Artificial intelligence is advancing space exploration. But how far will it go? To what extent will automated systems relieve humans of the burdens they must currently shoulder in space?
Predicting the future is speculation, but informed experts today can offer educated opinions. For this reason, Filling Space spoke with Ella Atkins, an aerospace engineering professor who directs the University of Michigan’s Autonomous Aerospace Systems Lab. Dr. Atkins shared her thoughts on the future of human-robot cooperation in space.
How did you come to focus your career on autonomous systems and astronaut-robot collaboration?
As an MIT aeronautics and astronautics undergraduate, my first job was an undergraduate research position in the Space Systems Lab. We investigated robots as a means to improve astronaut productivity in space. The goal was never to replace astronauts but rather to improve their productivity and safety. We also studied standalone robot systems for missions such as satellite servicing, structural assembly, and planetary surface exploration. I assisted graduate students in the development of hardware and software to support teleoperated and later autonomous robotic in-space free flight, docking, and structural assembly. Emphasis was placed on experimental validation of concepts in as realistic an environment as is possible on Earth: neutral buoyancy simulation of the zero-gravity environment found in space. Later, as a graduate student, I researched perception and control for automated docking systems.
During my PhD studies in artificial intelligence, I focused my attention more closely on developing the models and algorithms necessary for robotic systems to fully accomplish tasks and missions with minimal to no operator supervision. As I branched out into other domains, most notably the study of autonomy and safety for manned and unmanned aircraft, a central challenge has always been to identify the best roles for human and robotic systems to accomplish each mission efficiently and safely. When an autonomous robot can most effectively complete or assist a task, the challenge is then to develop the software, sensors, and mechanisms to realize that capability. Often the best solution involves a collaboration between humans and robots. In many cases the human is in a remote location from which communication may be delayed or unreliable.
What do you think some of the most promising areas are for autonomy to advance human space exploration?
Space science and exploration are challenged by communication delays and dropouts, a harsh environment, and a need to launch all needed components and systems into space at nontrivial cost per kilogram. Astronauts have historically devoted substantial time to maintaining space habitat systems when they could more productively engage in “fun” science, outreach, exercise, and leisure activities.
Autonomy capable of monitoring and maintaining space habitat systems with minimal to no astronaut oversight is a first step to increased productivity. Autonomy in edge computing and communication system planning and scheduling will also improve our ability to manage large quantities of science and habitat data collected and processed in situ and ultimately shared with mission operators and scientists on Earth. Engineers have recently proposed use of 3D printing technology in space and in planetary surface environments to minimize the need to launch and store all the parts needed for maintenance and ultimately construction of new exploration systems. Autonomy will be essential to manage in-space manufacturing processes to minimize the need for astronauts and human operators on Earth to interact with what is likely to be an ongoing material processing and component production pipeline.
What, if any, aspects of spaceflight do you think will always remain in human hands?
Human dexterity and perception capabilities have not yet been matched by robotic systems. Scientists will therefore continue to be more productive when personally conducting experiments in a space or planetary surface habitat. We also are interested in the potential for humans to colonize Mars and beyond. The only way we can continue to advance our understanding of how humans will survive and even thrive far from Earth is therefore to repeatedly send astronauts to space-based habitats and planetary surface environments. Such flights can further our understanding of physiological and psychological impacts of long-term space travel as a minimum. We will also further our understanding of how robotic and “big data” technologies can better support day-to-day operations so astronauts can achieve an unprecedented level of productivity and enjoyment as explorers in and ultimately “residents” of space and planetary surface habitats.
Human explorers have historically been most productive when they are able to move about in their environment. Astronaut extravehicular activity (EVA) has been essential in past missions to assemble, inspect, and repair in-space structures, but robots have also shown promise in improving productivity by supplementing or even replacing astronaut EVA activities. A major challenge of EVA is the reduction in astronaut dexterity and perception resulting from necessary spacesuit life support and environment protection materials. A human explorer needs to move through the environment. Next-generation spacesuits can employ new technologies such as virtual reality to augment perception and exoskeleton components to augment strength so that the astronaut-spacesuit system is a symbiotic human-robot exploration team.