AE graduate student Holly Dinkel awarded out-of-this-world opportunity
When Holly Dinkel was a fifth-year senior at the University of Missouri, balancing a challenging double major in two very different fields, she knew she had to decide on a career path. Would she continue toward a career in chemical engineering or clarinet performance? Dinkel chose neither.
“I knew I wanted to go to graduate school and decided music wasn't it but getting a Ph.D. in chemical engineering wasn't really it either. One of my professors offered a connection for work on the Space Launch System rocket. It is easy to get excited about space, so I said, ‘yeah, let’s start with that.’”
After working for a year at the NASA Marshall Space Flight Center in Huntsville, AL, she went to Stanford University for a master’s degree in aeronautics and astronautics.
“Stanford is where I explored a lot and stepped out of my comfort zone,” she said. “I assisted with the development of an undergraduate course in Positioning, Navigation and Timing; became involved with a few satellite projects; researched motion planning for robotic manipulation; and interned at NASA Ames Research Center on a distributed spacecraft PNT project. I had planned to intern in India at an industrial manipulation startup in the summer of 2020, but this fell through because of the pandemic. Then I was offered an internship at Ames researching deformable object manipulation for Astrobee. This was a dream opportunity that has since taken hold as my Ph.D. research.”
Dinkel received a NASA Space Technology Graduate Research Opportunities award and began a Ph.D. program in aerospace engineering in the fall of 2020 at the University of Illinois Urbana-Champaign working with Professor Tim Bretl. The NASA award includes funding for four years of her graduate studies. It also includes funding for four visiting technologist experiences which she plans to complete in collaboration with Brian Coltin and Trey Smith in the Intelligent Robotics Group at Ames, continuing her deformable object perception and manipulation research.
The Astrobee robots fly freely through the microgravity environment of the International Space Station using electric fans. Embedded sensors and a perching arm enable the robots to perform a variety of tasks within the space station environment like self-localization or construction of semantic maps. After performing these activities, the robots return to a dock on the wall of the ISS which serves as a recharging and communication station.
“My goal is to develop autonomous caretaking capabilities for Astrobee. I study deformable object perception, specifically perception of cabling because cables are ubiquitous in human-inhabited environments. An example of a deformable object perception challenge in a caretaking task is using Astrobee to detect that a cable has become unplugged. As a free-flying inspection platform, Astrobee could be used to identify this anomaly and a manipulation robot could then be used to restore the wired network to its nominal configuration according to a wiring schematic.”
To accomplish these manipulation tasks, Astrobee must distinguish individual cables from each other in complex scenes where cables may cross each other, be occluded by other objects in the environment, or may be knotted. With over eight miles of cables on the ISS, it is a big task for Astrobee to perceive cabling.
Because astronaut time is very expensive, Astrobee could also be used to relieve astronauts of chores such as unpacking cargo.
“That’s how I began studying deformable object manipulation,” Dinkel said. “Manipulating deformable objects like cargo bags is difficult because their shapes and mass distributions are not fixed. To manipulate my phone, I can command a gripper to grasp the object at desired contact points. I can be confident that the grasp is stable, the object will not slip, and I can access information about the object’s position and orientation within the grasp. But with deformable objects like cargo bags, desired contact points may not be accessible under the object’s current configuration and contact forces which change the shape of the object complicate downstream tasks like unpacking.”
Dinkel said Astrobee has several on-board camera sensors, so the perception challenge seemed like a natural starting point. “Given a lab or ISS image stream, I am working on detecting, localizing, and segmenting instances of cables from images. The object instance segmentation masks can then be used to segment the object in 3D depth images. Given an accurate 3D representation of an object from depth data, manipulation of the object can be performed with visual feedback.”
When Dinkel began her Ph.D. program, she said her biggest challenges were navigating working remotely during the pandemic and organizing her research directions within the project.
“Now I think the biggest challenge of graduate school is trusting the process,” she said. “I have a personal conviction that what I’m doing will work, but there is still a lot of uncertainty in research. Sometimes there are technical problems with hardware or software implementation. Sometimes there is uncertainty about the interpretability of results or the viability of solution candidates. I always wonder if I am asking the right questions. I’m still learning to navigate these issues.”