RESEARCH NEWS
XDLab: Agentic AI and Simulation-Driven Space Operation, System and Economy Dr. Di Wu The XDLab group is developing a foundational agentic AI framework for space systems, integrating reinforcement learning, numerical methods and large language models within physics-consistent dynamics. Foundation models serve as high-level reasoning components operating over validated numerical artifacts, emphasizing adaptability and explainability. The lab also advances a simulation-driven debris governance framework that integrates space situational awareness data with physics-based debris modeling. By incorporating structured policy iteration supported by large language models, this work enables quantitative assessment of mitigation strategies and their long-term environmental impact on uncertain surface conditions, supporting future exploration missions.
Space Technologies Laboratory Dr. Troy Henderson
Research in the Space Technologies Laboratory advances AI-driven autonomy for space missions, spanning both trajectory design and navigation. One effort develops reinforcement learning algorithms for low-thrust spacecraft guidance under deterministic and stochastic dynamics, enabling adaptive orbit transfers, inclination changes and asteroid rendezvous while reducing reliance on the Deep Space Network. Another effort focuses on autonomous landing navigation for lunar and Martian missions, combining IMU measurements, Kalman filtering and LSTM networks to estimate spacecraft heading in GPS-denied or disturbed environments, improving landing safety and robustness and supporting the transition from ground-monitored operations to fully autonomous deep-space navigation.
STAR Group: Machine Learning for Cislunar Dynamics and Rendezvous Safety Dr. David Canales Garcia
Asteroid Dynamics and Low-Thrust Deflection Modeling Dr. Francisco Crespo Cutillas
The STAR Group applies machine learning to challenges in cislunar mission design. One line of research analyzes chaotic trajectories in the Earth–Moon system. By leveraging dynamical systems tools and training ML models to extrapolate complex behavior efficiently, the team reduces computational cost while enabling large-scale prediction of trajectory stability. Another effort integrates neural-network-based Control Lyapunov–Barrier Functions to manage uncertainty while jointly enforcing stability and safety. The group also investigates LiDAR-based navigation using machine learning to process 3D point clouds for pose estimation, supporting autonomous docking and in-orbit operations.
This research develops stable Hamiltonian models to describe asteroid dynamics and evaluate low-thrust deflection strategies. Machine learning is used to construct efficient dynamical representations from high-fidelity gravitational simulations. By abstracting complex gravitational behavior into computationally efficient models, this work supports feasibility analysis for planetary defense missions.
4 | AERONEWS 2025 - 2026
Made with FlippingBook - professional solution for displaying marketing and sales documents online