Machine Learning: Improving Cloud Prediction Models Using ML: Fixing the Safety Culture Fixes at NASA

Author:
Nolan, Alice, School of Engineering and Applied Science, University of Virginia
Advisors:
Vrugtman, Rosanne, EN-Comp Science Dept, University of Virginia
Morrison, Briana, EN-Comp Science Dept, University of Virginia
Foley, Rider, EN-Engineering and Society, University of Virginia
Forelle, MC, EN-Engineering and Society, University of Virginia
Abstract:

The ability to send humans to Mars is impossible with the current state of technology (Redfern, 2021). Simply attempting to land on Mars has created a graveyard of probes and rovers due to the thinness of Mars’s atmosphere and the forces on the spacecraft as a result (Katz, 2021). A mission to Mars has physical limitations of mass, weight, and volume. With the demands of this odyssey, current communication systems are insufficient. Radio frequency systems are slow to transmit signals, take up precious payload space, and use too much of the spacecraft’s energy supply (Edwards, Israel, Wilson, Moores, & Fletcher, 2012). Being able to communicate quickly with astronauts is vital when traveling into deep space where disaster can strike at any moment. At the same time, the more payload space and resources taken up by communication systems, the less space there is for other vital supplies like food, water, and building materials for shelter. Furthermore, any inefficient drains on the power supply will limit the distance a spacecraft can travel (Williams, 2021).
Ensuring that astronauts can reach their destination on Mars safely is a top priority (Tran, 2019). Astronauts cannot be burdened with inadequate technology or a broken safety culture that could jeopardize their well-being. Astronaut safety can not only be disrupted by malfunctions of spacecraft technology but also by the failings of a safety culture that values schedules and budgets over safety and lives. The Columbia space shuttle completed its 16-day adventure through space and the seven astronauts inside the vessel were killed as it re-entered the Earth’s atmosphere. An investigation of the disaster found structural issues with the safety culture at NASA played a large role in the explosion (Smith, 2003). Astronauts must be aware and informed of the accepted risk and the safety procedures for their own personal well-being. Since ensuring technology can physically transport astronauts to Mars is foundational to future missions, improving the communication systems between astronauts and ground control stations and investigating the safety culture at NASA is vital to protecting the safety of astronauts on board.
The current state of laser optical communications requires greater certainty about cloud formation and prediction in order to discover optimal ground station placements. The atmospheric effects team comprised of meteorological experts at Northrop Grumman had decided to tackle this technical challenge by using machine learning to increase the availability of the optical network. The team addressed the problem using a convolutional neural net and image recognition to create cloud mask predictions of the sky dome. They used python-based libraries like TensorFlow and Keras to create the neural net, as well as data science libraries like Numpy and Pandas to process the data. The research conducted during the internship resulted in an improved cloud mask model that over-predicted clouds less often. The research could be applied to optical communication systems, solar panel placement, and missile launch sites. Future work on the project could include gathering more/better quality data, testing more predictors, and using improved thresholding on the weather-measuring device.
The research to investigate the safety culture at NASA centered around a case study of the Columbia explosion. By looking at reports and analyses of the tragedy, key structural issues like schedule pressures were compiled and discussed. Responses and proposed solutions from sociologists and ethics researchers were analyzed through the lens of technological determinism. The research and analysis found that the proposed solutions did attempt to address the structural faults with NASA’s safety culture. However, many of the previous recommendations did not consider the normalized deviance of accepted risk or the harms and fails associated with technological determinism and malfunction. The findings of the research recommended that the safety culture at NASA conduct periodic investigations, case studies, and recommendations for changes and improvements with the safety procedures. Each safety concern should be encouraged and investigated with recommendations for greater safety rather than dismissal. Normalized deviance and accepted risk should be more closely compiled and logged to ensure a more informed level of risk, even if not all risk may be known.
By working on both projects simultaneously, I was able to see how the ethical frameworks researched and applied in the STS research paper could be applied to the technical research paper. I found that assuming there to be a technical solution to the deep-space communication problem was present in my technical research. I learned to consider systems, ethics, and non-technical solutions rather than purely relying on technical solutions and code. When I was studying the harms of technical technologies on society, the reflection made me consider what the effects of my machine learning models could have on social groups I had not considered before. I gained a new understanding of engineering and the ethical principles and dilemmas baked into the field as well as how to better deal with these dilemmas.

References
Edwards, B., Israel, D., Wilson, K., Moores, J., & Fletcher, A. (2012). Overview of the Laser Communications Relay Demonstration Project. SpaceOps. doi: 10.2514/6.2012-1261897
Katz, O. (2021, May 21). Landing on Mars is difficult, often fails, and will never be risk-free. Phys. Retrieved from https://phys.org/news/2021-05-mars-difficult-risk-free.html
Redfern, M. (2021). Will we ever set foot on Mars?. BBC Earth. Retrieved from https://www.bbcearth.com/news/will-we-ever-set-foot-on-mars
Smith, M. S. (2003). NASA's Space Shuttle Columbia: Synopsis of the Report of the Columbia Accident Investigation Board. CRS Report for Congress. Retrieved from https://ntrs.nasa.gov/api/citations/20040040191/downloads/20040040191.pdf
Tran, L. (2019, August 7). How NASA Will Protect Astronauts From Space Radiation at the Moon. NASA. Retrieved from https://www.nasa.gov/feature/goddard/2019/how-nasa-protects-astronauts-from-space-radiation-at-moon-mars-solar-cosmic-rays
Williams, M. (2021, February 5). Every challenge astronauts will face on a flight to Mars. Phys. Retrieved from https://phys.org/news/2021-02-astronauts-flight-mars.html

Degree:
BS (Bachelor of Science)
Keywords:
Machine Learning, ML, Astronaut, Safety, Columbia, Testing, Technological Determinism, Laser, Optical, Communications, NASA, Mars, Mission
Notes:

School of Engineering and Applied Science
Bachelor of Science in Computer Science
Technical Advisor: Briana Morrison
STS Advisor: MC Forelle

Language:
English
Issued Date:
2023/05/12