate regions of interest, detect potential ice on the lunar surface, and map its location accurately.
- Most of the work will be done using Gazebo for simulations of the lunar surface and OpenCV in Python to detect obstacles or unstable terrain surrounding the robot.
- For detecting potential ice on the lunar surface, we’ll be training a convolutional neural network using simulated data. This will be useful for the robot’s external drill arm, which can take core samples of lunar debris.
- Additionally, we will be training a model on simulated data to improve SLAM (Simultaneous Localization and Mapping) accuracy given the constraints of movement on the lunar surface.
Minimum Viable Product: full obstacle detection software, accurate localization and mapping of the surrounding terrain, and detection of ice within lunar rocks.
Project Leads
