can provide transmission rates of about 54 Mbit/s [3]. These systems operate ..... controlled by the MHN, and the available transmission rate can be increased if ...
Hierarchical Cellular Multihop Networks M. Lott, M. Weckerle, W. Zirwas, H. Li, E. Schulz Siemens AG, Information and Communication Mobile, Gustav-Heinemann-Ring 115, 81730 Munich, Germany Abstract-Mobile communication systems beyond 3G will comprise cellular and self-organizing networks. In this paper an appropriate migration path towards the convergence of these two network types is highlighted and the resulting common radio network architecture is presented. By means of example scenarios the way to beneficially combine both network types will be described. Thereby, transport oriented aspects, especially resource allocation, multihop communication, routing, and special frequency reuse will be the focus of attention. Moreover, it will be described how other network types, like satellite systems, will complement and support cellular and self-organizing networks. Furthermore, we will present results of our research and show the challenges to be met in the design of future radio architectures.
and interference between such systems is difficult to predict and to control. Taking into account the advantageous, potentials and drawbacks of cellular network, WLAN, and self-organizing network architectures with respect to, e.g., coverage, capacity, mobility, cost of infrastructure, and flexibility, it becomes obvious that a combination of them is the logical consequence for nextgeneration network concepts. In this paper a wireless network architecture, called hierarchical cellular multihop network, is presented in Section 2. Based on this architecture in Section 3 technical issues are discussed. Introducing different example scenarios and challenges, the solutions and benefits from the proposed architecture will be discussed. Section 4 concludes the paper summarizing the benefits and potentials and gives an outlook on future research topics.
1. Introduction Future wireless communication will rely on different types of radio systems, for example satellite networks, high altitude platforms (HAPs), cellular networks (CN), point-to-multipoint (PMP) access systems, wireless local area networks (WLAN), and personal area networks (PAN). These systems can be organized in a layered structure and facilitate an optimized design for different application areas [1]. They serve to wirelessly extend the fixed network infrastructure, which is most likely based on IP protocol on the network layer, with the same and new services. The most common and widely used wireless access systems today are cellular radio systems, based on GSM, GPRS and in the near future on UMTS standards, and the recently deployed WLAN standards, e.g. IEEE802.11 and HIPERLAN/2 [2]. Besides an infrastructure mode for highspeed access in hot-spot areas to connect to the Internet, like in airports and hotels, WLAN also provide a self-organizing mode for communication between devices in a spontaneous manner, also referred to as ad hoc mode. In such networks nodes can communicate without the need of an infrastructure with each other by using intermediate nodes as relays, resulting in multihop communication. Unlike systems providing an ad hoc mode, cellular systems rely on an infrastructure of base stations (BS) and require network-planning and operation in licensed radio spectrums. For data rates up to 2Mbit/s UMTS provides an ideal platform. However, this cumulative data rate is still not enough for hot-spot areas where the number of mobile nodes (MNs) per area is very high. To increase the individual data rate of users, WLAN systems are introduced at these places, which can provide transmission rates of about 54 Mbit/s [3]. These systems operate in unlicensed radio spectrums and, generally, offer mobile data communications with very low cost. Nevertheless, transmission power in such kind of communication systems is limited, and hence the coverage is,
2. Hierarchical Cellular Multihop Network Communication systems beyond 3rd generation will comprise different types of wireless networks that partly cover the same area and complement each other. An example for such system architecture can be seen in Figure 1, which is called hierarchical cellular multihop network (HCMN). Example servi ce:
Example system:
Location service
S-UMTS
AAA, routing support
GSM / GPRS BS
VHO
HHO Resource allocation
MHN coverage MN
Multihop communication AP
BS: AP: VHO:
UMTS
BS
VHO
Bse Station Access Point Horizontal Handover
MN MHN MN: MHN:
H/2 with EPs
Mobile Node Multihop capable Node VHO: Vertical Handover
Figure 1. Hierarchicel system architecture An overlaying cellular mobile radio system, e.g. a 3G system, forms the basis of the proposed network architecture. The possible connection of each MN to a BS guarantees full coverage and this connection can always be taken as a fallback solution in the case that a MN looses connection to any other network it might be connected to. This requires interoperability of existing networks and future networks and the support of vertical handover (VHO), i.e. handover between different
wireless access networks (inter-system handover) [2]. The BS provides access to the backbone, which is most likely to be based on TCP/IP protocol suite [4]. In order to satisfy the increasing demand of higher data rates in hot-spot areas WLAN systems allow a broadband radio access to the Internet provided by access points (APs). The next evolutionary step towards a hierarchical multihop network structure is to introduce multihop capable nodes (MHN), which can be fixed or even mobile, see Figure 1. With the fixed MHNs the coverage of the APs can be extended. At the same time fixed MHN can be connected to a power supply to offer more potent services. Sub-cells can be established in a self-organizing manner. This means the MHN recognizes its AP and takes over control functionality within the sub-cell. A typical control function comprises the management of the medium access within the sub-cell. Furthermore, it provides the connections between MNs in the sub-cell, which can directly communicate with each other by means of a directmode. Moreover, the MHN and the MNs will be provided by signaling information by the overlaying cellular network, e.g., routing information that are provided by a BS. If geographic routing is performed the location information of an MN can be derived by a satellite system. Next to these exemplary services offered by the different hierarchies as shown in Figure 1, each layer can provide user data delivery, which is important for full coverage in areas where the underlying architecture is not deployed. Besides fixed MHNs also mobile MHNs are considered in a further evolutionary step. In the case that the required data throughput cannot be provided any more by the AP or fixed MHN, due to increasing penetration and/or to satisfy the future demands for packet-data services, a MN can become a MHN and establishes a sub-cell on demand. These cells can use the same or different frequencies. Independent on the fact whether a MHN is fixed or mobile it can act as a wireless gateway to the wired infrastructure for the MNs of the sub-cells. Furthermore, 3G systems can be used to provide signaling information for the MNs and MHNs.
3. Technical Issues of Multihop Networks Several benefits and advantageous over existing network architecture can be imagined, which can be made up with the novel proposed HCMN. However, these benefits have to be paid by the adaptation of existing and the development of new concepts and partly by more complex algorithms and terminals. Nevertheless, with the following technical challenges and examples it will be shown, that with the continuous development of the existing network structures a smooth migration path towards the HCMN can be realized with the exploration of already existing infrastructure.
3.1 Extending the Coverage with Multihop Current research efforts concentrate on the investigation of multihop communication and several results indicate that multihop provides a potential for enhanced coverage [5][6][7][8]. These approaches assume that multihop communication over mobile terminals with relay capabilities mitigate typical problems of cellular networks, i.e. the existence of deadspot locations, e.g. due to shadowing of buildings. Different from the concept of mobile relay stations, in the
EPMCC03_HCMN_v03.doc
COVERAGE project it is proposed to extend the coverage by means of fixed MHN, so called extension points, EPs. They are installed at certain locations and serve to extend the infrastructure in a flexible manner. A typical scenario is shown in Figure 2. EP EP
AP1
IP-based network
EP
Public places AP2 AP: Access Point EP: Exten sion Point
Figure 2. Typical hot spot scenario for public places with fixed MHNs (EP) The APs have direct connection to the IP-based backbone network, while the EPs are situated at planed locations according to the actual environment. It is assumed mainly traffic of MNs into/from the wired backbone- similar to the situation of e.g. a cellular GSM network. In case of traffic between MNs in the range of an MHN this could be handled by direct links nonetheless. For the broadband wireless Internet access, HIPERLAN/2 (H/2) is envisaged, which supports aggregate data rates up to 54Mbit/s. Not shown in Figure 2 is the cellular overlay network, which becomes essential for full coverage, since in the beginning full coverage is not possible with EPs. For rapidly and smooth deployment into an existing infrastructure and running system, it is assumed that MNs are standard conform, i.e. in the case of COVERAGE conform to H/2. The MN should not see any difference whether connected to an AP or MHN. This allows users to roam between office, home- and public environments. For the MHN it means it has to support AP functionality plus additional multihop functions. The MHN is not a simple direct repeater but has additional intelligence and processes the forwarding data in the base band. The planned and fixed locations of the MHNs guarantees full coverage within the area of the HCMN. In contrast mobile MHNs might be switched of or the owner of the MHN might despise relaying data for other users, e.g. because his battery will be emptied. Thus, planing guarantees that the MHNs are situated at locations, where they yield the highest benefit. In Figure 3 the gain in data rate by the use of MHNs for the inner city of Munich can be seen. The colors indicate the additional data rate, which reaches typically values between 0.5 and 3 Mbit/s. The location of EP2 has been chosen, so that it transmits it’s power mainly into side streets. The simulation has been done with the ray-tracing tool WINPROP. The mentioned SFN concept (Single Frequency Network) is beyond the scope of this paper (see [10]).
Gain in Netto Data Rate (with SFN and EG predist.)
( c)
7
bit/s x 10 5.5
Ptx N
5
0 MH
= æç1 + è
( c ) ( ) ö÷ø ( ) − ( ) , ( 2 ) Ptx N
n
1 γ 0 one d
d γ n
d γ n
4.5 4 3.5 3 2.5
EP2
whereas the constant c0 takes into account the wavelength and the units. The resulting SNR at the transmitter for the MH connection over the respective SNR value for a one-hop connection is shown in Figure 4.
2 1.5
AP
γ= 2
1
EP1
2
3
4
0.5
n=2
0
Figure 3: Gains in data rates with forwarding and SFN concept compared with a single AP
3.2 Transmit Power and Capacity One crucial topic in future radio systems is the required transmit power of mobile nodes and at the same time the stand-by time and emission. With multihop connections the required transmit power can be considerably reduced, since several low-power links are established to connect arbitrary end-user-nodes with each other over several intermediate nodes [8]. At the same time interference is reduced, which in turn provides potentials for capacity increase. E.g., it has been demonstrated for H/2 that under special constellations capacity can be increased with multihop connections [11]. Different from these results in [8] it has been shown for a CDMA-based scheme that capacity decreases considerably as a result of the relaying option. For the direct link to the BS only a single resource is needed, whereas the use of relaying requires several resources, which are related to the number of individual links. Hence, it is worth mentioning that relaying requires more transmissions and receptions, which might end up in higher interference, respectively power consumption. Moreover, it introduces additional delays. A capacity bottleneck occurs especially near the BS, where often only a few nodes are used to relay all packets in the cell. However, in [8] it has not been considered link adaptation and the authors expect to increase capacity with multihop when a reuse of resources within a cell is considered. To come to a more general view of the potential gains of multihop communication the Shannon capacity for a one and an n-hop connection is examined. On each link of the multihop (MH) connection the data rate has to be n times higher than the data rate of the corresponding one-hop connection to end-up in the same end-to-end throughput. The required SNR for the multihop connection over an AWGN channel for a given bandwidth, W, noise power, N, and received signal power, Prx, becomes
( )
Prx N MH
( ( ) ) −1 ( 1 )
= 1+
n Prx N one
On the one side, the required transmit power has to be increased to achieve the same data rate, but on the other side the distance on the individual links decrease with an increasing number of hops. To incorporate this dependency in the capacity calculation, the receive power is substituted by the transmit power, Ptx, and the distance, d,
EPMCC03_HCMN_v03.doc
(Ptx/N c0)OH in [dB]
n=4 d = 200 γ
d = 100
γ=2 γ=3 γ=4
(Ptx/N c0)one in [dB]
Figure 4: Required signal-to-noise ratio at the transmitter for one-hop and multihop connections It can be recognized that the gain for a multihop connection strongly depends on the transmit power, the distance between source and destination, the attenuation coefficient, γ, and the number of hops, denoted by n. The smaller the transmit power, the larger the distance, and the stronger the attenuation, the more attractive becomes the introduction of intermediate hops. It can further be seen that with increasing number of hops the breakeven point is shifted to smaller SNR values and the gradient increases beyond this point. Nevertheless, high gains can be obtained under high attenuation, e.g. approx. 10dB / 20dB for a γ°=°4 and n°=°2 / n°=°4 hops, respectively. Nevertheless, it is noteworthy that with the introduction of intermediate nodes the attenuation coefficient will most probably decrease, which in turn reduces the potential benefit of relaying, not mentioning the additional delay and protocol overhead. All the aforementioned facts lead us to the conclusion that a number of 4 hops should not be exceeded to benefit from relaying. Different from the concept of multihop in ad hoc networks in HCMN a reduction of signaling overhead by means of centralized management of the system is envisaged, which in turn increases the system capacity. As already demonstrated in the previous sections, multihop communication over fixed MHN allow also to increase capacity. And power consumption becomes less critical with fixed MHN connected to power supplies. But also mobile MHN provides potentials for capacity enhancements. Moreover, with intelligent scheduling and resource allocation the frequency reuse distance can be decreased. The spectral efficiency of the system can be enhanced when reusing the same frequency in different sub-cells controlled by the MHN, and the available transmission rate can be increased if using different frequencies, whereas the
coordination of the resources is beneficially managed at the BS in the overlay network. In line with central coordination of multihop communication and reuse of resources in a cell, within the MIND project the benefits of multihop communication are investigated [16]. Specifically, it is proposed to simultaneously assign the same resource to separated user groups. Thereby, the direct-mode is explored, which is supported by the extension of the H/2 standard for the home-environment [18]. The concurrent scheduling of resources under the control of the AP is referred to as smart direct-link (SDiL) appliance [17]. With the SDiL mobile MHN are introduced between an MN and the AP, which allows to considerably increase the data rate for high attenuation values and for large distances between the MN and the AP.
3.3 Routing in Multihop Wireless Networks In the MANET sub-group of IETF, several routing algorithms for mobile ad hoc networks such as DSDV, AODV, DSR and TORA have being investigated. An assumption of the MANET group is that no infrastructure exists in multihop networks, hence the network management, radio resource management and routing algorithm are performed among mobile nodes in a selforganised and distributed way. Earlier study of those algorithms has shown high overhead of radio transmission protocols and low efficiency in network throughput. As studied by Perkins [15], in a medium size, moderate loaded ad hoc network, generally more than 80% of the transmission capacity is used for routing, medium access and radio link control protocol packets. Based on this observation, a hybrid routing scheme under the HMCN architecture is studied within IPonAir1, next to radio resource management schemes. The proposed hybrid routing scheme works as follows: A mobile node, say C, wants to access the Internet and sends a route request (RREQ) to the control centre in the cellular network. After searching the location controller (LC), which contains information of locality and neighbourhood of each mobile node, and the service controller (SC), which provides the registered AP provider, the control centre finds a route. This route comprises one series of mobile nodes that are willing to relay the data traffic between mobile node C and an access point to the Internet. The BS of the cellular network then unicasts the route information (route reply, RREP) asking each mobile node along the path to establish a logical data connection. In this case, the service and route discovery is done in the cellular network and data packet transmissions are done in the multihop network. In this way we add the value of cellular networks on the one side, and on the other side the efficiency of multihop transmissions are significantly improved for the routing. A cellular based routing algorithm has been compared with ad-hoc routing algorithms DSR and AODV [15] in a simple multihop cellular network topology, where all mobile nodes are uniformly distributed inside the cell, and each one has the same distance to its six neighboring nodes. Hence the nodes form a central symmetric hexagon. It is assumed that half of the nodes on the border have data packets destined for its central symmetric counterparts. The 1
IPonAir is a research project partially funded by the German Government and has started in November 2001.
EPMCC03_HCMN_v03.doc
distance between two neighbor nodes is r, where r is set up in such a way that a node can communicate with its direct neighbors, but not further neighbors. The decoding radius of the base station is R=N*r, which covers all MN under investigation. In addition, every t seconds the route can brake with a probability p, which invokes a route maintenance process. The cellular based algorithm used here is similar with one in [6]. In Table 1 the average number of routing packets per connection per second produced by the three routing algorithms as outcome of event-driven simulations are given. It is assumed that there is no collision during radio transmissions. Table 1. Comparison of routing protocol overhead P=0.5, t=20
N=3
N=9
Low mobility
Unicast
Broadcast
Unicast
Broadcast
DSR
0.34
0.145
0.97
0.93
AODV
0.225
0.875
0.675
6.725
Cellular based
0.625
0.525
0.625
0.525
P=0.7, t=10
N=3
N=9
High mobility
Unicast
Broadcast
Unicast
Broadcast
DSR
1.70
0.95
5.04
6.76
AODV
0.63
2.45
1.89
18.83
Cellular based
0.85
0.57
0.85
0.57
The AODV algorithm can always find the best route. The DSR algorithm can find several routes, which have the least number of hops. Note that the protocol overhead is divided into unicast and broadcast, because both might cause different overheads of radio transmission layers. The results show that AODV sends much more broadcast signaling packets than DSR, while DSR sends much more unicast signaling packets than AODV. This phenomenon is caused by using different cache mechanisms and has been also confirmed in [15]. The cellular based routing protocol has the best performance and excellent scalability in dynamic multihop networks, because the protocol overhead is not affected by the network size and is not heavily influenced by high mobility.
3.4 Adaptive Antennas in Mobile Multihop Networks With the increasing demand for higher data rates the concepts of adaptive antennas are proposed to increase the individual user data rate. The use of adaptive antennas allow to reduce the interference and improve the link qualities between the AP and the MNs which allow the use of higher data rates in order to improve the throughput on the links between the MNs and the AP. One basic challenge with these concepts are the determination of the radio channel and interference of other nodes being active at the same time. This becomes even worse when all stations move in an unpredictable manner and have to provide connections between each other, as this is the case in typical mobile ad hoc networks.
When establishing sub-cells as proposed in HCMN, each subcell can be controlled by an MHN, which can be fixed or mobile, and only one MHN directly communicate with the AP, which reduces the number of communication links at the AP significantly compared to conventional cellular systems. When reducing the number of communication links between MNs and the AP one can expect that in the azimuth the spatial distances between impinging waves from different MHNs is larger than the spatial distances of the impinging waves originating from all MNs. Due to the reduced number of impinging wave fronts the use of adaptive antennas at the AP is more profitable [13]. Furthermore, the use of adaptive antennas might allow a simultaneous allocation of resources for different links between the AP and the MHNs and MNs in the sense of space division multiple access (SDMA) [14]. In SDMA the interfering signals are suppressed due to a specific shape of the antenna diagram. In this case the number of antenna elements used at the AP must be larger than the number of propagation paths that should be suppressed. With a smaller number of existing communication paths at the AP in HCMN than in cellular networks the number of antenna elements can be reduced, thus, leading to reduced costs for the implementation of the adaptive antennas.
Editorial, IEEE Personal Communications, vol. 8, no. 5, October, 2001.
[3] A. Doufexi, S. Armour, M. Butler, A. Nix, D. Bull, "A Study of the Performance of HIPERLAN/2 and IEEE802.11a physical layers," IEEE VTC´01, Rhodes, Greece, May 6-9, 2001.
[4] M. Annoni, R. Hancock, T. Paila, E. Scarrone, R. Tönjes, L. Dell'Uomo, D. Wisely, R. Mort, “Radio Access Networks beyond the 3rd Generation: A First Comparison of Architectures from 4 IST Projects, “ in Proc. of PIMRC´01, San Diego, USA, Sep. 30 – Oct. 3, 2001.
[5] Y.-D. Lin and Y.-C. Hsu, “ Multihop Cellular: A New Architecture for Wireless Communications”, in IEEE INFOCOM, 2000.
[6] R. Ananthapadmanabha, B.S. Manoj and C. Siva Ram Murthy, “Multi-hop Cellular Networks: The Architecture and Routing Protocols”, Proc. PIMRC, 2001.
[7] 3GPP, Opportunity Driven Multiple Access,” Technical Specification Group Radio Access Network, 3G TR 25.924, Dec. 1999.
[8] O.C. Mantel, N. Scully, and A. Mawira, “Radio aspects of
4. Conclusion Within this paper the hierarchical cellular multihop network (HCMN) architecture for future generation wireless systems is presented. It takes into account the advantageous, potentials and drawbacks of cellular network, WLAN, and self-organizing network architectures, which are expected to be present in next generation systems. Taking into account the typical challenges and demands of next generation wireless systems, e.g., coverage, capacity, resource management, and routing, these different technical issues are highlighted and the benefits of the proposed architecture are described and exemplary demonstrated in example scenarios and research projects. Specifically, the performance gain of multihop communication is studied. From the results it can be concluded that introducing fixed multihopcapable nodes (MHNs) increases coverage, and that mobile MHNs offer the potential to increase capacity and decrease power consumption. Furthermore, with the introduction of the hierarchical structure the different demands for network operation can be beneficially served at that level where the service can be most efficiently provided. Thereby, existing and future networks are combined in an appropriate manner that allows a smooth evolution towards a powerful next-generation mobile system. Besides the advantageous of the proposed concepts there are further potential benefits, like low installation costs, ease of deployment, simple centrally controlled security schemes, etc, which have not been discussed because of space reasons, and which make Hierarchical Cellular Multihop Networks a very promising candidate as future radio network architecture.
Reference [1] Wireless World Research Forum (WWRF), Book of Visions, http://www.wireless-world-research.org/.
[2] A. Jamalipour, and S. Tekinay, “Fourth Generation Wireless Networks and Interconnecting Standards,” Guest
EPMCC03_HCMN_v03.doc
hybrid wireless ad-hoc networks,” in Proc. of IEEE VTC´01, Rhodes, Greece, May, 2001.
[9] W. Zirwas, et. al., “Broadband multihop networks with reduced protocol overhead,” in Proc. European Wireless 2002, Florence, February, 2002.
[10] W. Zirwas, et. al, “Synchronization of single frequency network,” in Proc. of Int. OFDM Workshop, Hamburg, Germany, September 2001.
[11] S. Mengesha, H. Karl, A. Wolisz, “Improving Goodput by relaying in transmission-power-limited wireless systems,” in Proc. of Informatik 2001, Vienna, Austria, Sept. 2001.
[12] J.H. Winter, “Smart antennas for wireless systems,” in IEEE Personal Communications, Feb. 1998.
[13] R.A. Monzingo and W.T. Miller, Introduction to adaptive arrays, John Wiley & Sons Inc., New York, 1980.
[14] M. Tangemann, “Introducing adaptive array antenna concepts in mobile communication systems”, Proc. RACE Mobile Telecommunications Workshop, Amsterdam, 1994, pp. 2/714-727.
[15] C.E. Perkins, et. al, “Performance comparison of two ondemand routing protocols for ad hoc networks”, IEEE Personal Communications, Feb. 2001,pp16-28.
[16] MIND: Mobile IP based Network Developments (IST2000-28584), http://www.ist-mind.org.
[17] M. Siebert, M. Lott, M. Weckerle, “Enhanced radio resource management introducing smart-direct link concepts,” submitted to EPMCC 2003, Glasgow, Scotland, April 22-25, 2003.
[18] “Broadband Radio Access Networks (BRAN); HIPERLAN Type 2; Data Link Control (DLC) Layer; Part 4: Extension for Home Environment”, ETSI TS 101 761-4, 2000.
