A Neuromorphic Stereo Vision System for On-Orbit Object Acquisition

krtkl and its teaming partner, University of Texas at Arlington, propose to develop a versatile, energy-efficient, and real-time neuromorphic (event-based) stereo vision solution using field-programmable gate arrays (FPGAs) for deployment in On-Orbit Servicing, Assembly, and Manufacturing (OSAM) applications. The proposed system will be crucial for approaching and rendezvousing with low-Earth orbit objects (e.g., space debris) under extremely challenging and unpredictable lighting conditions. In this project, the team will build upon existing research by assessing algorithms (e.g., object detection, classification, and tracking) using event cameras and power-/performance-optimized deep neural networks (DNNs) via state-of-the-art FPGA Systems on Chip (e.g., Xilinx Zynq Ultrascale+).  Event-based stereo vision systems will play a key role in the development of future autonomous systems, especially for OSAM tasks.