Implemented a Multi-Agent Proximal Policy Optimization (MAPPO) framework for remaining useful life prediction from engine sensor data. The approach uses sensor-level agents with a centralized critic to learn cycle-by-cycle degradation dynamics and estimate remaining life, while also providing sensor-importance insights through policy-confidence analysis. [NASA Open Source Data]
Developed AI-trained artificial viscosity stabilization algorithms learned from the existing Persson Sensor, and integrated them into the existing customized CFD (BoSSS Solver) for online CFD simulations in high-performance computing (HPC). [Paper]
Characterized uncooperative space debris client using a scannerless LiDAR servicer and a customized 3 degree of freedom tumbling rig for ground truth. Employed 3D online template matching for initial pose acquisition and iterative closest point algorithms for dynamic tracking. [Paper] | [Design Review Presentation Slides]
