So Young Kim

JPL-NASA, U.S.A.

Lecture. Designing Human-Machine Teaming in Robotic Space Exploration Missions

Abstract.

   Human-machine teaming (HMT) in robotic space exploration missions that JPL designs, builds and operates is in a different context compared to other HMT domains. Humans cannot maintain continuous communications with machines in robotic space exploration missions. Spacecraft may experience solar or other conjunctions where the Sun or other planets are in the line of sight to Earth, creating a communication blackout. Sometimes, the mission operations team does not have the “track” from deep space network facilities to communicate with their spacecraft. (It is like a phone line becoming busy, so you cannot use it.) Therefore, the communications between humans and machines could encounter unplanned blackouts, regardless of the best planning efforts. In addition, communication latency depends on the distance from Earth, limited by the speed of light. When humans did get the chance to communicate with their spacecraft, due to the latency, there is no concept of real-time communication that we tend to assume in other HMT domains.

   Another interesting aspect of human-machine teaming in this context is that the relationship between the mission operations teams and their spacecraft is less intimate than in other domains. Unlike pilots with their aircraft or drivers with their self-driving cars, one spacecraft interacts with an entire team and teams of operators, which may involve about 200 people in a given day. Operating a robotic space mission entails numerous negotiations between what various groups of scientists want to observe (often competing given limited resources like energy, memory, etc.) and what operations engineers care about (maintaining spacecraft health and safety). Therefore, trust in autonomy in this context is not only an operator's trust but also the entire team's trust (also often the institution's trust) in the spacecraft's ability to manage situations.

   Specifically, these contexts both propel and hinder the infusion of autonomy capabilities into spacecraft. The limited real-time commanding capability of the spacecraft required the development and deployment of autonomous fault management capabilities (because there is no way for the operations team to intervene on time). Also, the limited observability of the spacecraft's progress enabled the infusion of an autonomous navigation system for Mars rovers. These are typically referred to as Functional Autonomy. On the other hand, the so-called System Autonomy (or Executive Autonomy) has yet to be infused into the missions. System Autonomy's function is to be able to achieve goals without the operations team specifying every step to achieve those. Currently, most space missions are operated by the operations team, providing the full list of activities to do until the next time they can communicate with the spacecraft. The spacecraft can barely veer off from that unless there is a fault, at which time the predefined fault management capabilities kick in. This is, I would call it, a “human micro-managing machine-teaming paradigm.” The basis of human-machine teaming for robotic space missions is “do exactly what I say. If you can't, call me and wait until I tell you what to do.”

   However, along with recent advancements and the proliferation of AI, machine learning, and autonomy technologies, more daring mission concepts with more complex capabilities onboard spacecraft are pushing the institutional appetite toward onboard autonomy. From an HMT perspective, this shift towards highly autonomous missions leaves the mission operations, science, and management communities with one of the fundamental questions of HMT. What are the roles and responsibilities of the ground operations team in operating highly autonomous spacecraft?

    What decision can and should the humans make, and what decisions could, and should the spacecraft make on its own? Do humans have only a supervisory oversight role in autonomous operations?

   How do the humans specify their intents (science goals and objectives) to the spacecraft? That is, how to develop a shared language and a shared common meaning between humans and the spacecraft, that achieves the set-out goals by the humans, while also minimizing risks during the autonomous execution.

   How can the human understand enough of the spacecraft's decisions, actions, and their consequences? Note that we have significant constraints in volume, timing of availability, and computing resources of data. Explainability is already a challenging subject but what do we do when we cannot get data we want and when we want?

   What does trust mean in HMT for robotic space exploration? What are the verification and validation steps to develop trust in the autonomous actions of the spacecraft?

   In this lecture, we will go through representative examples of current and future planetary missions and highlight how the mission goals are addressed through the spacecraft hardware as well as the flight system systems and mission operations. Exploring these questions will lead us to the design of a new HMT paradigm for highly autonomous space exploration missions.

Bio. Dr. So Young Kim is the Technical Point of Contact for Human Systems Integration (HSI) for NASA JPL. She also serves as the Chief Engineer for the Enterprise and Information Systems Engineering Section. At JPL, she leads and executes tasks related to how HSI can contribute to designing and building of JPL missions. In addition to her HSI role, she works as the spacecraft systems engineer for Mars Sample Return Lander. Her past work at JPL includes Human-Autonomy Teaming Design efforts for Europa Lander and Human-Centered Design efforts on projects such as AR/VR system designs and Model-Based Systems Engineering capabilities for spacecraft design. Prior to that, she worked at General Electric’s Global Research, shaping and leading programs such as Future Flight Deck and Future Power Plant. Her specialty is Human Systems Integration and, in particular, Human-Autonomy Teaming. She received her Ph.D. and M.S. in Aerospace Engineering from Georgia Institute of Technology, Atlanta, GA.