NASA creates project plans with the hope that everything will work as it should. However, space is tricky and schedules slip, costs increase, and unexpected roadblocks pop up throughout the life of a project. But what if you could predict risk?
That was the task of participants in the Risky Space Business: NASA AI Risk Prediction Challenge. The public was asked to design project management tools that could extract past project risk information and utilize artificial intelligence/machine learning (AI/ML) to predict risks on future projects. The challenge received 27 submissions. Of those entries, six solutions have been awarded a share of the $50,000 award. The winners are:
- First Place: Christopher Milo (United States) supported by team members Zach Pryor, Benjamin Walzer, Daniel Mask, Jacob Walzer, Sean Mellott. The winning solution employed algorithms programmed in Python that leveraged open-source ML and Natural Language Processing (NLP) models to train text classifiers on historical NASA documents to recognize language patterns relating to risk.
- Second Place: Richárd Ádám Vécsey Dr. (Hungary) supported by his team member Axel Ország-Krisz Dr. This entry predicts risks in two different ways from project-related text documents: 5×5 risk severity-likelihood matrix and a list of risk categories. It contains three different neural network models: One predicts the chance of the occurrences of each unique risk category in the text, the second predicts severity and likelihood values for each risk, and the third recognizes whether the text contains any risk.
- Third Place: Thomas Ilin (United Kingdom). The proposal focused on a solution architecture that can make a difference to the way project risks are identified and predicted by NASA, across all of its directorates and projects.
- Risk Prediction Category: Alexander Poplavsky (Poland). The solution comprises several AI text-processing models and related algorithms to estimate the unknown project risk based on historical lessons learned data and target Project Plan information. The risk is estimated following the standard NASA classification: cost, schedule, technical, and programmatic affinities with green, yellow, and red levels.
- Data Extraction Category: Petra Galuscakova (Slovakia). This idea reformulates the problem of predicting the project risks to the problem of searching similar earlier NASA projects. It focuses on creating a searchable index of the materials, such as lessons learned, presentations, and proposals, used in the earlier projects.
- Data Formatting Category: Dean Koucoulas (Canada) supported by team member Snezana Kirova. The solution implemented techniques in NLP in Python to compare the information within project documents with a risk-based lexicon/dictionary. The information was then categorized according to what Condition, Departure, Asset, and Consequence best matched the risk details, drawing on classification schemes inherent in the NASA Continuous Risk Management Process.
“Wouldn’t it be great if we could be better at predicting risks, make more informed decisions to help reduce the chance that they will occur, and ultimately minimize negative impacts that challenge the success of projects?” said Amanda Cutright, chief engineer for NASA’s Game Changing Development Program at Langley Research Center in Hampton, Virginia. “NASA is now one step closer to that reality thanks to these innovative ideas! Being able to identify and manage risk is essential to effectively implementing any technology development project, therefore, we need to continue to adapt, innovate, and enable breakthroughs that expand our capabilities.”
The challenge allowed participants to utilize their unique set of experiences and skills to focus on portions of the overall solution. Sifting through unstructured project data to pull out potential risks takes new methods on machine learning and natural language processing and an overall solution is not currently readily apparent. Mitigating risks could greatly lead project teams toward mission success. The challenge solutions will help NASA expand its risk predictive capabilities and maximize the impact of its lessons learned data, streamlining program management and important decision-making.
The effort supported NASA’s Game Changing Development (GCD) Program and Digital Transformation (DT) Initiative. GCD advances space technologies that may lead to entirely new approaches for NASA’s future space missions and provide solutions to significant national needs. DT is a NASA strategic initiative hosted by the Office of the Chief Information Officer that serves as a catalyst for NASA to transform the way we work, the experience of our workforce and the agility of our workplace.
Freelancer.com, in collaboration with LMI Inc., administered the challenge. The NASA Tournament Lab, part of the Prizes, Challenges, and Crowdsourcing program in the Space Technology Mission Directorate, managed the challenge. The program supports public competitions and crowdsourcing as tools to advance NASA research and development and other mission needs.
Learn more about opportunities to participate in your space program via NASA prizes and challenges at:
https://www.nasa.gov/solve