Page 2 Lecture Notes in Computer Science 2865 Edited by G. Goos ...

118 F. Gamba, J.-F. Wagen, and D. RossierIn current deployment of WLAN hot spots, the frequency channel is fixed manually[6] and will be changed only if consumers complain or if a new survey is undertakenby WLAN operators.We developed an autonomous management system that will detect the new situationand adapt the radio resource allocation so that the quality of service is maximized.2.1 Towards Agent-Based ManagementThis paper describes the architecture of an autonomous and adaptive managementsystem dedicated to on-line optimization of access point channel assignment; theprocess mainly relies on environmental information issued from the surrounding accesspoints and information exchange between enhanced access points.The target architecture resorts to autonomous software agents which can run onphysically distributed or centralized platforms. From a logical viewpoint, each softwareagent is delegated to a single access point. Technical information such as thecurrently used frequency channel, the number of users, the number of rejected packets,etc. are gathered by each software agent and constitute their internal knowledge,i.e. their internal representation of the local environment. The implementation ofadvanced mechanisms to exchange internal knowledge between agents will enable theenhanced access point to perform a local optimization.One of the major characteristics of the proposed solution resides in its ability todeal with currently deployed WLAN networks in concordance with the establishedIEEE 802.11 standards [7][8] and related management systems. Thus, managementtechniques directly rely on standardized protocols and information models making avendor-independent implementation possible. Simple Network Management Protocol(SNMP) has become the de facto standard for IP network management. SNMP iscommonly use to manage IP based elements and also for wireless elements.3 Intelligent Agent ApproachThe management of future WLAN networks will have to cope with highly competitiveenvironment where several operators will deploy their own infrastructure in thesame geographical areas. Furthermore, the diversity of WLAN equipments will makesuch a network very heterogeneous. As a consequence, centralized network managementsystems will be not sufficient to control fine-grained resources allocation withrespect to the available frequency spectrum. On the other hand, legacy managementsystems such as SNMP-based systems can not be ignored in the overall architecture.The elaboration of hybrid centralized and de-centralized network management systemstherefore constitute a general trend, not only for WLAN networks but also forother mobile networking technology like ad-hoc networks [9].In this context, intelligent agents can be considered as one of the most promisingapproaches addressing issues related to distributed applications in the rapidly expandingcommunication industry [10]. Intelligent agents can be seen as a software programthat can perform specific tasks for a user and possesses a degree of intelligencethat permits it to perform parts of its tasks autonomously and to interact with its envi-