Session

Technical Session II: Commercial Applications

Abstract

Earth observation using small satellites is leaving demonstration status and being proposed for more and more commercial applications. By analysing such a service from both end-user and satellite operator point-of-view, it is hoped to provide the information such as service performance, on-board resource status and key parameters for system optimisation before the spacecraft is designed, launched and put into service. In this paper, queueing theory, traditionally used to perform tra_c and e_ciency analysis in tele-communication and other queueing system, is applied in this new area - a commercial imaging service using small satellites. The introduction of queueing theory in our application will eliminate the main di_culty of using the traditional solution - operation simulation which is huge computational complexity that arises when the operation spans a long period. In this paper, only the imaging download process, which is usually the system bottleneck for Low Earth Observation spacecraft, will be analyzed. Unlike traditional queueing systems where the service is continuous, our application su_ers regular idle periods when the satellite is not visible from the ground-station for image download. The distribution and duration of such idle periods are the subject of orbital-dynamics. Therefore available queueing theory is not applicable directly and needs some extension to handle this speci_c problem of satellite imaging services. In this paper, three queueing models are discussed: M/G/1, M/Gx/1 and GI/G/1 together with the analysis of their suitability to our application. An extension to using M/Gx/1 is outlined which can provide a better approximation of the service than traditional queueing models. Some basic service parameters, such as queue length distribution, mean service occupation and mean service waiting time, can thereby be calculated. All results presented are compared with that from operation simulation. Limitation and constraints of using queueing theory in this application are also discussed. As a conclusion of this research work, it is shown that queueing theory will be appropriate for the early stage performance analysis in a quick but gross manner which can provide some basic performance parameters, while operation simulation can be treated as a re_nement and a method capable of providing more complete solutions that will certainly take much longer time.

Share

COinS
 
Aug 14th, 9:45 AM

Satellite Imaging Service Analysis using Queueing Theory

Earth observation using small satellites is leaving demonstration status and being proposed for more and more commercial applications. By analysing such a service from both end-user and satellite operator point-of-view, it is hoped to provide the information such as service performance, on-board resource status and key parameters for system optimisation before the spacecraft is designed, launched and put into service. In this paper, queueing theory, traditionally used to perform tra_c and e_ciency analysis in tele-communication and other queueing system, is applied in this new area - a commercial imaging service using small satellites. The introduction of queueing theory in our application will eliminate the main di_culty of using the traditional solution - operation simulation which is huge computational complexity that arises when the operation spans a long period. In this paper, only the imaging download process, which is usually the system bottleneck for Low Earth Observation spacecraft, will be analyzed. Unlike traditional queueing systems where the service is continuous, our application su_ers regular idle periods when the satellite is not visible from the ground-station for image download. The distribution and duration of such idle periods are the subject of orbital-dynamics. Therefore available queueing theory is not applicable directly and needs some extension to handle this speci_c problem of satellite imaging services. In this paper, three queueing models are discussed: M/G/1, M/Gx/1 and GI/G/1 together with the analysis of their suitability to our application. An extension to using M/Gx/1 is outlined which can provide a better approximation of the service than traditional queueing models. Some basic service parameters, such as queue length distribution, mean service occupation and mean service waiting time, can thereby be calculated. All results presented are compared with that from operation simulation. Limitation and constraints of using queueing theory in this application are also discussed. As a conclusion of this research work, it is shown that queueing theory will be appropriate for the early stage performance analysis in a quick but gross manner which can provide some basic performance parameters, while operation simulation can be treated as a re_nement and a method capable of providing more complete solutions that will certainly take much longer time.