Aerospace doctoral student wins Molly K. Macauley Award

Published: Nov 30, 2023 3:00 PM

By Dustin Duncan

Kanak Parmar envisions spacecraft autonomously navigating themselves using artificial intelligence (AI).

Parmar, a doctoral student in aerospace engineering in Auburn University’s Samuel Ginn College of Engineering, shared her vision during a presentation at the American Astronautical Society (AAS) von Braun Space Exploration Symposium in Huntsville on Oct. 27 after receiving the Molly K. Macauley Award for the science/engineering category.

The award aims to support and recognize future leaders in the space industry by awarding grants to exceptional collegiate students. Kanak was awarded $2,500 for her research, which she can use to attend any future conference related to her research.

Parmar’s dissertation, “Towards Autonomous Spacecraft Optical Navigation Using Strategic ML-Driven Methodologies,” seeks to strategically program AI models to further enable onboard autonomy for body-centric spacecraft navigation without constantly contacting ground control for guidance, navigation and control solutions.

If her work is successful, it could reduce reliance on the Deep Space Network (DSN), which faces an increasing workload while NASA is dealing with a reduced budget of $50 million for the DSN.

Parmar said one barrier to space is affordability. Typically, large agencies have the funds to build, develop and launch spacecrafts. Now, small satellites can be assembled for a few thousand dollars.

Small satellites, or smallsats as they are commonly referred to, are low-mass and size satellites that require smaller and cheaper launch vehicles. However, though assembly costs may be lower, smallsats still utilize the DSN for ground communication for mission operations and navigation solutions, which dramatically increases mission costs. Kanak said that reducing reliance on the DSN will lower the cost of entry for smallsats.

Parmar said every spacecraft must communicate to earth-based ground control, and most missions use the DSN, which is NASA’s international array of giant radio antennas that supports space missions. Paying for a slot on the DSN is a significant expense.

Due to the increasing use of smallsats, more operators are using the DSN than ever.

“Back when few could afford to go to space, you could use the DSN regularly,” Parmar said. “Now, it's become so overstrained that people are fighting over a couple minutes.”

When a time slot is secured, DSN antennas align with a spacecraft and transmit a signal. The spacecraft receives it and sends a return signal, providing ground control with a noisy state estimate. From there, operators decide to change course or continue their current path. Each time slot on the DSN can cost thousands of dollars.

The solution could be rooted in AI, specifically for spacecraft navigation, Parmar said. However, she said the higher levels of computational power, the better the navigation. Large spacecraft from NASA can afford to have ray tracing engines on board to digitally render surface map products and compare them to real-time images from the camera suite. The renderings can determine a location estimate for an onboard state estimation filter.

While most smallsats cannot afford ray tracing technology at the same level, AI may be able to close the playing field. Not every spacecraft has the computational power, but they do have cameras that can take an image in space to determine where they are using star navigation.

Parmar's goal is to program AI to triangulate the spacecraft's relative position using an innate and strategic knowledge representation that can be generally applicable to various orbital regimes.

Parmar likened it to blindfolding someone and taking them to a place in the United States. When they take the blindfold off and see the Statue of Liberty, they know Staten Island is nearby.

“Planets have very good features like that, so if you see something distinctive, you automatically know your general vicinity,” she said. “Images contain so much information. If we can train AI to remember the right features in images, it can automatically understand that it’s seen the information before and can provide enough information for a navigation estimate.”

Another solution is understanding how AI behaves and why it provides inaccurate information.

“AI is still not well explained, especially for flight operations. Because of this, many missions aren’t going to trust it because they don’t know how it will behave in a critical situation,” Parmar said. “One of the things I really want to focus on is explaining AI. I want to see when AI fails and figure out why it failed. That’s something not many people are doing.”

She said if her project works to an acceptable degree, she can push for a technology demonstration in space. That would mean sending a model into space on a mission. If successful, it could reduce the reliance of spacecraft on ground control, possibly allowing them to check in once every two weeks rather than almost daily, thus alleviating some of the pressure on the DSN.

The award from the AAS is more motivation to find answers for the astronautical industry that doesn’t involve building more antennas, which would be costly and time-consuming.

“This award signifies that this work is very industry-applicable and can have far-reaching capabilities,” she said. “Even if the algorithm doesn’t work perfectly, someone can take it and run, developing it into something better.”


Media Contact: Dustin Duncan,,
woman stands outside in front of trees

Kanak Parmar, a doctoral student in aerospace engineering, won the Molly K. Macauley Award for the science/engineering category from the American Astronautical Society.

Recent Headlines