NASA Fellowship Award Selection

Congratulations to Timothy Darrah for his recently awarded 3 year NASA Fellowship (at Ames).

Dr. Gautam Biswas submitted the proposal on behalf of Tim, making his acceptance 1 of 5 major university selections out of 279 applications across the nation.

A little bit about Tim's proposal:

Automated reasoning in the presence of uncertainty is a fundamental task performed during space mission operations. Reasoning tasks include choosing science observations to perform, path and motion planning in the presence of hazards, and contingency planning in the presence of faults. Constraints on computational resources such as processor speed, utilization time, and memory allocation substantially increase the complexity of choosing the appropriate reasoning algorithm at design time. Currently, mission designers and their teams spend thousands of hours manually evaluating different algorithms for a given space mission, which are error prone and not guaranteed. “Designers of autonomous systems do not have the right information at their disposal to evaluate the tradeoffs between different approaches to autonomous systems.” Complex decision-making algorithms typically do not end up deployed on-board due to software complexity, resource consumption, or lack of solution quality guarantee given the resource limitations of the on-board processors. On NASA’s recently chartered High-Performance Space Computing initiative, A researcher at Airforce Research Laboratory was quoted as saying “since we do not yet have an actual HPSC for tests, we can make some educated guesses as to what its performance may be like,” meaning space-based computing will remain resource-limited for some time to come.

Under the proposed research, a framework for evaluating different approaches will be developed, and the results of evaluating a select number of algorithms will be presented as hardware/software trade metrics to space mission designers. The trades may be performed for a number of automated reasoning tasks; the results of the research will also provide a template for evaluating other automated reasoning algorithms needed by future space missions. The results of this research will provide the information mission managers need to evaluate the tradeoffs between solution quality, software complexity, and resource utilization, allowing them to deploy better decision-making algorithms, and increase autonomy for more complex missions. The problem space we consider in this work is contingency planning under faulty conditions, whereby the system develops a fault that is detected, isolated, and now must recover or continue operating autonomously.