An agent-based architecture for handover initiation and ... - IEEE Xplore

An Agent-based Architecture for Handover Initiation and Decision in 4G Networks Vassilis E. Zafeiris, Emmanuel A. Giakoumakis Athens University of Economics and Business (AUEB) Department of Informatics Patission 76, 104 34 Athens, Greece {bzafiris, mgia}@aueb.gr Abstract The roll out of 3G networks coincides with the emergence of a variety of wireless access technologies (IEEE802.11x, Bluetooth etc.) supporting high bitrates in restricted areas and acting complementary to wide coverage mobile networks. Integration of the different wireless access technologies is the general trend towards 4G. Our work focuses on the area of handover management in the context of 4G, and specifically on the initiation and decision phases of the handover mechanism. In 3G or earlier generation networks, handovers are initiated and decided by the base station on the basis of measurements of perceived SNR, cell congestion, terminal speed etc. In 4G, due to the diversity of available radio access services additional factors need also to be taken into account. We propose an agent-based architecture that determines the timing and target network of handovers with respect to user, application and terminal constraints.

1. Introduction The next generation of mobile communications systems is orientated towards the integration of available wireless access technologies, enabling a user to seamlessly roam across heterogeneous networks in a way that best reflects her preferences, application requirements and terminal device capabilities. The envisaged architecture of a fourth generation (4G) network comprises a variety of wireless access (WLAN, Bluetooth etc.) and cellular 3G/4G networks interfacing with access routers to a common core IP network, that serves their interconnection and integration with earlier generation networks (PSTN, ISDN, 2G/2G+ etc.) through appropriate gateway routers [10]. Handover management, a standard procedure in mobile communications systems that ensures a terminal’s connectivity as it roams in different service regions (cells), is more

complicated in 4G networks especially in the case of vertical handovers. In contrary to horizontal handovers, i.e. handovers between cells of the same radio access technology, a vertical handover requires integration and interoperation of the networks’ respective AAA (Authorization, Authentication and Accounting), location and address management procedures. In addition, a mapping between their data transfer service classes is essential in order to preserve the QoS of the terminal’s current sessions. Besides handover execution, initiation and decision phases, that precede it in the handover mechanism, pose further requirements in 4G in relation to earlier generation wireless access networks [6]. Initiation is an iterative task, executed either by the terminal, the base station (or access point) in charge or both, that determines when a handover should be triggered. In the decision phase an appropriate target cell is selected for handover execution. In homogeneous networks, perceived power or signal to noise ratio (SNR) are basic factors for handover initiation and decision. Other factors are user velocity, cell congestion etc. In 4G networks, a terminal’s alternatives in certain locations may span a variety of radio access networks with diverse service attributes (cost, bandwidth, QoS). In such cases the role of handover is extended to ensure efficient connectivity with respect to user, applications, terminal constraints. As a matter of fact, during handover initiation and decision additional factors need to be taken into account such as user’s cost tolerance and contractual constraints, applications’ requirements and priority, terminal device status (battery status, available radio interfaces etc.) [5]. We propose an agent-based architecture with purpose to assist a user, roaming in a 4G wireless access environment, in efficiently utilizing available radio access services with respect to her profile, applications’ requirements and current terminal’s capabilities. The architecture’s focus is on the initiation and decision phases of the handover mechanism. It defines agent types and their deployment in user

Proceedings of the Sixth IEEE International Symposium on a World of Wireless Mobile and Multimedia Networks (WoWMoM’05) 0-7695-2342-0/05 $20.00 © 2005 IEEE

terminals and access networks, as well as their reasoning and collaboration in order to determine the timing and target of handovers. Handover target is selected among wireless networks with coverage in the user’s current location, on the basis of user, terminal and application constraints. The rest of the paper is organized as follows: Section 2 presents the merits of the agent-based approach along with a brief description of the architecture. Sections 3 and 4 focus on the agents that comprise the architecture, expanding their functionality and interaction respectively. In Section 5 an implementation of a system that simulates the main features of the architecture is described. Section 6 presents related work and the paper closes with conclusions and future extensions on the architecture.

2. Architecture Overview The architecture deals mainly with vertical handovers, or horizontal inter-domain handovers. Our assumption is that horizontal handovers in the bounds of a certain network, are handled transparently by the network’s mobility management mechanism itself. Network interconnection is enabled through a common core IP network or through the public internet. The first approach applies in the case of a single provider e.g. a 3G or 2G+ operator that enhances service level in highly congested cells (hot spots) with WLAN access points. We consider the second, more general approach where the interconnected networks may belong to multiple wireless access providers. In this case, access and mobility between the various networks is enabled by a third entity, not necessarily owning access infrastructure, that maintains roaming agreements with the providers and will be referred to as Multi-Access Provider (MAP). We can envisage the MAP as a service provider that offers value-added wireless access by integrating the services offered by various wireless providers. The clients of the MAP do not deal directly, in terms of contractual commitments, with the wireless access providers-partners of the MAP. The MAP is responsible for their authentication and bills them by aggregating their respective charges from the various network providers. It is evident that the MAP plays supportive and managerial role in the architecture, enabled by appropriate technological infrastructure. Specifically, it operates an intranet, attached to the internet backbone, that hosts the execution of systems such as authentication servers, software for billing purposes, GIS with the coverage of available wireless networks, databases for persistent storage of user profiles etc.

2.1. An agent-based approach The adoption of software agents as architectural elements and structural components of the proposed system

is due to the special attributes of the application domain. At first, a natural way to conceptualize a distributed system that serves the interests of different cooperating authorities (MAP, wireless providers, users), is through software agents that represent them and interact in order to achieve their delegated goals. System components will be distributed in different administrative domains (terminals, network provider servers) and limited assumptions can be made on their execution environment. Autonomy, a basic property of software agents, renders them self-contained active components that depend exclusively on agent platform services for their execution. The platform is an agent run-time environment providing basic services that enable agent communication, life-cycle management and resource discovery (agents and their respective services). Thus, any network node hosting an agent platform, provides a standard execution context for system components. Moreover, system manageability at the component level is enabled as both agents and platforms (at least those conforming to FIPA standards [3]) provide a standard management interface. Interoperation of system components is a critical design issue as they will be implemented by different parties (MAP, network providers) with possibly different perspectives on the problem domain. A straightforward approach for interoperation is based on the definition of APIs and messaging protocols. Such specifications are implementation specific, focus on the syntax of the messages and may have different interpretation by the various providers. Agent interoperation is based on the exchange of messages expressed in one of the widely accepted Agent Communication Languages (ACLs), FIPA ACL or KQML [8]. An ACL message encapsulates the communication payload and describes it in a domain independent way with a predefined set of attributes. The payload is expressed in a content language (e.g. KIF, RDF) with the use of vocabulary from shared ontologies. Consensus, thus, among providers is reached at a higher, conceptual level that ensures unambiguous interaction. In addition, the well defined semantics of the message content enable the use of automated reasoning during its processing. Finally, an ACL provides a framework for integration into the system of non-agent software (e.g. GIS, network management systems etc.) by the implementation of ACL transducers that manipulate it according to ACL statements [4]. System requirements that relate to personalization and availability of user profile in various terminals and access networks, underpin the agent-oriented approach. Personalization can be supported by user representative agents that encapsulate their principal’s profile and act upon it. Thus, frequent access to a central profile database is avoided, while disconnected personalized operation of system components is enabled.


2.2. Agent platform deployment The software agents that constitute the proposed architecture are executed in agent platforms deployed in the administrative domains of the system actors (Figure 1). Multi-Access Provider has attached to its intranet a platform populated by agents that serve management and support functionality. Their duties include location-based retrieval of available networks, software distribution, management of user profiles etc. Wireless Network Providers (e.g. WLAN, GSM, UMTS providers) incorporate in their core network a platform containing agents that publish available transfer services and their attributes (e.g. cost, QoS). In addition, the platform includes agents, that represent the currently connected users and reason on the timing and target of handovers. User Terminals execute different versions of agent platforms matching their computational capabilities. The main task of the agents hosted in a terminal’s platform, is handover execution, initiated from their counterparts in the access network, and information provision related to user activities, perceived QoS, device status (e.g. battery level) etc.

the presence of the AMS (Agent Management System) and DF (Directory Facilitator) agents. These agents, along with Message Transport Service (MTS) [3], the default communication method between agents on different platforms, are the basic logical components of the FIPA agent management reference model. AMS is a mandatory component that provides agent registration, life cycle control and white page services in an agent platform. DF is an optional component that provides yellow page services to the agents of a platform. The ownership of AMS and DF correlates with the administration of the platform, so, according to Figure 2, each authority manages its own platform.

AMS

DF

Support Agents

AMS

DF

CM-Agent P-Agent

FIPA ACL/KQML IIOP/HTTP/WAP TCP

Application Software

IP Operating System Device Drivers MAP Support Services User Terminal agent platform MAP agent platform AMS

DF

NM-Agent NP-Agent

AF-Agent

MAP representatives Network Management System

UMTS Provider

Multi-Access Provider

Wireless Provider representatives Wireless Provider agent platform

User representatives

Agent Platform

Agent Platform

Figure 2. Agents comprising the architecture. Internet

PDA

3.1. Multi-access provider agents WLAN Provider

Agent Platform

User Terminal

Agent Platform

Figure 1. Agent platforms’ distribution.

3. Agent types and functionality Figure 2 depicts the different types of agents that constitute the proposed architecture along with the platforms that support their execution. Moreover, agents are distinguished, as indicated by the different shading, with respect to the authority they represent. Specifically, dark-grey agents execute on behalf of the MAP, light-grey agents represent one user while striped ones correspond to one wireless access provider. We assume that agent platforms follow the FIPA agent management reference model [3] as it is evident by

Multi-Access Provider agents’ main duty is to support the activities of other agent types residing in wireless networks and user terminals; hence they will be referred to as ”support agents”. Support agents wrap services offered by the MAP’s infrastructure and provide them via an ACL interface. Among the main support services offered by the MAP agents are: • Software distribution. Support agents distribute on demand latest versions of drivers and software libraries, in order to enhance the utilization of the terminal’s network interface(s) and enable the employment of various encoding schemes depending on the features of the current access network. • Storage and retrieval of profile data. User profile information and the configuration of her terminal devices are permanently stored in a database administered by the MAP and transferred to authorized agents in order to reason and decide upon them.


• Location-based network retrieval. The geographical area that is covered by each wireless access network as well as information concerning its type are stored in a geographical database (GIS) under the MAP administration. Support agents interface with the GIS and serve requests regarding: a) the wireless networks of a certain type that cover a geographical location and b) the wireless networks that have overlapping coverage and the same type with a given network.

3.2. Wireless provider agents A wireless provider’s platform is populated by agents that represent all the physical entities that participate in the architecture. The platform includes exactly one Network Provider agent (NP-agent), representing the wireless provider, one representative of the MAP, the Network Monitor agent (NM-agent) and several Access Facilitator agents (AF-agents) corresponding to the currently connected users. Network Provider-agent’s duty is to publish the classes of data transfer services offered by the network its platform is connected to, henceforth referred as home network. Each data transfer service is characterized by a set of attributes, including bandwidth, QoS, cost etc. The attributes and availability of each transfer service are educed from relevant network management information. NP-agent is the only agent in the system that has access to such information and does not disclose any part of it, except for service descriptions, to third-party agents. NP-agent implementation is a basic requirement for a network provider in order to cooperate with the MAP. Network Monitor-agent aggregates information on the transfer services of various available networks and provides it, upon request, to user representative agents in order to support their decision making. Transfer service descriptions are retrieved with periodical requests to the respective NP-agents. NM-agent contacts the NP-agent of its home network and networks of the same type that overlap with its coverage area. For example, a NM-agent residing on a WLAN agent platform aggregates information regarding its home network and other WLANs in its geographical range. The latter ones are discovered with a request to a MAP ”support” agent asking for the contact addresses of NP-agents corresponding to networks that overlap with its home network. Access Facilitator-agent is a user proxy in the network side responsible for handover initiation and decision. It incorporates its principal’s profile (preferences, contractual constraints), the MAP’s policy on the usage of the various providers and the current terminal’s hardware and software configuration. This information, complemented with the terminal’s status (location, battery level) and the current applications’ requirements in terms of bandwidth and QoS,

form the basis for its decision making. AF-agent is instantiated in the MAP agent platform, at the terminal’s initial log on to a network, and thereafter migrates and executes to the current access network, thus reducing communication latency with the terminal. In the case of a multi-mode terminal that maintains connections to more than one network, it migrates to the one supporting the fastest connection. AF-agent contacts periodically a set of NM-agents, one for each type of radio interface supported by the terminal. Among them is its home network’s NM-agent while the others correspond to networks with presence in the terminal’s current location. Each NMagent provides descriptions of data transfer services offered by a certain network type. Service descriptions that correspond to networks with no coverage in the user’s current location are ignored. AF-agent keeps track of available networks in each location by subscribing to a notification service of the MAP that informs it when the terminal enters or passes out of a network’s coverage area.

3.3. Terminal device agents A terminal’s agent platform executes two types of agents that are exclusively user representatives: Profile-agent (Pagent) and Connection Manager-agent (CM-agent). Profile-agent’s role is to communicate the perceived QoS, user preferences and application requirements to AFagent in order to make informed decisions on the target and timing of handovers. Moreover P-agent is shaping the user profile by monitoring her daily data transfer activities and seeking for interesting patterns. This information is forwarded and used by AF-agent in order to predict and thus respond timely to user demands. Connection Manager-agent is responsible for successful execution of handovers initiated by AF-agent. Prior to handover execution AF-agent informs CM-agent of the appropriate driver versions and protocol implementations that the terminal should support in order to connect effectively to the selected network. CM-agent checks the terminal’s software configuration and reports any deficiencies. Finally, appropriate software modules are downloaded from the MAP and CM-agent proceeds with handover execution.

4. Agent communication and interaction High level communication of agents’ beliefs, goals and intentions is realized with the exchange of messages expressed in an ACL. An ACL (Agent Communication Language), as its name suggests, focuses on the structure and communication related attributes of a message, such as sender/recipient address, message type (e.g. assertion, request, query etc.), ontological commitments, supported content languages and interaction protocols, rather than its


content per se. The content of communication is encapsulated in ACL messages and is represented in mutually understandable content languages (KIF, SL, RDF etc. [3]) with the use of vocabulary borrowed from shared domain ontologies. Message transfer is carried out over TCP/IP with the employment of protocols such as HTTP, IIOP, or WAP in case of wireless connections. A more detailed view of the architecture with emphasis on the interaction of the various agent types is presented below.

4.1. Access and authentication At initial log on of a disconnected terminal, network selection is not aided by user representative agents. Their instantiation takes place after user authentication to a network provider-partner of the MAP. Thus the terminal scans the frequency spectrum and attempts to connect to the first available network from a predefined list of wide coverage providers (GSM, UMTS etc.). After user authentication, the executable code of P-agent and CM-agent is loaded either from the MAP or the terminal’s cache (if available from previous connections) and agent instantiation takes place. At the same time AF-agent instantiates on the MAP agent platform and migrates to the current access network. Authentication and identification credentials can be usersupplied (e.g. username and password) or embedded in the terminal (e.g. in the SIM/USIM smart card of a 3GPP terminal). In either case, user authentication requires interoperation between the AAA systems of the MAP and the wireless providers. From the perspective of the terminal device, authentication is carried out by the current network’s AAA server. However, authentication information regarding each MAP client as well as authorization data on the usage of the network are maintained in the MAP. Thus, the network provider’s AAA forwards user credentials for verification to the MAP’s respective server and retrieves (if verified) its usage permissions.

4.2. Handover initiation and decision Handover initiation is an iterative task, distributed among user representative agents, that triggers handovers with purpose to preserve the terminal’s efficient connectivity. Events that trigger handovers are perceived by both P-agent and AF-agent. Figure 3 presents the message exchange sequence during a handover initiated by P-agent. P-agent requests a handover (1) in case of: a) link quality degradation, b) increased bandwidth or QoS requirements set by an application c) special network infrastructure required by an application e.g. for the establishment of a VoIP session. AF-agent may also initiate a handover if the terminal passes out of the coverage area of the current network

or a new more efficient network becomes available. In either case, selection of the target for handover is carried out by AF-agent (2), that incorporates a handover decision algorithm specified and implemented by the MAP. Once handover is decided, AF-agent informs CM-agent of the target network type and identifier (3). CM-agent reports its software configuration to AF-agent that determines whether an update is required (4). Note that if current connection does not allow efficient and timely software download, due to cost or bandwidth constraints, handover is cancelled. If no update is required AF-agent informs CM-agent to proceed with handover execution (5). During handover execution (6), AF-agent migrates to the target network’s agent platform in order to serve its principal with the minimum network latency (7). AF-agent

P-agent

CM-agent

1: Request Handover

2: Handover Decision 3: Initiate handover to network 4: Report terminal's software configuration 5: Proceed with handover 6: Handover execution to netID

7: Migration to netID's agent platform

Figure 3. Handover initiation and decision.

5. Implementation The agents’ basic functionality along with their interaction protocols have been implemented in a simulated environment using an agent development framework. Specifically we used JADE (Java Agent DEvelopment Framework) [7], a framework comprising of an agent platform for agent deployment and execution, and tools for debugging and monitoring the implemented agents. Our primary purpose was to test our approach in handover initiation and decision and evaluate the efficiency of the agent-based approach in the implementation of the architecture. The simulated parts of the architecture were the wireless networks and their operation, the user terminal and its connection shifts to various networks and finally the mobility of the user across the networks’ service areas. Our experience from the simulation system implementation confirmed our choice of software agents as design and implementation units of the architecture. Implementation of system components’ communication and coordination, a


basic issue in distributed systems development, was simplified by using the JADE’s infrastructure. Such infrastructure is present in most agent platform distributions due to the standardization efforts that take place in agent communication and interaction (e.g. FIPA). The system can be easily deployed in several hosts without requirements of any software infrastructure apart from the java runtime environment and the libraries implementing the JADE platform. Finally remote management, debugging and instantiation of the system’s components were supported by respective JADE tools.

6. Related work Requirements for enhanced user experience in a 4G environment, including best utilization of available access devices and networks, seamless mobility, single sign on and application adaptation, are generally characterized by the notion of Always Best Connected (ABC). In [5] ABC is approached from a business perspective, identifying actors and business relationships, as well as from a technical perspective by defining a reference model and basic functional blocks for an ABC solution. Our work provides an agentbased design for the access discovery, access selection and profile handling functional areas of an ABC solution. The DRiVE (Dynamic Radio for IP Services in Vehicular Environments) IST project [2] and the Japanese Government project MIRAI (Multimedia Integrated network by Radio Access Innovation) [9] represent two significant efforts towards the design and implementation of a next generation heterogeneous network. Both approaches enable access through a variety of wireless networks, attached to a common IP backbone. MIRAI also prescribes a broad coverage Basic Access Network (BAN) for signalling traffic that is accessible from terminals through a dedicated radio interface. Handovers are initiated and decided either in the core network (MIRAI) or the terminal (DRiVE) on the basis of various factors. Transfer service descriptions in MIRAI are retrieved from a core network service while in DRiVE each access network publishes its own services. In [1] the architecture of a multi-mode terminal is presented that employs fuzzy logic for initiation and decision of vertical handovers among its radio interfaces. The approach introduces minimal modifications in the access networks but charges the terminal with extra processing load.

7. Conclusions and Future Work We proposed an architecture for handover initiation and decision in the context of fourth generation mobile communications systems. The architecture’s basic structural elements are software agents that either represent users,

network providers or the MAP. The adoption of the agent oriented approach simplifies conceptualization and modelling of the system. Moreover, it enables efficient deployment, management and maintenance of its components. The above concepts were practically verified with the implementation of an agent-based system simulating the architecture’s functionality. A basic point of our approach, is detachment of the handover initiation and decision procedures from the terminal device, and their delegation to a user-representative mobile agent that migrates and executes in the platform of the currently used access network. This results in considerable saving of the terminal’s usually limited power and computational resources. Agent mobility enables any terminal, that incorporates an agent platform, to be used for accessing the MAP services. Moreover, the presence of the user profile in the network side, allows personalized service offering independently of the currently used terminal device. Going a step further, incorporation of the user’s application preferences in his/her profile can extend the architecture to support personal mobility. Our future work will focus on enhancements to the proposed architecture. In the first place, a detailed specification of the decision making procedure is deemed necessary. In addition, the system will be complemented with the integration of a handover execution mechanism. Finally we will focus on performance issues by integrating our simulation system in a real wireless access environment.

References [1] P. M. L. Chan, R. E. Sheriff, Y. F. Hu, P. Conforto, and C. Tocci. Mobility management incorporating fuzzy logic for a heterogeneous IP environment. IEEE Communications, 39(12):42–51, December 2001. [2] DRiVE project. Home page, http://www.ist-drive.org, 2005. [3] Foundation for Intelligent Physical Agents. FIPA Specifications, http://www.fipa.org/specifications/index.html, 2005. [4] M. R. Genesereth. An agent-based framework for interoperability. In J. M. Bradshaw, editor, Software Agents, pages 317–345. AAAI Press, 1997. [5] E. Gustafsson and A. Jonsson. Always Best Connected. IEEE Wireless Communications, 10(1):49–55, February 2003. [6] S. Y. Hui and K. H. Yeung. Challenges in the migration to 4G mobile systems. IEEE Communications, 41(12):54–59, December 2003. [7] JADE. Home page, http://jade.cselt.it, 2005. [8] Y. Labrou, T. Finin, and Y. Peng. Agent Communication Languages: The current landscape. IEEE Intelligent Systems, 14(2):45–52, March/April 1999. [9] G. Wu, M. Mizuno, and P. J. M. Havinga. MIRAI architecture for heterogeneous network. IEEE Communications, 40(2):126–134, February 2002. [10] T. Zahariadis. Trends in the path to 4G. IEE Communications Engineer, pages 12–15, February 2003.
