Enhancing Interoperability in Heterogeneous Mobile ... - CiteSeerX

2424

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 8, NO. 5, MAY 2009

Enhancing Interoperability in Heterogeneous Mobile Wireless Networks for Disaster Response F. Richard Yu, Senior Member, IEEE, Jie Zhang, Helen Tang, Henry C. B. Chan, Senior Member, IEEE, and Victor C. M. Leung, Fellow, IEEE

Abstract—Efficient communications are crucial for disaster response and recovery. However, most current public safety land mobile radio (LMR) networks only provide narrowband voice service with limited support of low-speed data services. In this paper, we study to enhance the interoperability of LMR with commercial wireless cellular networks, by which a wide variety of benefits can be offered to disaster responders, including new multimedia services, increased data rates and low cost devices. Our approach is based on Session Initiation Protocol (SIP) and a joint radio resource management framework. We use a novel SIPbased seamless handoff scheme to support the interoperability between cellular and LMR networks. In addition, an optimal radio resource management scheme is proposed to maximize the overall radio resource utilization and at the same time guarantee service availability and continuity quality of service (QoS) for disaster responders. The effectiveness of the proposed schemes is illustrated by numerical examples. Index Terms—Interoperability, public safety land mobile radio, wireless cellular networks.

I. I NTRODUCTION

D

ISASTER response and recovery require timely interaction and coordination of disaster responders in order to save lives and property. Efficient communications are crucial during disasters. With recent advances of wireless technologies, mobile wireless networks play an increasingly important role in disaster response. Currently, public safety land mobile radio (LMR) is used by public safety agencies for coordinating teams and providing rapid emergency response. Most public safety mobile wireless networks currently being deployed throughout the world are based on two digital narrowband LMR technologies: Association of Public Safety Communications Official (APCO) Project 25, standardized by Manuscript received October 18, 2007; revised April 15, 2008; accepted June 22, 2008. The associate editor coordinating the review of this paper and approving it for publication was H.-H. Chen. F. R. Yu is with Carleton School of Information Technology and the Department of Systems and Computer Engineering, Carleton University, 1125 Colonel By Drive, Ottawa, ON, Canada K1S 5B6 (e-mail: [email protected]). J. Zhang and V. C. M. Leung are with the Department of Electrical and Computer Engineering, University of British Columbia, 2332 Main Mall, Vancouver, BC, Canada V6T 1Z4 (e-mail: {zhangj, vleung}@ece.ubc.ca). H. Tang is with Defense R&D Canada - Ottawa, ON, Canada (e-mail: [email protected]). H. C. B. Chan is with the Department of Computing, The Hong Kong Polytechnic University, Hong Kong, P.R. China (e-mail: [email protected]). This work is supported by Natural Science and Engineering Research Council of Canada. Digital Object Identifier 10.1109/TWC.2009.071158

the Telecommunications Industry Association (TIA) and Electronic Industries Alliance (EIA), and Terrestrial Trunked Radio (TETRA), standardized by the European Telecommunications Standards Institute (ETSI). There are also some efforts within TIA Technical Requirement Working Group 8.08 (TR8.08) and Project Mobility for Emergency and Safety Applications (MESA), a partnership between TIA and ETSI. Although some technologies (e.g., IEEE 802.11a/j in the 4.9 GHz band) are being considered as candidate access technologies for future public safety radio networks, most current deployed networks provide narrowband circuit-switched digital voice service with limited support of low-speed data services [1]. In contrast, current commercial wireless cellular networks, e.g., third generation (3G) wideband code division multiple access (WCDMA) systems, can support packet-switched, broadband services with a variety of multimedia applications that include voice, data, web browsing and video [2]. High speed downlink packet access (HSDPA), defined in Release 5 of the third generation partnership project (3GPP), can support 14.4 Mbps downlink data rate [3]. The fourth generation (4G) cellular can provide very high data rates (exceeding 100 Mbps), which will emerge between 2010 and 2015 [4]. The significant differences (in terms of services and data rates) between these two kinds of wireless networks are largely due to market forces, requirements, spectrum policy and other factors [5]. For example, the commercial wireless cellular user community is two orders of magnitude larger than the public safety LMR base. As a consequence, the R&D investments in commercial wireless cellular networks dwarf those made in public safety LMR networks. During disasters, efficient communications are crucial for disaster responders in disaster response and recovery. For example, it is desirable for the disaster responders to have the access to the Internet to share real-time multimedia information with off-site commanders and specialists providing expert assistance. However, these communication services are not available in the current public safety LMR. Whereas in commercial cellular networks, less service availability means less revenue; in public safety arena, less service availability may impact lives. Therefore, it is attractive to enhance the interoperability of these two wireless networks, by which a wide variety of benefits can be offered to disaster responders, including new multimedia services (e.g., video), increased user data rates and low cost devices. In the interoperable cellular and public safety LMR networks, disaster responders can access the services in cellular

c 2009 IEEE 1536-1276/09$25.00

YU et al.: ENHANCING INTEROPERABILITY IN HETEROGENEOUS MOBILE WIRELESS NETWORKS FOR DISASTER RESPONSE

networks that are not available in public safety LMR networks to increase the service availability. Furthermore, when a disaster responder moves out of the coverage of public safety LMR networks with an ongoing communication session, the session should be handoffed to cellular networks instead of being dropped to provide the communication continuity. There are some schemes proposed in the literature for the interoperability of heterogeneous wireless networks. Authors of [6] propose a location management scheme, including location update and paging, in heterogeneous systems. The optimal conditions under which vertical handoff should be performed is studied in [7]. Authors of [8] and [9] study several admission control schemes in cellular/WLAN integrated networks to improve the performance of voice and data services. Scalable routing techniques are proposed in [10] for heterogeneous mobile networks. In this paper, we study to enhance the interoperability between cellular and LMR networks based on Session Initial Protocol (SIP) [11] and a joint radio resource management framework, which are different from the schemes in previous work [5]-[10]. SIP is designed by the Internet Engineering Task Force (IETF) to provide application-layer signaling for voice and multimedia session management, which can achieve true end-to-end mobility management. In addition, SIP has excellent extensibility and scalability due to its operation at the highest layer and use of text-based control messages. Several wireless technical fora (e.g., 3GPP, 3GPP2 and MWIF) have agreed to use SIP to provide session management. However, traditional SIP-based handoff scheme may have considerable handoff delays due to the exchange of application layer messages, which may be unacceptable for real-time multimedia services [12]. Moreover, since multimedia applications are resource-intensive in wireless networks, radio resource management is one of the major challenges in designing the interoperable cellular and public safety LMR networks. The objective of this paper is to enhance the interoperability between cellular and LMR networks by addressing these challenges. The contributions of this paper are as follows. •

•

•

We use a novel SIP-based seamless handoff scheme (SSIP) to support the interoperability between cellular and LMR networks. This scheme employs a “make-beforebreak" handoff procedure to efficiently reduce handoff delays. We propose a joint radio resource management framework in interoperable wireless cellular and public safety LMR networks to manage the overall radio resource in these two heterogeneous networks. An optimal radio resource management scheme is proposed to maximize overall radio resource utilization and at the same time guarantee service availability and continuity quality of service (QoS) for disaster responders.

Extensive numerical examples illustrate the effectiveness of the proposed schemes. The rest of the paper is organized as follows. Section II describes the interoperable cellular and public safety LMR networks. Section III presents the joint radio resource management framework. Section IV presents the optimal radio resource management scheme. Some numerical examples are

2425

given in Section V. Finally, we conclude this study in Section VI. II. I NTEROPERABLE C ELLULAR AND P UBLIC S AFETY LMR N ETWORKS A common Internet Protocol (IP)-based core network can be used to bridge cellular networks and LMR networks. In this section, we describe the IP Multimedia Subsystem (IMS) architecture in 3G cellular networks. We then present the interoperable cellular and public safety LMR networks. The novel SIP-based seamless handoff scheme is also presented in this section. A. IP-Based Networks in Wireless Cellular Networks IMS [13] is introduced in 3GPP release 5 and is being updated in releases 6 and 7. This is a key evolution of the core networks as it opens 3G networks to the seamless provision of multimedia services. The 3GPP defines the IMS as the component that supports multimedia services (e.g., voice and video) based on IP with QoS and authentication, authorization, and accounting (AAA) provision. The IMS architecture is composed of a set of call session control function nodes (CSCFs). They are signaling proxies whose task is to establish, modify, and release media session with guaranteed QoS, AAA and charging (AAAC) support. The tasks of the CSCFs include functionalities of SIP proxies, but are not limited to the above mentioned QoS and AAAC tasks. There are several types of CSCFs: proxy CSCF (PCSCF, which acts on behalf of the mobile terminal in the IMS), serving CSCF (S-CSCF, which implements user registration and session control), and interrogating CSCF (I-CSCF, with proxy and topology hiding functions between operators). The home subscriber server (HSS) is similar to the home location register (HLR) in a GSM network: a centralized database that stores user authentication and profile information. Moreover, application servers (ASs) can be connected to the IMS in order to provide advanced services. SIP is the core protocol chosen by the 3GPP to perform signaling tasks in the IMS. SIP is a general-purpose application layer protocol designed to establish, modify, and release sessions in IP networks. SIP is not a vertically integrated communications system. SIP is rather a component that needs to be used with other IETF protocols (e.g., RTP and RTCP) to build a complete multimedia architecture. SIP supports five basic aspects of multimedia sessions: user location, user availability, user capabilities, session negotiation, and session management. B. Interoperable Cellular and Public Safety LMR Networks Several schemes are available in the literature to solve the handoff problem in heterogeneous networks. Mobile Internet Protocol (MIP)-based scheme [14] works at network layer. Authors in [15] present a transport layer approach based on Stream Control Transmission Protocol (SCTP). SIP-based handoff scheme is proposed in [12]. Among these solutions, the SIP-based solution can achieve true end-to-end mobility without the need to modify the network architecture or enduser terminals. In addition, SIP is also adopted in the IMS

2426


Remote Node (RN)

Internet

P-CSCF

I-CSCF

DHCP

GGSN

SIP Proxy

S-CSCF Access Control

Public Safety LMR Networks

HSS Application Servers Cellular Networks

Fig. 1.

Mobile Node (MN)

the UA will register its current address with the SIP registrar service in the home domain. Therefore, other nodes can always track it. Such mobility can happen between or during active sessions, and is referred as out-of-session mobility and insession mobility, respectively. Out-of-session mobility can be easily solved by sending a REGISTER message to the registrar (HSS in the IMS). In-session mobility is usually enabled by sending an INVITE message to the remote node (RN) notifying the new domain/IP address and session parameters. This will cause a certain period of disconnections, and data packets will be lost. This delay is made up by several latency factors, such as authentication latency, IP address acquisition latency and media-redirection latency. The delay incurred by handoff is too long to support real-time multimedia services [12]. Therefore, a more efficient handoff scheme is needed to enhance the interoperability between cellular and LMR networks.

Interoperable wireless cellular and public safety LMR networks.

C. SIP-based Seamless Handoff Scheme and has excellent extensibility and scalability. The IMS scope is now extended in Release 7 standardization for other access networks. To provide the interoperability between cellular and public safety LMR networks, LMR networks can be treated as other access networks within the IMS framework. Fig. 1 presents the interoperability model used in this paper, showing the signaling interfaces between both networks. A general network model is used for public safety LMR networks. In this figure, the I-CSCF proxy is acting as the signaling entry point in the interconnection between cellular and public safety LMR networks, according to the 3GPP CSCF’s role definition [13]. Since natural disasters or terrorist attacks often occur in a localized region, we assume that the coverage of the LMR is under the coverage of the cellular network. The mobile devices used by disaster responders are equipped with multiple radio interfaces that enable them access both the LMR and the cellular network within the coverage of the LMR. However, for commercial users, only the cellular network can be accessed. IP-based multimedia services (e.g., video streaming) are available to disaster responders via the cellular network, and mission-critical services (e.g., tactical group voice) are provided to them via the LMR. Since disaster responders are free to move in the interoperable LMR/cellular systems, the support of handoff between these two networks, which provides ongoing service continuity, is needed in this integration. In this interoperable system, disaster responders are efficiently communicated with state-of-the-art applications during a disaster. A SIP-enabled mobile device has a SIP user agent (UA) to manage SIP messages on behalf of the user. UAs are identified by SIP uniform resource identifiers (URI) with an email-like address: user@domain. Various control messages are defined in [11] to manage sessions between UAs. A client-server interaction model is used for the exchange of messages between UAs. To setup (terminate) a session, an INVITE (BYE) message is sent from the requesting UA client to the target UA server. If a mobile device’s network address is changed,

Unlike the traditional SIP that provides “break-beforemake" handoff, the novel SIP-based seamless handoff (SSIP) scheme offers a “make-before-break" scheme to reduce the handoff delay and achieve seamless handoff. We use an example to illustrate how a mobile node (MN) handoffs from a LMR network to a cellular network. Fig. 2 shows the message sequence for the handoff procedure. Initially, a MN with SIP URI MN@LMR in the LMR network sends an INVITE message to a RN. The RN agrees to establish a session by replying with an OK message. An ACK message is then sent back to the RN. After the setup procedure, a session is setup to exchange data. During the session, the MN wants to handoff to the cellular network for more advanced multimedia services. It then authenticates with the cellular network and acquires IP and/or domain addresses. We assume that the dynamic host configuration protocol (DHCP) server is used in the cellular network and the MN is assigned with an address of MN@Cellular. After obtaining the new address, the MN sends an INVITE message to the RN via the cellular network. In this way, the RN knows that the new session wants to join the ongoing session between the RN and MN via the LMR network. After negotiating the parameters, another session is established between the MN and RN via the cellular network. The RN synchronously sends data to both networks. After the handoff transaction, the MN and RN communicate through the two sessions independently and synchronously. The MN will discard any duplicate RTP packets. When the new session is setup, the MN will send a BYE message to the RN to terminate the session via the LMR network. It also updates the contact address in the HSS with the new address MN@Cellular. The handoff period of the proposed scheme can be determined as follows. Note that the data sent during this handoff period can still reach the MN due to the seamless handoff nature. From Fig. 2, we can see that the handoff procedure starts with message 4 and ends with message 16. As the registration to the HSS can be executed in parallel with message 10-14, the handoff time only counts from message 4 to message 14. Let Dhandoff be the handoff delay and DA↔B be the delay of messages transmitted between nodes A and B.


UA_LMR

UA_Cellular

CSCF

DHCP

RN

HSS

(1) Invite (2) Ok Setup Session

(3) ACK Data (4) AAA Request (5) AAA Reply

Attach to New Session

(6) DHCP Discover (7) DHCP Offer (8) DHCP Request (9) DHCP ACK (10) Invite, Join, referred by UA_LMR

Setup New Session

(11) OK (12) ACK Data (13) Bye

Register

(14) Ok (15) Register

End Old Session

Fig. 2.

(16) Ok

LMR to cellular network handoff procedure.

We have: Dhandoff = 5DMN ↔RN + 2DMN ↔CSCF + 4DMN ↔DHCP . (1) Generally, there are three types of delay: processing delay at end nodes, transmission delay (wireless and wireline) and wireless propagation delay. The propagation delay over a wireless access network is very small and can be neglected. The delay incurred by messages involves: (a) processing delay at end nodes (MN or RN), (b) transmission delay over wireless links, and (c) transmission delay over the Internet. Therefore, we have: DMN ↔RN = DP,MN + DT,L (DT,C ) + DT,I + DP,RN , (2) DMN ↔CSCF/DHCP = DP,MN + DT,C + DP,CSCF/DHCP , (3) where DP,MN and DP,RN are the process delay at the MN and RN, respectively, DT,L , DT,C and DT,I are the transmission delay over the LMR, cellular and Internet, respectively. Substituting (2) and (3) into (1), we have: Dhandoff = 11DP,MN + 5DP,RN + 2DT,L + 9DT,L

+5DT,I + 2DP,CSCF + 4DP,DHCP .

(4)

III. J OINT R ADIO R ESOURCE M ANAGEMENT IN I NTEROPERABLE C ELLULAR AND P UBLIC S AFETY LMR N ETWORKS Radio spectrum is one of the most important resources in wireless networks. Multimedia applications are resourceintensive in wireless networks. To provide good quality of service to end users and optimally utilize the overall radio resource simultaneously is one of the major challenges in designing the interoperable cellular and public safety LMR networks. In this paper, we propose a joint radio resource management framework in interoperable wireless cellular and public safety LMR networks. The framework can be used to manage the overall radio resource in these two heterogeneous networks. A. QoS in Interoperable Cellular and Public Safety LMR Networks Service availability and continuity are important issues in public safety arena, because less service availability and

2427

continuity may impact lives during disasters. In the interoperable systems, the key session-level QoS measures, P N , new session blocking probability, and P H , handoff (from LMR to cellular networks) dropping probability, experienced by disaster responders, should be kept below a target value. For simplicity of the presentation, we consider an interoperable LMR/cellular system with a single LMR cell and a single cellular cell. Disaster responders can access both the LMR and the cellular networks, whereas commercial users can only access the cellular network. When a first responder with an ongoing communication session moves out of the coverage of the LMR, the session can be handoffed to the cellular network if resource is available there. There are J classes of traffic. Class j, j = 1, 2, ..., J, new sessions arrive according to a Poisson distribution with the rate of λc,n,j (λl,n,j ) in the cellular (LMR) area. Class j handoff sessions depart from the LMR to the cellular network according to a Poisson distribution with the rate of μl,h,j . Session duration time for class j traffic is exponentially distributed with the mean 1/μc,t (1/μl,t ) in the cellular (LMR) area. In the LMR, for class j mobile devices used by disaster response personnel, the new session blocking probability, PjN , should be kept below a target value to guarantee the service availability. Let T PjN denote the target value of the new session blocking probability of class j. PjN ≤ T PjN . (5) The handoff (from the LMR area to the cellular area) dropping probability, PjH , should be kept below a target value to guarantee the service continuity. Let T PjH denote the target value of the handoff dropping probability of class j. PjH ≤ T PjH .

(6)

Assume that there are totally C channels in the LMR. The number of channels used by a class j session is cj . Let nl,j denote the number of active class j users in the LMR. Define a vector (7) xl = (nl,1 , nl,2 , ..., nl,J ). The admissible set in the LMR can be expressed as ⎧ ⎫ J ⎨ ⎬ XL = xl ∈ ZJ+ : nl,j cj ≤ C . ⎩ ⎭

(8)

j=1

An important physical layer QoS requirement in cellular networks for class j users is the signal-to-interference ratio, SIRj , which should be kept above the target value ωj . In this paper, we consider WCDMA cellular systems with variable spreading gain [2]. Let W denote the total cell bandwidth. The average bit rate of a class j user is Rj . The orthogonality factor is ρ. The ratio between intercell interference and total intracell power is γ. The path loss of user i is Li . Let nc,j denote the number of active class j users in the cellular network. The downlink capacity can be nc,jusing the cell load factor, evaluated J (ρ+γ)/((W/ωj Rj )+ρ). which is defined as η = j=1 i=1 The transmit power needed at the base station to guarantee the SIR requirements is PT =

Pp + PN Λ , 1−η

(9)

2428


where Pp is the power used by common control channels, PN is the background noise power. J nc,j Λ = j=1 i=1 Li /((W/ωj Rj ) + ρ). Define a vector xc = (nc,1 , nc,2 , ..., nc,J ).

(10)

The admissible set in the cellular network can be expressed as

XC = xc ∈ ZJ+ : PT ≤ PTMAX , (11) where PT is defined in (9) and PTMAX is the maximum available power of the base station. B. Joint Radio Resource Management In the interoperable networks, define the state vector of the systems as x = (xl , xc ). The state space X can be derived from (8) and (11). The state space X of the interoperable system is given as X = {x = [xl , xc ] = [nl,1 , nl,2 , . . . , nl,J , nc,1 , J 2J nc,2 , . . . , nc,J ] ∈ Z+ , : nl,j cj ≤ C, PT ≤ PTMAX }, (12) j=1

where C is the capacity in the LMR, PT is defined in (9) and PTMAX is the maximum available power of the cellular base station. For each given state x ∈ X, an action a(x) is performed by a joint radio resource management scheme. The action space is a set of all possible actions. The action is done according to a joint radio resource management scheme u ∈ U , where U is defined as U = u : X → A. (13) {x(t), u}t∈R+ is a Markov process under each radio resource management scheme. Let πu (x) denote the equilibrium probability that the system is in state x under scheme u. Define el,j ∈ {0, 1}J as a row vector containing only zeros except for the jth component, which is 1. x + (−)el,j corresponds to an increase (decrease) in the number of class j sessions in the LMR by 1. ec,j ∈ {0, 1}J is defined similarly for the cellular network. The global balance equations for the Markov Chain under scheme u are [16] J

[πu (x − el,j )λl,j aj (x − el,j ) +

j=1

πu (x − ec,j )λc,j aj (x − ec,j ) + πu (x + el,j )μl,j (nl,j + 1) +πu (x + ec,j )μc,j (nc,j + 1)] J = [λl,j aj (x) + μl,j (nl,j + 1)]πu (x) j=1

+

J

[λc,j aj (x) + μc,j (nc,j + 1)]πu (x), x ∈ X,

(14)

j=1

where λl,j , λc,j , 1/μl,j and 1/μc,j are class j sessions arrival and departure rates in the LMR and cellular network, respectively. These global balance equations can be solved using any linear equation procedure, such as Jacobi and GaussSeidel methods. Once the equations are solved, network layer

blocking probability QoS, can be directly calculated. The blocking probability for a class j session is π(i), (15) Pjb = i∈Xj

where Xj ⊆ X is the set of states that the system will move out of X with the addition of one session of class j. This approach is general enough to be applicable to a variety of radio resource management schemes. However, there is a problem with the above approach. The computation complexity of solving the global balance equations is extensive when the cardinality of the state space is large. Feasible solutions are difficult to obtain in real networks due to the problem of large dimensionality. We can consider a set of coordinate convex schemes that have a product form of the equilibrium probabilities. The coordinate convex schemes form several important resource management schemes, such as complete sharing, complete partitioning and threshold schemes. The name of coordinate convex scheme comes from the concept of coordinate convex set. A coordinate convex scheme is characterized by a coordinate convex set, which is any nonempty set Δ ⊆ X with the following property: if x ∈ Δ and nj > 0 then x − ej ∈ Δ, where x − ej corresponds to a decrease in the number of class j sessions in the system. In a coordinate convex scheme associated with coordinate convex Δ, a session arrival is admitted to the system if and only if the system state remains in Δ after the admission. The equilibrium probabilities of the system can be obtained from the theory of multiservice loss networks. ⎧ J ⎨ (λj /μj )nj π0 , if n ∈ Δ, nj ! π(n) = (16) j=1 ⎩ 0, otherwise, where π0 is a normalization constant, π0 =

1 J n∈Δ j=1

(λj /μj )nj nj !

.

(17)

IV. O PTIMAL R ADIO R ESOURCE M ANAGEMENT The above joint radio resource management schemes may not be able to maximize the radio resource utilization and guarantee the service availability and continuity QoS simultaneously. In this section, we propose an optimal radio resource management scheme that can achieve this goal. Specifically, the problem is formulated as a semi-Markov decision process (SMDP) [17]. An optimal solution can be obtained from a linear programming algorithm in this formulation. In the interoperable network, when a new or handoff session arrives, a decision must be made as to whether or not to admit and to which network (LMR or cellular network) to admit the session request based on the current state of the system. In the SMDP framework, these decision time instants are called decision epochs. The state information is the number of sessions of each class of traffic in the system. The optimality criterion for the SMDP is the long-run average reward per unit time.


A. SMDP formulations The system state vector at decision epoch t can be defined as x(t) = [xl (t), xc (t)], (18) where xl and xc are defined in (7) and (10), respectively. Similar to [18], we choose the decision epochs to be the set of all session arrival and departure instances. At each decision epoch tk , k = 0, 1, 2, ..., the network makes a decision in the time interval (tk , tk+1 ], which is referred to as an action. Action a(tk ) is defined as a(tk ) = [al,n (tk ) ∈ {−1, 0, 1}J , ac,n (tk ) ∈ {0, 1}J , al,h (tk ) ∈ {0, 1}J ],

(19)

where al,n (tk ), ac,n (tk ), al,h (tk ) are defined and interpreted as follows. 1) Define row vector al,n (tk ) = [al,n,1 (tk ), al,n,2 (tk ), . . ., al,n,J (tk )], where al,n,j (tk ) denotes the action for class j new session arrivals in the LMR area. If al,n,j (tk ) = 1, a new class j session that arrives in the LMR area is admitted to the LMR network. If al,n,j (tk ) = −1, a new class j session that arrives in the LMR area is admitted to the cellular network. If al,n,j (tk ) = 0, it is rejected. 2) Define row vector ac,n (tk ) = [ac,n,1 (tk ), ac,n,2 (tk ), . . ., ac,n,J (tk )], where ac,n,j (tk ) denotes the action for class j new session arrivals in the cellular area. If ac,n,j (tk ) = 1, a new class j session that arrives in the cellular area is admitted to the cellular network. If ac,n,j (tk ) = 0, it is rejected. 3) al,h (tk ) is defined similarly for handoff session arrivals to the cellular area. If al,h,j (tk ) = 1, a handoff class j session from the LMR to the cellular area is admitted to the cellular network. If al,h,j (tk ) = 0, it is dropped. For a given state x ∈ X, a selected action should not result in a transition to a state that is not in X. In addition, action (0, 0, ..., 0) should not be a possible action in state (0, 0, ..., 0). Otherwise, the system cannot evolve. The action space of a given state x ∈ X is defined as

where S1 = (λc,n,j ac,n,j + λl,n,j δ(−al,n,j ))τx (a), S2 = (μl,t,j + μl,h,j (1 − al,h,j ))nl,j τx (a), δ(x) = 0, if x ≤ 0 and δ(x) = 1, if x > 0. The average reward criterion is considered as the performance criterion in this paper. The blocking probability can be expressed as an average cost criterion. The reward for stateaction pair (x, a) can be expressed as r(x, a) =

+wc,n,j ac,n,j + wl,h,j al,h,j ],

B. Service Availability and Continuity Constraints The service availability (for disaster responders) constraint is that the new session blocking probability should be kept N N below a target value, Pl,j ≤ T Pl,j . The service continuity (for disaster responders) constraint is that the handoff session (from the LMR to the cellular network) blocking probability H H ≤ T Pl,j . Since should be kept below a target value, Pl,j we have derived the expected sojourn time τx (a) for a given state-action pair, the new blocking probability for class j in the LMR can be defined as the fraction of time the system N ⊂ X and the chosen action is is in a set of states Xl,n,j N in a set of actions AxN ⊂ A, where xN l,n,j ∈ Xl,n,j and l,n,j AxN = {a ∈ A : al,n,j = 0}. The above derivation follows l,n,j from the Poisson arrival see time averages (PASTA) theorem, which requires Poisson arrivals. Therefore, the blocking probability constraints in the system can be addressed in the linear programming formulation in (25) by defining cost functions related to service availability constraints, cN l,n,j (x, a) = 1 − |al,n,j |, j = 1, 2, ..., J.

if y = [xl + ej , xc ] if y = [xl , xc + ej ] if y = [xl − ej , xc + ej ] , if y = [xl , xc − ej ] if y = [xl − ej , xc ] otherwise (21)

(23)

Similarly, we can define cost functions related to the service continuity constraints, cH c,h,j (x, a) = 1 − al,h,j , j = 1, 2, ..., J.

J

λl,n,j δ(al,n,j )τx (a), ⎪ ⎪ ⎪ ⎪ S1 , ⎪ ⎨ μl,h,j nl,j al,h,j τx (a), pxy (a) = μ c,t,j nc,j τx (a), ⎪ ⎪ ⎪ ⎪ ⎪ S2 , ⎩ 0,

(22)

where wl,n,j ∈ R+ , wc,n,j ∈ R+ and wl,h,j ∈ R+ are the weights associated with class j new sessions in the LMR network, new sessions in the cellular network and handoff sessions to the cellular network, respectively.

a = (0, 0, . . . , 0) if x = (0, 0, . . . , 0)}, (20) where ej ∈ {0, 1} denotes a row vector containing only zeros except for the jth component, which is 1. xl + ej corresponds to an increase in the number of class j sessions by 1 in the LMR. xc + ej corresponds to an increase in the number of class j sessions by 1 in the cellular cell. The state transition probabilities of the embedded chain and the expected sojourn time τx (a) for each state-action pair can be used to characterize the dynamics of the system, J where τx (a) = [ j=1 (λl,n,j |al,n,j |+λc,n,j ac,n,j +μl,h,j nl,j + μl,t,j nl,j + μc,t,j nc,j )]−1 . The state transition probabilities of the embedded Markov chain are ⎧

J [wl,n,j δ(al,n,j ) + wc,n,j δ(−al,n,j ) j=1

Ax = {a ∈ Ax : al,n,j = 1 and if [(xl + ej ), xc ] ∈ X, al,n,j = −1, ac,n,j = 0 and al,h,j = 0 if[xl , (xc + ej )] ∈ X, j = 1, 2, . . . , J, and

2429

(24)

C. Linear Programming Solution The optimal policy u∗ of the SMDP is obtained by solving the following linear program. max r(x, a)τx (a)zxa zxa ≥0,x∈X,a∈Ax

subject to a∈Ay

zya −

x∈X a∈Ax

x∈X a∈Ax

pxy (a)zxa

x∈X a∈Ax

=

zxa τx (a) =

(1 − |al,n,j (x)|)zxa τx (a) ≤ x∈X x a∈A (1 − al,h,j (x))zxa τx (a) ≤ x∈X a∈Ax

0, y ∈ X 1, N T Pl,j ,

(25)

H T Pl,j .

The decision variables are zxa , x ∈ X, a ∈ Ax . The term zxa τx (a) can be interpreted as the steady-state probability of


the system being in state x and a is chosen. The first constraint is a balance equation and the second constraint can guarantee that the sum of the steady-state probabilities to be one. The new session blocking probabilities in the LMR are expressed in the second constraint to guarantee the service availability of the disaster responders. The handoff (from the LMR to the cellular network) blocking probabilities are expressed in the third constraint to guarantee the service continuity of the disaster responders. Since sample path constraints are included in (25), the optimal policy obtained will be a randomized policy: The optimal action a∗ ∈ Ax for state x is chosen probabilistically according to the probabilities τx (a)zxa / a∈Ax τx (a)zxa .

1400

1000

800

600

A. Handover Delay Improvement Fig. 3 shows the sequence numbers of packets at the corresponding time on the X axis. The MN leaves the LMR and initiates the handoff at time 3.5s. From this figure, we can see that, in the traditional SIP-based handoff scheme, the MN cannot have communication with the RN between 3.5s and 4.7s, which is unacceptable for real-time multimedia applications. This is caused by the handoff delay in the traditional SIP-based handoff scheme, in which the old session in the LMR is broken before the new session is setup in the cellular network. Comparatively, it is evident that the seamless-handoff scheme supports seamless handoff between these two networks. During the handoff process, the MN attaches to the cellular network, obtains a new domain/IP

Handoff delay

400

New connection established 200

V. N UMERICAL R ESULTS AND D ISCUSSIONS

0

Fig. 3.

Activate handoff

1

2

Handoff to cellular

3

4

Time (s)

5

6

7

Packet sequence number in different schemes.

1500

Average handoff delay (ms)

In this section, numerical examples are used to illustrate the performance of the proposed schemes and simulations have been used to validate our model. We compare the handoff delay in the novel S-SIP scheme with that of existing general SIP-based handoff scheme. We show that the handoff delay is very close to zero, which means that most handoffs are seamless. In addition, we also compare the proposed interoperable system with the system in which the LMR is not interoperable with the cellular network. We show that the proposed interoperable system can significantly improve the service availability and continuity QoS for disaster responders. We also show that the scheme can guarantee the QoS constraints by keeping new service blocking and handoff dropping probabilities below the target values. To guarantee the QoS requirements, some bandwidth should be reserved in the cellular network. The optimal reserved bandwidth will also be given in this section. We consider a LMR/cellular interoperable system with a single LMR cell and a single cellular CDMA cell. One class of video traffic is considered. The data rate of each video flow is 64 Kbps. We assume the capacity of the LMR is between 144384 Kbps. The numerical values for the system parameters are given in Table I. We set the transmission delays over the LMR wireless link and the Internet as variable values. The new session arrival rate in the system is λn = λl,n + λc,n , where λl,n and λc,n are new session arrival rates in the LMR and the cellular network, respectively. In the numerical examples, μc,t = 0.005, μl,t = 0.002, μl,h = 0.002 and wl,n = wc,n = wl,h = 1.

New S−SIP scheme in LMR New S−SIP scheme in cellular Traditional SIP scheme in LMR Traditional SIP scheme in cellular

1200

Packet sequence number

2430

1000 Traditional SIP scheme (Internet delay = 150 ms) Traditional SIP scheme (Internet delay = 50 ms) New S−SIP scheme (Internet delay = 150 ms) New S−SIP scheme (Internet delay = 50 ms) 500

0 50

Fig. 4.

55

60

65

70

75

80

85

90

95

Transmission delay over the LMR network (ms)

100

Handoff delay in different schemes.

address and joins the session with the new address before the old session with the LMR is terminated. No packets are lost during this seamless handoff procedure. Different network configurations may have effects on the handoff delay performance. We set the transmission delays over the Internet as 50ms and 150ms. Fig. 4 shows the handoff delay with different transmission delays over the LMR. We can see that the handoff delay in the proposed scheme is very close to zero, which means that most handoffs are seamless independent of the network configurations. In contrast, the transmission delay over the Internet has effects on the handoff delay in the traditional SIP-based scheme. The larger the Internet transmission delay, the larger the handoff delay. B. Radio Resource Utilization, Service Availability and Continuity QoS Improvement Fig. 5 shows the radio resource utilization in different schemes. In this example, we assume that 40% of the total new session arrivals occur in the LMR area. We can observe that the radio resource utilization can be increased significantly in the proposed interoperable system compared to the existing


Parameter target SIR for video traffic bandwidth in the cellular network orthogonality factor intercell/intracell ratio data rate for video traffic common control channels power background noise power maximum base station power capacity in the LMR processing delays processing delays processing delays processing delays transmission delay over the LMR transmission delay over the cellular network transmission delay over the Internet

Notation ω W ρ γ R Pp PN PTM AX C DP,M N DP,RN DP,CSCF DP,DHCP DT,L DT,C DT,I

2431

Value 8 dB 3.84M 0.4 0.55 64 Kbps 33 dBm -106 dBm 43 dBm 144-384 Kbps 20 ms 20 ms 20 ms 20 ms 50-100 ms 30 ms 50-200 ms

TABLE I S YSTEM PARAMETERS .

0.7

Network utilization

0.6 0.5 0.4 0.3 0.2 Proposed scheme (Analysis) Proposed scheme (Simulation) Existing scheme (Analysis) Existing scheme (Simulation)

0.1 0 0.01

0.02

0.03

0.04

0.05

0.06

New session arrival rate (sessions/second) (λ

p,n

Fig. 5.

0.07

+λ )

Service availability and continuity QoS

0.8 0

10

−1

10

−2

10

−3

10

−4

10

−5

10

Handoff dropping probability in the existing scheme New session blocking probability in the existing scheme Handoff dropping probability in the proposed scheme New session blocking probability in the proposed scheme

−6

10

−7

10

0.01

0.02

0.03

0.04

non-interoperable one. We can also see that the simulation results roughly agree with those from the analysis, which is true for all of the following examples. For clearness of the presentation, we omit the simulation results in the following figures. Fig. 6 shows the service availability and continuity QoS in different schemes. We can see that for new sessions from disaster responders, the blocking probability in the proposed scheme is significantly less than that in the existing scheme, in which the LMR and the cellular network are not interoperable. A similar observation is true with the handoff dropping probability of disaster responders. In the existing scheme, when a disaster responder moves out of the LMR with an ongoing session, the ongoing session must be dropped. In contrast, in the proposed scheme, the ongoing session can be handoffed to the cellular network. The service availability and continuity QoS can be improved significantly in the proposed scheme compared to the existing scheme. If there are 30% of the total new session arrivals occuring in the LMR area, we have the similar results, which are shown in Fig. 7. In the following examples, we assume that 40% of the total new session arrivals occur in the LMR area.

0.06

p,n

c,n

Radio resource utilization.

0.05


0.07

+λ ) c,n

Fig. 6. Service availability and continuity QoS without QoS constraints (40% of the total new session arrivals occur in the LMR area).

C. Guaranteed Service Availability and Continuity QoS The proposed scheme can also guarantee the QoS constraints, which is shown in Fig. 8. In this example, the new session blocking probability QoS constraint is 3% for disaster responders, which means the new session blocking probability for disaster responders cannot exceed 3%. The handoff dropping probability constraint is 0.5%. From Fig. 8, we can see that the proposed scheme can always guarantee the QoS constraints with a variety of traffic loads. This is achieved by reserving some bandwidth in the cellular network for disaster responders. D. Optimal Reserved Bandwidth The optimal reserved bandwidth can be obtained by solving the linear program (25). Fig. 9 shows the optimal bandwidth that needs to be reserved to guarantee the QoS requirements for disaster responders. Some bandwidth is reserved exclusively for disaster responders. Some bandwidth is reserved partially for them. For example, when the new session arrival rate is 0.06, the new session arrivals from the cellular network will be rejected when there are 8 (or 9 and 10) users in the

2432


10 0

9

Bandwidth in the cellular network


10

−1

10

−2

10

−3

10

−4

10

−5

10


−6

10

−7

10

0.01

0.02

0.03

0.04

0.05

0.06


p,n

+λ )

0.07


7

6

5

4 Reserved bandwidth for first responders Partially reserved (with probabilities) bandwidth for first responders Free bandwidth for commercial users

3

2

1

0 0.025

0.03

0.035

0.04

0.045

0.05

0.055

0.06

0.065

0.07

New session arrival rate (sessions/second) (λp,n + λc,n)

0.075

c,n

Fig. 7. Service availability and continuity QoS without QoS constraints (30% of the total new session arrivals occur in the LMR area).

Fig. 9. Optimal reserved bandwidth in the cellular network to guarantee the QoS constraints for disaster responders.

optimal scheme were presented. Numerical examples were used to show the performance of the proposed schemes. We have shown that the proposed schemes can significantly improve the service availability and continuity QoS for disaster responders. Further study is in progress to consider other QoS requirements, such as packet delay and loss, in the interoperable cellular/LMR wireless networks. The proposed schemes are not limited to the interoperability problem in cellular/LMR networks. Therefore, it is interesting to apply the schemes to other heterogeneous wireless networks.

0

10

−1

10

−2

10

−3

10

−4

10

−5

10


−6

10

−7

10

8

0.01

0.02

0.03

0.04

0.05

0.06

0.07

New session arrival rate (sessions/second) (λp,n + λc,n)

Fig. 8. Service availability and continuity QoS (New session blocking probability QoS constraints: 3%; Handoff dropping blocking probability QoS constraint: 0.5%).

cellular network; whereas session arrivals from the LMR will always be accepted whenever some bandwidth is available. When there are 7 users in the cellular networks, the new session arrivals from the cellular network will be accepted with a probability 0.1129. This randomized policy is due to the QoS constraints, which is explained in Subsection IV-C. VI. C ONCLUSIONS AND F UTURE W ORK We have studied the interoperability problem in public safety LMR networks and commercial cellular networks for disaster response. The interoperability can be enhanced by using SIP and a joint radio resource management framework. In addition, we have formulated service availability and continuity QoS for disaster responders as new session blocking probability and handoff dropping probability constraints, respectively. We have presented an optimal joint radio resource management scheme in the interoperable system to maximize the overall radio resource utilization while guaranteeing the QoS constraints. A semi-Markov decision process formulation and linear-programming-based algorithms for computing the

Acknowledgment

We thank the reviewers for their detailed reviews and constructive comments, which have helped to improve the quality of this paper. R EFERENCES [1] TIA, “Apco project 25 system and standards definition," TIA/EIA Telecomm. Syst. Bull., TSB102-A 1995. [2] H. Holma and A. Toskala, WCDMA for UMTS: Radio Access for Third Generation Mobile Communications. New York: Wiley, 2004. [3] J. Gozalvez, “HSDPA goes commercial," IEEE Veh. Technol. Mag., vol. 1, pp. 43-53, Mar. 2006. [4] S. Frattasi, H. Fathi, F. Fitzek, R. Prasad, and M. D. Katz, “Defining 4G technology from the user’s perspective," IEEE Network, Jan. 2006. [5] K. Balachandran, K. C. Budka, T. P. Chu, T. L. Doumi, and J. H. Kang, “Mobile responder communication networks for public safety," IEEE Commun. Mag., pp. 56-64, 2006. [6] A. D. Assouma, R. Beaubrun, and S. Pierre, “Mobility management in heterogeneous wireless networks," IEEE J. Select. Areas Commun., vol. 24, pp. 638-648, Mar. 2006. [7] E. Stevens-Navarro, Y. Lin, and V. W. S. Wong, “An MDP-based vertical handoff decision algorithm for heterogeneous wireless networks," vol. 57, pp. 1243-1254, Mar. 2008. [8] W. Song, H. Jiang, and W. Zhuang, “Performance analysis of the WLAN-first scheme in cellular/WLAN internetworking," vol. 6, pp. 1932-1952, May 2007. [9] W. Song, Y. Cheng, and W. Zhuang, “Improving voice and data services in cellular/WLAN integrated networks by admission control," vol. 6, pp. 4025-4037, Nov. 2007. [10] T. Yagyu, M. Jibiki, and K. Yoshida, “SMART: scalable mobility adaptive routing techniques for heterogeneous mobile networks," in Proc. IEEE WCNC’07, 2007. [11] J. Rosenberg and et al., “SIP: session initiation protocol," IETF RFC 3261, June 2002. [12] W. Wu, N. Banerjee, K. Basu, and S. K. Das, “SIP-based vertical handoff between WWANs and WLANs," IEEE Wireless Commun., vol. 12, pp. 66-72, June 2005.


[13] G. Camarillo and M. A. Garcia-Martin, The 3G IP Multimedia Subsystem (IMS): Merging the Internet and the Cellular Networks. Wiley, 2004. [14] C. E. Perkins, IP mobility support for IPv4. RFC 3220, Jan. 2002. [15] L. Ma, F. Yu, and V. C. M. Leung, “A new method to support UMTS/WLAN vertical handover using SCTP," IEEE Wireless Commun., vol. 11, pp. 44-51, Aug. 2004. [16] K. Ross, Multiservice Loss Models for Broadband Telecommunication Networks. Springer-Verlag, 1995. [17] M. Puterman, Markov Decision Processes. John Wiley, 1994. [18] F. Yu, V. Krishnamurthy, and V. C. M. Leung, “Cross-layer optimal connection admission control for variable bit rate multimedia traffic in packet wireless CDMA networks," IEEE Trans. Signal Processing, vol. 54, pp. 542-555, Feb. 2006. F. Richard Yu (S’00-M’04-SM’08) received the PhD degree in electrical engineering from the University of British Columbia (UBC) in 2003. From 2002 to 2004, he was with Ericsson (in Lund, Sweden), where he worked on the research and development of 3G cellular networks. From 2005 to 2006, he was with a start-up in California, USA, where he worked on the research and development in the areas of advanced wireless communication technologies and new standards. He joined Carleton School of Information Technology and the Department of Systems and Computer Engineering at Carleton University, Canada, in 2006, where he is currently an Assistant Professor. His research interests include cross-layer design, QoS provisioning and security in wireless networks. He has served on the Technical Program Committee (TPC) of numerous conferences and as the TPC Co-Chair of IEEE IWCMC’2009, VTC’2008F Track 4, WiN-ITS’2007. He is a senior member of the IEEE. Jie Zhang received the B.S. degree in computer science and technology from Nanjing University, Nanjing, China in 2001 and M. Phil in computing from The Hong Kong Polytechnic University, Hong ong in 2004. Since 2004, she has been working toward the Ph.D degree in Electrical and Computer Engineering from the University of British Columbia, Canada. Her area of research include the mobility management over the 4-th Generation wireless networks, the IP Multimedia Subsystem and electronic commerce. She has published more than a dozen of papers on networks and electronic commerce area.

2433

Helen Tang received her Ph.D. degree in the Department of System and Computer Engineering at Carleton University, Ottawa, Canada in 2005. From 1999 to 2005, she had worked in a few R&D organizations in Canada and USA including AlcatelLucent, Mentor Graphics and Communications Research Center Canada. In Oct. 2005, she joined Network Information Operations Section at Defence R&D Canada as a Defence Scientist. She is a member of IEEE. She has published more than 20 research papers in international journals and conferences including IEEE T RANSACTIONS ON W IRELESS C OMMUNICATIONS, J OURNAL OF S ECURITY AND C OMM . N ETWORKS , IEEE ICC, IEEE VTC, IEEE Milcom, and IEEE Globecom. She has served as reviewer, session chair and technical committee member for various conferences. Her research interests include ad hoc and sensor networks, wireless network security, communication protocols and performance analysis. Henry C. B. Chan received his BA and MA degrees from the University of Cambridge, England and his PhD degree from the University of British Columbia, Canada. From October 1988 to October 1993, he worked with Hong Kong Telecommunications Limited primarily on the development of networking services in Hong Kong. Between October 1997 and August 1998, he worked with BC TEL Advanced Communications on the development of high-speed networking technologies and ATM-based services. Currently, he is an associate professor in the Department of Computing at The Hong Kong Polytechnic University. His research interests include networking/communications, wireless networks, Internet technologies and electronic commerce. He has authored/co-authored a textbook on electronic commerce, three book chapters and over 60 journal/conference papers. Dr. Chan is a member of IEEE, ACM and IET. He is currently serving as an executive committee member of the IEEE Hong Kong Section Computer Chapter. Victor C. M. Leung (S’75-M’89-SM’97-F’03) received the B.A.Sc. (Hons.) degree in electrical engineering from the University of British Columbia (U.B.C.) in 1977, and was awarded the APEBC Gold Medal as the head of the graduating class in the Faculty of Applied Science. He attended graduate school at U.B.C. on a Natural Sciences and Engineering Research Council Postgraduate Scholarship and obtained the Ph.D. degree in electrical engineering in 1981. From 1981 to 1987, Dr. Leung was a Senior Member of Technical Staff at MPR Teltech Ltd., specializing in the planning, design and analysis of satellite communication systems. In 1988, he was a Lecturer in the Department of Electronics at the Chinese University of Hong Kong. He returned to U.B.C. as a faculty member in 1989, where he is a Professor and holder of the TELUS Mobility Research Chair in Advanced Telecommunications Engineering in the Department of Electrical and Computer Engineering, and a member of the Institute for Computing, Information, and Cognitive Systems. His research covers broad areas in wireless networks and mobile systems, and has published more than 380 journal and conference papers in these areas. Dr. Leung is a Fellow of IEEE, a Fellow of the Engineering Institute of Canada, a Fellow of the Canadian Academy of Engineering, and a voting member of ACM. He serves on the editorial boards of several journals, including the IEEE T RANSACTIONS ON W IRELESS C OMMUNICATIONS and IEEE T RANSACTIONS ON C OMPUTERS . He has served on the Technical Program Committee (TPC) of numerous conferences. He was the TPC-Vice Chair of IEEE WCNC 2005, General Co-chair of ACM/IEEE MSWiM 2005, General Chair of QShine 2007, and TPC Chair of the wireless networks and cognitive radio track of IEEE VTC-fall 2008.