Multi-Agent Passivity-based Control for Perception-based Guidance
Aditya Deole, Shahriar Talebi, Spencer Kraisler, Niyousha Rahimi, Mehran Mesbahi, Saptarshi Bandyopadhyay, Vinod P. Gehlot, William Seto, Amir Reza Rahmani, Jonathan Becktor · AIAA SCITECH 2023 Forum · 2023
View Video Presentation: https://doi.org/10.2514/6.2023-2156.vid This work presents a passivity-based control approach for tracking an uncooperative target spacecraft within the ego camera center via attitude guidance stabilization. A camera and ML based state estimation are used to generate target pose and position estimates in ego camera frame in a photo-realistic simulation environment. The control approach is robust to noisy observations and uncertainties in the perception map. The sensing error is modeled using sector bounded non-linearities and passivity conditions on the perception map. We show that feedback interconnection with CNN based observation is stable under set constraints and derive an optimal control law for the corresponding feasible set. The modular nature of this passivity-based synthesis also allows us to incorporate observations from multiple ego spacecraft, and subsequently, use a consensus protocol on the target position in order to improve the tracking performance.