RADIO PLANNING IN MULTIBEAM GEOSTATIONARY SATELLITE NETWORKS C. Touati * , E. Altman * , J. Galtier *,‡ , B. Fabre § , I. Buret § * INRIA BP93, 06902 Sophia Antipolis Cedex, France ‡ France Telecom R&D, 06921 Sophia-Antipolis Cedex, France § Alcatel Space Industries, Toulouse, France We consider the problem of how a geostationary satellite should assign bandwidth to several service providers (operators) so as to meet some minimum requirements, on one hand, and to perform the allo- cation in a fair way, on the other hand. In the paper, we firstly address practical issues (such as integrity constraints), whereafter we provide a computational method for obtaining an optimal fair allocation in polynomial time taking the practical issues into account. keywords: bandwidth allocation, fairness, geosta- tionary satellite. 1 Introduction We consider a multi-spot geostationary satellite sys- tem for which a manager wants to assign bandwidth between various service providers (operators) that op- erate in different geographical areas. An actual as- signed unit of bandwidth may correspond to different amounts of throughput, depending on many factors, and in particular on weather conditions. Indeed, dur- ing bad weather in an area, a local operator may have to use a larger part of its bandwidth for redundant in- formation (a higher coding rate for error correction), and thus the effective throughput of information de- creases. Therefore, if the objective was to maximize the global throughput, it would become non-profitable to assign bandwidth to operators in areas that suffer from bad weather if this bandwidth could be assigned to other operators instead. It is therefore interesting to understand and then to propose bandwidth assign- ment schemes that are more fair and do not systemat- ically penalize operators that suffer from bad weather conditions. The geographical area covered by a geostationary satellite is divided into hexagonal areas called spots. Each spot uses a certain frequency range denoted as its color. We denote S(c) the set of spots of a given color c. In practice, as the number of available fre- quencies is limited, each satellite has a fixed number of colors that it can use. Two spots having the same color could interfere with each other. That is why they need to be geographically distant. The colors are statically assigned to the spots. A spot s is further di- vided into a set Z (s) of zones. They are small enough to assume that the weather condition is the same in any point of a given zone. Then, every operator in a given zone uses the same coding rate and maximizing the global throughput does not penalize any operator with respect to any other within the same zone. 1.1 Technical framework A central difficulty for solving such systems is that integrity constraints may arise: we might not be able to assign any value of bandwidth between the mini- mum and maximum given values. Instead, each oper- ator in the set O of operators can be assigned one or more carriers in each zone among the set of N types of carriers: T = {1,...,N }. Let B t be the overall bandwidth of a type-t carrier. We assume that B 1 > ... > B N . B t is not directly proportional to the actual through- put of information of a type-t carrier. Firstly, as men- tioned before, the throughput depends on the coding rate, which may be different from one zone to another due to atmospheric conditions. Secondly, the effective throughput is lower due to overheads (around 10%) for signaling, frequency margins, etc; the percentage of overhead depends on the carrier type. To handle that, each carrier is associated to the utility C t (z,o). The utility is the value that operator o at zone z is willing to pay for a carrier of type t. It can be cho- sen as a function of the amount of redundancy (in the channel coding) which depends on the atmospheric conditions at each zone. Thus the utility can be made proportional to the actually assigned throughput, so the problem solved becomes how to fairly (or opti- mally) assign throughput. We assume that there is a minimum and a maximum number of type-t carriers per zone z required by each operator o, denoted by D min t (z,o) and D max t (z,o) re- spectively. We assume that the minimum requirement can be satisfied; if not, our algorithm can be adapted to find new fair minima, see Appendix A. The actual implementation of bandwidth assign- ment to users involves two phases. The first concerns the allocation of the global bandwidth to each opera- tor in each zone. In that phase we ignore the problem of interferences between carriers of the same colors as- 1 of 8 American Institute of Aeronautics and Astronautics 21st International Communications Satellite Systems Conference and Exhibit AIAA 2003-2271 Copyright © 2003 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.