Sandia Labs FY22 Laboratory Directed Research & Development Annual Report

R-INSULATED 3 MV SWITCHES.

done in subscale with 800 kV switches using high pressure air, has generated data for a 3 MV switch.

This advance in materials and an understanding of the high pressure physics has led to the potential for building compact 3 MV or greater switches. (PI: Randy Curry)

Electromagnetic field plot of enhanced electrodes used to evaluate the Paschen Curve of a subscale high- pressure switch operating at over 30 atmospheres of air.

REAL-TIME EVASIVE MANEUVERS IN CONTESTED, UNCERTAIN ENVIRONMENTS. This research is concerned with real-time evasive maneuvering in contested, uncertain environments. New algorithms for real-time guidance were graduate students from Sandia Alliance partners Georgia Tech and Purdue, and George Mason University. This work led to two journal publications,

developed by leveraging recent advances in Deep Reinforcement Learning. Central thrusts involved: (1) developing high-level action spaces, and (2) concatenating high-level actions (e.g., motion primitives) into feasible, evasive trajectories with Deep Reinforcement Learning and robotics planning algorithms. The methods developed in this project produce trajectories which are feasible, that is, trajectories which do not violate vehicle capabilities and other physical system limits (e.g., heating), and can execute on board the vehicle in real-time. This approach significantly deviates from existing approaches of tracking a trajectory generated offline. Several academic collaborations were instrumental in this effort, including professors and

one in the IEEE Transactions on Aerospace and Electronic Systems Journal and the other in the Journal of Aerospace Information Systems in 2022. (PI: Kyle Williams)

(Top) Real-time evasive maneuvering, and (left) the new algorithm pipeline for real-time guidance.

25

LABORATORY DIRECTED RESEARCH & DEVELOPMENT