| || || |
Contents PCoS Parallel SimulatorPCoS Deployment
About usModelling and Simulation
Research and Development
Standards and Methods
Parallel Constellation Simulator (PCoS)
PCoS is a study funded in the frame of the ESA-Portugal Task Force. It focuses on the use of parallel simulation techniques to support spacecraft constellation missions and formation flying. The contract has been executed by EDISoft (PT) as prime contractor responsible for the parallel simulation techniques, and Deimos (PT) as sub-contractor responsible for the formation flying algorithms.
PCoS Serial Simulator
As a first step, a so-called serial simulator has been created in Matlab/Simulink. This serial simulator is then providing the use-case for parallelisation. The Darwin mission has been selected as a reference mission, which uses a constellation of 6 observatory spacecraft and a central communication hub spacecraft. In particular the formation reconfiguration manoeuvres were analysed in detail. As a first step towards parallelisation, the Matlab/Simulink simulator was auto-coded into C-code using the Simulink real-time workshop. This stand-alone version of the serial simulator needed some additional components, e.g. to provide a command line interface needed to run stand-alone. A validation has been performed to check if the results of the stand-alone serial simulator were in line with the serial simulator running in Matlab/Simulink.
PCoS Parallel Simulator
PCoS model development approach
The PCoS parallel simulator is based on the Grid concepts and infrastructure. The latest Grid developments have been in the area of Web services, the so-called Grid services. This technology allows a simulation to be distributed over a WAN, and defines an XML message based SOAP interface. As part of PCoS, automation of the process to convert a stand-alone serial simulator as mentioned above into a Grid service based simulator as described here has been investigated. The resulting simulator has been divided into various grid services, which communicate to each other using standard Grid protocols based on SOAP messages. Furthermore, support for Monte-Carlo analysis has been provided; this allows running (randomly) parameterised simulations in a distributed way. A validation has been performed between the results of the parallel simulator against the original serial simulator.
As it may take a lot of effort to deploy a Grid network, one of the goals of PCoS was to simplify this task. As such, it provides an auto-run CD which sets up any PC as a Grid working node. In a Grid network, there shall be exactly one Master node, which manages all working nodes and e.g. distributes any tasks on the various available working nodes. The Master node provides a portal to communicate with the Grid users, who e.g. want to deploy their Grid services and to run simulations. Furthermore, the Master Grid node is responsible for all data storage, both the input data as well as the simulation results. Finally, it manages all certification and security issues. The working nodes provide a portal as well, but this one has very limited functionality, mainly to configure the node within the Grid network.
Last update: 18 December 2006
| || |