Kazuki Toma

Physics-Based Diagnosis Model and Digital Twin Simulation for Space Habitats

Aug, 2024 - May, 2028 (Ongoing)

Research Ph.D. Ongoing
Back to Research
Advisor:
Prof. Daniel Selva
Institution:
Texas A&M University, Systems Engineering, Autonomy and Knowledge Laboratory (SEAK)
Funding:
NASA Human Research Program (Grant #80NSSC19K0656) & NASA STTR Phase II (Award #80NSSC24CA028)

Research on virtual assistants (VA) for spacecraft anomaly treatment during long duration exploration missions

Supported by NASA's Human Research Program (Grant #80NSSC19K0656) and NASA STTR Phase II (Award #80NSSC24CA028, with TRACLabs)

Contribution: Developing a physics-based anomaly diagnosis capability for Daphne-AT, a prototype VA, using hybrid model-based reasoning and simulation-based fault trend analysis with real-time telemetry.

Current Research

Development of a Physics-Based Diagnosis Model and Digital Twin Simulation of a Space Habitat for End-To-End Autonomous Anomaly Resolution

Co-authors:
J. Elston, D. Kortenkamp, A. Diaz-Artiles, D. Selva
Institutions:
Texas A&M University, TRACLabs Inc.
Funding:
NASA Human Research Program (Grant #80NSSC19K0656) & NASA STTR Phase II (Award #80NSSC24CA028)

Abstract: Astronauts in deep space exploration missions will experience long communication delays with mission control. In such environments, spacecraft anomalies or unexpected sudden environmental changes can occur, some of which could lead to loss of the mission or loss of life in a short period of time. Virtual Assistants (VAs) that can seamlessly retrieve information from several heterogeneous data sources and combine it to produce natural language answers are a promising technology to support astronauts during these situations.

Daphne-AT was developed to study the impact of such VAs on human performance and has been used for several laboratory and analog studies. This poster describes the ongoing development of a physics-based reasoning capability in Daphne-AT. A real-time telemetry data feed can be assimilated into the physics-based model to produce a digital twin of the habitat, significantly improving anomaly detection and diagnosis capabilities.

The integrated loop demonstrated end to end

Fault in BioSim → Daphne-AT diagnosis with operator approvals → PRIDE procedure → robot inspection and repair → telemetry back to nominal

Daphne-AT time-series matching analysis and chatbot approval request
Daphne-AT time-series matching analysis and chatbot approval request
Interaction sequence between BioSim, Daphne-AT, operator, PRIDE, and robot
Interaction sequence between BioSim, Daphne-AT, operator, PRIDE, and robot
CRAFTSMAN robot arm at the air filter rack in simulation
CRAFTSMAN robot arm at the air filter rack in simulation
Experiment overview: anomaly scenario player, telemetry simulator and robotics environment, Daphne-AT, PRIDE, and crew / CRAFTSMAN, with mission control across a communication delay
Experiment overview: anomaly scenario player, telemetry simulator and robotics environment, Daphne-AT, PRIDE, and crew / CRAFTSMAN, with mission control across a communication delay
habitat-dx architecture
habitat-dx architecture

Research Posters and Presentations

Recent Conference Presentations:

  • 2025 NASA HRP IWS: "Development of a Physics-Based Diagnosis Model and Digital Twin Simulation of a Space Habitat for End-To-End Autonomous Anomaly Resolution" - Galveston, USA, January 2025
  • 2025 NASA HRP IWS: "Bayesian Network Approach to Collaborative Anomaly Diagnosis in Human-Agent Environments" (Co-authored) - Galveston, USA, January 2025

Research Materials:

Interactive HTML slide deck (habitat-dx)open in the browser; use ←/→ to navigate
Related Publications:Initial Study of a Physics-Based Virtual Assistant for Real-Time Fault Diagnosis in Space HabitatsA Bayesian Framework for Human-Agent Collaborative Fault Diagnosis in Space HabitatsDevelopment of a Physics-Based Diagnosis Model and Digital Twin Simulation of a Space Habitat for End-To-End Autonomous Anomaly ResolutionBayesian Network Approach to Collaborative Anomaly Diagnosis in Human-Agent Environments