Dusty Challenge

Craters created by asteroids and smaller meteorites struck the Moon. When an impactor strikes the solid surface of a planet, a shock wave spreads out from the site of the impact. The shock wave fractures the rock and excavates a large cavity. The impact sprays material out in all directions. The impactor is shattered into small pieces and may melt or vaporize. The more strikes that had happened in the area, the more dust that can be accumulates over the time. Because the weather on the moon is static, we use crater density to predict the dust accumulation on moon.

Project Description

Phase I

Fortress Guard is a multi-layer AI project that detects and reduces the lunar dust from accumulating on the astronaut and the spacecraft.

Phase II

Dust melting base carpet using the robot to clear the landing area before astronaut are exposed to the dust environment using microwave

Phase III

Build magnet fence to repel dust away from the base area.

Development Process

We start by using model training on computer to match craters on the moon, then test the model accuracy on 200 other image: manually labeling 200 crater images and implementing 23 layered Convolutional Neural Networks and we used 9 convolutional layers for the purpose of activation and pooling which recurred after every 2 convolutional networks.

Total number of parameters in the model are 642,935. We executed 100 epochs on it with each epoch taking 180 seconds on average to run on GPU enabled Tensorflow back-end. The accuracy started with 0.3058 accuracy and jumped to 0.3866 in just 3rd epoch. Within the 10th epoch the accuracy jumps to 0.5193. Overtime then by the 100th epoch accuracy had only progressed to 0.5751.

The model detects the craters in the testing set on the moon. This model can also be implemented in the live stream, so astronauts can use this to detect craters in the real time. After the model is implemented, we marked the areas in different colors with the varying levels of lunar dust. We used Keyhole Markup Language to visualize the data over Google Moon by using the open sourced API.

Project Impact

Our Fortress Guard offers an unprecedented solution to predict the danger zones and map out the safer route for astronauts on the moon using crater's location. We are providing an optimized solution with intelligent AI system on security and safety measures.This could be a strong foundation for future human interspace migration project.

Github link https://github.com/Grasin98/Model_Function_23/blob/master/model_function.py

Project Movie: