A High Performance, Onboard Multicore Intelligent Payload Module for Orbital and Suborbital Remote Sensing Missions
Wednesday, February 18, 2015
Building 3 Auditorium - 11:15 AM
(Coffee and cookies at 10:30 AM)
The purpose of the Intelligent Payload Module (IPM) is to provide a secondary science processor for high data rate science missions which will provide access to subsets of the instrument data in realtime and then provide the ability to send the subset of data rapidly to the ground or to process the data onboard to build quicklook products which would be useful to low latency users such as emergency workers. The target mission for this research was HyspIRI, a hyperspectral instrument mission, with a composite instrument data rate of 930 Mbps and with requirements for access to the survey data within about 2 weeks from the archive. The mission specifications required the IPM to be able to capture onboard a subset of the high rate data and then to rapidly send or rapidly process and then send the data products to the ground via a Direct Broadcast antenna. Since the start of this research effort, the HyspIRI mission has been delayed and so the focus of the research has evolved to include smallsat and cubesat as target platforms in addition to HyspIRI type of missions. The target onboard computation platforms are multicore processor systems augmented with Field Programmable Gate Array (FPGA) circuits. The original target multicore processor was Maestro architecture developed under a NASA/DoD collaboration which is radiation hardened, however, we used the non-radiation hardened Tilera Tile Pro and Tile GX as proxies for our prototyping efforts. We are using SpaceCube and the Xilinx MicroZed for the FPGA portion our effort. We have performed various benchmark tests for the key data processing chain software which consists of Level 0 processing, Level 1 Radiometric Correction, Atmospheric Correction, Geometric Correction and various classification algorithms. This also included testing of our Web Coverage Processing Service (WCPS) onboard which enables users to define in realtime, their own custom algorithm to run on the IPM. Our tests are being conducted by flying the IPM on various airborne platforms such as a helicopter, a Cessna and a Citation with a few different instruments including a ChaiV640 hyperspectral instrument. We are trying to see what it would take to make the IPM compatible with smallsats and cubesats. This effort serves as a stepping stone to infusing high performance computation into sensors within space based SensorWebs.
Dan Mandl has been at NASA GSFC since 1980 and is the Earth Observing 1 (EO-1) Mission Manager and the SensorWeb Project PI. He has been using the EO-1 mission as an orbiting testbed in conjunction with various SensorWeb experiments. Testbed efforts included onboard autonomy flight software (Autonomous Sciencecraft Experiment-JPL), onboard model-based diagnostic and recovery software (Livingstone - Ames), onboard cloud detection (MIT-Lincoln Lab), onboard Delay Tolerant Network (GSFC), Open Geospatial Consortium (OGC) SensorWeb Enable (SWE) web service access and control of EO-1 sensors, compute cloud-based automated data processing chain for EO-1 data products, user defined and executed algorithms for higher level data products and the use of OpenStreetMap and crowdsourcing for standardized decision support data products for various applications like emergency response. Many of the SensorWeb experiments were conducted in collaborative efforts with NASA Jet Propulsion Laboratory, NASA Ames, the Open Geospatial Consortium(OGC), Open Cloud Consortium (OCC), Group on Earth Observations (GEO), Committee on Earth Observation Satellites (CEOS), United Nations Platform for Space-based Information for Disaster Management and Emergency Response (UN-SPIDER), Centro del Agua del Tropico Humedo para America Latina y el Caribe (CATHALAC) in Panama, Regional Center for Mapping of Resources for Development (RCMRD) in Kenya, and National Hydrological Services in Namibia to provide real world test cases for SensorWeb for disaster management decision support. The present effort is focused on accelerating selected SensorWeb functionality to enable high speed onboard multicore processing (low power, radiation tolerant processors) augmented with Field Programmable Gate Arrays (FPGA), automated parallelization of single threaded programs, high speed data processing chain software. Research for these efforts were funded under various NASA Earth Science Technology Office (ESTO) research grants. He won the Robert H. Goddard Award of Merit for 2013 for SensorWeb and other related technical leadership efforts and also won the 2007 NASA Exceptional Achievement Award.
IS&T Colloquium Committee Host: Karen Moe
Sign language interpreter upon request: 301-286-7040