IN019-08
Evaluation of Remote-Sensing Architectures using the Virtual Constellation Engine

Thursday, 10 December 2020: 10:51
Virtual
Matthew French1, Marco Paolieri2, Vivek Menon3 and Andrew Schmidt3, (1)USC, Information Sciences Institute, Arlington, VA, United States, (2)University of Southern California, Los Angeles, United States, (3)USC, Information Sciences Institute, Arlington, United States
Abstract:
Remote sensing systems are experiencing a transition toward autonomous and distributed architectures with massive scale. UAVs, CubeSats and SmallSats can now be equipped with science-quality instruments and compute accelerators, allowing increased volumes of data to be acquired and processed on-board; in turn, on-board processing enables autonomous coordination among satellites, e.g., to repeat Earth science measurements or to request close-range measurements by UAVs. These distributed and autonomous aspects pose new challenges in how Earth science missions are conceived, developed, and deployed: the performance of computing and networking becomes critical to the success of the mission, as nodes of the constellation have limited storage buffers and communication windows.

The Virtual Constellation Engine (VCE) is a framework to test multi-satellite applications using cloud resources. VCE emulates network latency and bandwidth between nodes of a constellation (satellites, ground stations, UAVs) based on their current coordinates, while different compute capabilities can be assigned to each node using one of the VM types available on AWS (including GPUs and FPGAs) or emulated in software. Execution is monitored by VCE, which collects system metrics such as CPU, memory, disk and network usage, and allows applications to receive instrument outputs (e.g., GPS coordinates) or to record data or events for later debug analysis.

More recently, development efforts in VCE are tackling the integration with the Trade-Space Analysis Tool for Constellations (TAT-C), a tool exploring alternative mission architectures to find optimal designs; VCE is introducing a REST API to automate its use, together with support for execution on Docker containers, and with mission templates of remote-sensing constellations optimized by TAT-C. This integration will allow TAT-C to evaluate quantitative performance metrics for trade-space analysis of remote-sensing satellite applications.