Efficient topology control algorithms for ad hoc networks

Transcription

1 University of Wollongong Research Online University of Wollongong Thesis Collection University of Wollongong Thesis Collections 2006 Efficient topology control algorithms for ad hoc networks G. Srivastava University of Wollongong, Recommended Citation Srivastava, Gaurav, Efficient topology control algorithms for ad hoc networks, PhD thesis, School of Electrical, Computer and Telecommunications Engineering, University of Wollongong, Research Online is the open access institutional repository for the University of Wollongong. For further information contact the UOW Library:

2

3 Efficient Topology Control Algorithms For Ad Hoc Networks A thesis submitted in fulfilment of the requirements for the award of the degree Doctor of Philosophy from The University of Wollongong by Gaurav Srivastava Bachelor of Computer Engineering (Hons) School of Electrical, Computer and Telecommunications Engineering 2006

4 Abstract An Ad-hoc network is a collection of devices equipped with computing and wireless communication capabilities, co-operating together to form a network. An ad-hoc node can communicate with other nodes within its transmission range, or use intermediate nodes to establish communication paths to nodes outside its transmission range. Intermediate nodes in a route can collaborate to act as routers and forward their own traffic as well as the traffic generated by other nodes in a network. This strategy of routing is different to wired networks were communication is made possible through specialised networking devices including hubs, switches and routers interconnecting Local Area Networks (LANs) and Wide Area Networks (WANs). The routers used in LANs and WANs are devices with large processing capabilities, and high speed communication links. The topology of a network consists of a set of links and nodes which are used to discover and maintain communication paths and assist in coordinating the flow of packets in a network. The topology information can be used for many purposes including evaluating the connectivity/bi-connectivity of a topology and construction of routing paths for network applications. Nodes generally rely on each other to acquire topology information. Topology information can be disseminated by one centralized node by using a specialised process known as flooding or can also be disseminated in a distributed manner by broadcasting partial link state information and generating local topology views. Such local views can be put together to generate a larger topology view of a network. ii

5 Abstract iii Topology control is defined as a process where the topology of a network can be controlled by selective addition of nodes and links within a network. This process of selective addition can significantly impact the power usage and connectivity of network devices and improve longevity of a wireless network. In this thesis we analyse the impact of topology control and propose new algorithms and strategies that improve the connectivity, fault tolerance and communication reliability of topology graphs. We review, classify, categorise existing literature and discuss problems and issues associated with topology construction and maintenance proposed with and without Global Positioning System (GPS) aided techniques. Distributed topology control algorithms proposed for constructing minimum node degree graphs [Bettstetter, 2002] including K-Neigh [Blough et al., 2003], Location Information No Topology (LINT) [Ramanathan and Rosales-Hain, 2000], Location Information Link State Topology (LILT) [Ramanathan and Rosales-Hain, 2000] and MobileGrid (MG) [Liu and Li, 2002], do not necessarily generate bidirectional connected topology graphs and may result in isolated nodes and disconnected clusters. Such isolated nodes and disconnected clusters affect the overall connectivity of a network. The bidirectional connectivity of a network is an important factor in order to determine its performance. A bidirectional connected network, generally known as a connected network, is able to provide access to all crucial parts of a network and allows delivery of applications through a network. The bidirectional links are critical as they facilitate two way communication between the transmitter and receiver nodes. We proposed two mechanisms, Collaborative Algorithm (CA) [Srivastava et al., 2004a] and Probable Critical Links (PCL) [Srivastava et al., 2004d] to improve the connectivity of minimum node degree graphs and analyse their performance [Srivastava et al., 2006]. Power aware topology graphs use low power communications in order to reduce the overall power consumption of network nodes. A power optimised topology graph reduces the total number of links in the topology graph by eliminating

6 Abstract iv long distance links and replacing them with a number of small links. Removing links in a network topology can impact the fault tolerance of a network. Fault tolerance of a network is the ability of a network to cope with link and node failures. We analyse the fault tolerance issues related to power aware topology graphs including Minimum Spanning Tree (MST) [Prim, 1957] and Relative Neighbourhood graphs (RNG) [Huang et al., 2002]. We propose Link Redundancy (LR) algorithm [Srivastava et al., 2005a] to improve the fault tolerance of topology control algorithms. We analyse the performance of LR through worked examples and simulations [Srivastava et al., 2005a]. The LR algorithms yields a higher fault tolerant topology as opposed to the distance based approaches where link selections are made on the basis of the separation distance and may not necessarily yield a higher degree of connectivity [Srivastava et al., 2005a]. Power control also allows nodes to improve the spatial reuse of a wireless channel by reducing interference on other network communications and allowing nodes to communicate simultaneously. We analyse the spatial reuse issues related to topology graphs which can limit the performance of the algorithms due to the presence of hidden nodes [Poon and Li, 2003]. We propose Distributed Range Assignment (DRA) algorithm [Srivastava et al., 2004b] to reduce the hidden nodes in a network topology. We and apply DRA to MST and RNG topology graphs and analyse their performance through worked examples and simulations [Srivastava et al., 2004b]. The distance based topology graphs including MST, RNG, K-Neigh, and CA- PCL [Srivastava et al., 2006] do not model obstacles in a network. The connectivity of distance based topology graphs can be severely limited due to the presence of obstacles as nodes alter transmission range on the basis of the separation distance of links. Including signal attenuation characteristics can construct an accurate view of a network. We analyse the disconnected nature of distance based topology graphs under Lognormal-shadowing [Cox et al., 1984] [Cox et al., 1987] signal attenuation model. The Lognormal-shadowing model is

7 Abstract v used to analyse the impact of signal strength variations due to shadowing and scattering on the connectivity of power based topology graphs. We propose CA [Srivastava et al., 2006] in conjunction with the with power based topology graphs to improve the connectivity of a network [Srivastava et al., 2005b].

8 Statement of Originality This is to certify that the work described in this thesis is entirely my own, except where due reference is made in the text. No work in this thesis has been submitted for a degree to any other university or institution. Signed Gaurav Srivastava 1 July, 2006 vi

9 Acknowledgments I wish to thank my supervisors Dr. Paul Boustead and Prof. Joe Chicharo for their advice, assistance and support during the course of this thesis. I like to thank my family, Lalit, Neelam and Garima. A special thanks to Melanie Kruppa, for her companionship, patience, persistence and encouragements. Finally, I like to thank Justin Lipman, for his friendship and advice. vii

10 Contents 1 Introduction Background Thesis Overview Accomplishments and Contributions Publications Literature Review Introduction Medium Access Control Protocols Hidden Node Problem Exposed Node Problem IEEE Routing Protocols Proactive Routing Protocols Reactive Routing Protocols Hybrid Routing Protocols Signal Propagation Models Free Space Signal Attenuation Model Two Ray Signal Attenuation Model viii

11 CONTENTS ix Okumura Signal Attenuation Model Hata Signal Attenuation Model Lognormal Shadowing and Scattering Signal Attenuation Model Topology Control in Adhoc Networks Classification of the Topology Control Algorithms Centralised Topology Control Algorithms Minimum Spanning Tree (MST) Relative Neighbourhood Graph (RNG) Delaunay Triangulation (DT) Graph Novel Topology Control Algorithm (NTC) Connect Algorithm Biconn-Augment Algorithm Minimum Radius Graph (minr) Distributed Topology Control Algorithms Local Information No Topology (LINT) Algorithm Local Information Link-State Topology (LILT) Algorithm K-Neigh Algorithm Local Minimum Spanning Tree (LMST) Graph Distributed Relative Neighbourhood Graph (DRNG) Distributed Novel Topology Control Algorithm (DNTC) MobileGrid (MG) Common Power level (COMPOW) CLUSTERPOW Comparison of the Topology Control algorithms Research Issues

12 CONTENTS x Power Conservation Issues Spatial Reuse Issues Discussion of Open Research Issues Considered In The Thesis Summary of Research Issues Identified in thesis Connected Fixed Node Degree Graphs Introduction Collaborative Algorithm (CA) CA Implementation CA Worked Example CA Limitations Probable Critical Links (PCL) PCL implementation CA-PCL Worked Example Simulation and Analysis CA Scenario CA Results CA Summary CA-PCL Scenario CA-PCL Results CA-PCL Summary Conclusion Improving The Reliability Of Topologies Introduction Fault Tolerance

13 CONTENTS xi Simulation Results Summary Link Redundancy (LR) Selection LR Implementation LR Worked Example Simulation and Analysis Results Summary Conclusion Reliable Channel Access and Reuse Introduction Transmission Power and Channel Reuse Overview Impact of Transmission Range on Hidden and Exposed Nodes Identifying Exposed nodes and Hidden nodes in a topology graph Distributed Range Assignment (DRA) Algorithm DRA Implementation Impact of Global and Local topology Information on Hidden Node Evaluation Worked Example Simulation and Analysis Scenario Overview Results

14 CONTENTS xii 5.6 Conclusion Power Based topology Graphs Introduction Connected Topology Construction Problem Power Based Topology Graphs Implementation Worked Example Simulation and Analysis Scenario Overview Results Conclusion Conclusions and Future Work Overview Significant Results Further Work Bibliography 169 Appendices 179 A Chapter 3 Simulation Data 180 B Chapter 4 Simulation Data 191 C Chapter 5 Simulation Data 201 D Chapter 6 Simulation Data 208

15 List of Figures 2.1 (a) The Hidden Node problem, node A and C cannot sense each other s carrier and transmit simultaneously to node B (b) The Exposed Node problem, node B prevents node C from transmitting (a) Infrastructured mode of IEEE (b) Infrastructureless mode of IEEE (a) PCF frame exchange sequence in IEEE [Group, 1999a] (b) RTS/CTS/DATA/ACK with Virtual Carrier Sensing in the DCF mode of IEEE [Group, 1999a] (a) The Hidden Node solution, node C defers access after receiving a CTS from node B (b) The Exposed Node problem, CTS from node B prevents node C from transmitting A Classification of Topology Control Algorithms A Lune of node i and j is the area covered by the intersection of the two arcs (a) Maximum Transmission Power Based Routing Topology (b) Power Optimised Routing Topology (a) Channel Reuse in Maximum Power Topology (b) Channel Reuse in Low Power Topology (a) MST based network topology. (b) RNG based network topology (a) K-Neigh based topology graph with nd=2. (b) K-Neigh based bidirectional topology graph with nd=2. (c) CA based network topology with nd= xiii

16 LIST OF FIGURES xiv 3.2 (a) Alone Soldier problem in K-Neigh based topology graph with nd=2. (b) CA based network topology with nd= Two disjointed clusters due to the fixed node degree value of A lune of i and j is the region between the two arcs (a) Maximum Power Topology of a 5 node network (b) K-Neigh/CA(nd=2) based topology of a 5 node network (c) CA-PCL based topology of a 5 node network A comparison of the average power adaptation decisions vs total number of network nodes A comparison of the average end-to-end connectivity vs total number of network nodes A comparison of the average one hop neighbours vs total number of network nodes A comparison of the average power vs total number of network nodes A comparison of the average network connectivity vs total number of network nodes A comparison of the average one hop bidirectional neighbours vs total number of network nodes A comparison of the average power vs total number of network nodes A comparison of the network connectivity at different node failure rates vs total number of network nodes A comparison of the network connectivity at different node failure rates vs total number of network nodes A comparison of the network connectivity at different node failure rates vs total number of network nodes A comparison of the average network connectivity of the topology control algorithms [Srivastava et al., 2004c] A comparison of the fault tolerance of the topology control algorithms [Srivastava et al., 2004c]

17 LIST OF FIGURES xv 4.3 A comparison of the transmission power of the topology control algorithms A comparison of the hop diameter of the topology control algorithms A comparison of the average one hop neighbours of the topology control algorithms A Comparison of the RNG and LR-RNG topology graphs A comparison of the fault tolerance of the topology control algorithms A comparison of the average one hop bi-directional neighbours of the topology control algorithms A comparison of the average transmission power of the topology control algorithms A comparison of the hop diameter of the topology control algorithms A DRNG based network topology DRA applied to a DRNG based network topology A comparison of the average number of hidden nodes per link against the total number of network nodes in Distributed Relative Neighbourhood Graphs A comparison of the average number of hidden nodes per link against the total number of network nodes in Minimum Spanning Tree based Graphs A comparison of the average number of exposed nodes per link against the total number of network nodes in Distributed Relative Neighbourhood Graphs A comparison of the average number of exposed nodes per link against the total number of network nodes in Minimum Spanning Tree based Graphs A comparison of the average transmission power of the RTS and CTS packets per link against the total number of network nodes in Distributed Relative Neighbourhood Graphs

18 LIST OF FIGURES xvi 5.8 A comparison of the average transmission power of the RTS and CTS packets per link against the total number of network nodes in Minimum Spanning Tree based Graphs A Comparison of the DRNG topology graphs constructed using Free Space and Log-normal shadowing signal propagation models Power based DRNG topology, assuming a Free Space signal attenuation model Power based DRNG topology, assuming log-normal shadowing Power based bidirectional DRNG topology, assuming log-normal shadowing Power based bidirectional DRNG topology with Convert-TCN(), assuming log-normal shadowing A comparison of the connectivity of the RNG and RNG-CA topology graphs A comparison of the average one hop neighbors of the RNG and RNG-CA topology graphs A comparison of the average transmission power per link of the RNG and RNG-CA topology graphs

19 List of Tables 2.1 Summary of characteristics of the centralised topology control algorithms Summary of characteristics of the distributed topology control algorithms A.1 A comparison of the average power adaptation decisions vs total number of network nodes illustrated in Figure A.2 A comparison of the average end-to-end network connectivity vs total number of network nodes illustrated in Figure A.3 A comparison of the average one hop bidirectional neighbours vs total number of network nodes illustrated in Figure A.4 A comparison of the average transmission power vs total number of network nodes illustrated in Figure A.5 A comparison of the average network connectivity vs total number of network nodes illustrated in Figure A.6 A comparison of the average one hop bidirectional neighbours vs total number of network nodes illustrated in Figure A.7 A comparison of the average transmission power vs total number of network nodes illustrated in Figure A.8 A comparison of the average network connectivity at different node failure rates vs total number of network nodes illustrated in Figure A.9 A comparison of the average network connectivity at different node failure rates vs total number of network nodes illustrated in Figure xvii

20 LIST OF TABLES xviii A.10 A comparison of the average network connectivity at different node failure rates vs total number of network nodes illustrated in Figure B.1 A comparison of the average network connectivity vs total number of network nodes illustrated in Figure B.2 A comparison of the average network connectivity vs the node failure rate illustrated in Figure B.3 A comparison of the average transmission power vs total number of network nodes illustrated in Figure B.4 A comparison of the average hop diameters vs total number of network nodes illustrated in Figure B.5 A comparison of the average one hop neighbours vs total number of network nodes illustrated in Figure B.6 A comparison of the average network connectivity vs the node failure rate illustrated in Figure B.7 A comparison of the average one hop neighbours vs total number of network nodes illustrated in Figure B.8 A comparison of the average transmission powers vs total number of network nodes illustrated in Figure B.9 A comparison of the average hop diameter vs total number of network nodes illustrated in Figure C.1 A comparison of the average hidden nodes per link in Distributed Relative Neighbourhood Graphs vs the total number of network nodes illustrated in Figure C.2 A comparison of the average hidden nodes per link in Minimum Spanning Tree Based Graphs vs the total number of network nodes illustrated in Figure C.3 A comparison of the average exposed nodes per link in Distributed Relative Neighbourhood Graphs vs the total number of network nodes illustrated in Figure C.4 A comparison of the average exposed nodes per link in Minimum Spanning Tree Based Graphs vs the total number of network nodes illustrated in Figure

21 LIST OF TABLES xix C.5 A comparison of the average transmission power of the RTS and CTS messages per link in Distributed Relative Neighbourhood Graphs vs the total number of network nodes illustrated in Figure C.6 A comparison of the average transmission power of the RTS and CTS messages per link in Minimum Spanning Tree Based Graphs vs the total number of network nodes illustrated in Figure D.1 A comparison of the average network connectivity vs the total number of network nodes illustrated in Figure D.2 A comparison of the average one hop neighbours vs the total number of network nodes illustrated in Figure D.3 A comparison of the average transmission range vs the total number of network nodes illustrated in Figure

22 List of Abbreviations ABR ACK AODV AP ARA ARTP BS BSS CA CA-PCL CFP CI CO COMPOW CP CSMA CSMA/CA CTS DARPA DBTMA DCF DDR DE Associativity-Based Routing Acknowledgment Ad-hoc On-demand Distance Vector Access Point Ant-colony-based Routing Algorithm Articulation Point Base Station Basic Service Set Collaborative Algorithm Collaborative Algorithm with Probable Critical Links Contention Free Period Contention Index Communication Overhead Common Power Level Contention Period Carrier Sense Multiple Access Carrier Sense Multiple Access With Collision Avoidance Clear To Send Defense Advance Research Project Agency Dual Busy Tone Multiple Access Distributed Coordinate Function Distributed Dynamic Routing Directed Edges xx

23 List of Abbreviations xxi DFS DHCP DIFS DNS DNTC DRA DREAM DRNG DRNG DS DSDV DSR DSSS DST DT FAMA FHSS FSR GAMA GB GPS GSR IEEE IR LAN LAR LILT LINT LMST LR Depth First Search Dynamic Host Configuration Protocol Distributed Coordinate Function Inter Frame Space Domain Name Service Distributed Novel Topology Control Algorithm Distributed Range Assignment Distance Routing Effect Algorithm For Mobility Distributed Relative Neighbourhood Graph Distributed Relative Neighbourhood Graph Distribution System Distance-Sequenced Distance Vector Dynamic Source Routing Direct Sequence Spread Spectrum Distributed Spanning Tree Delaunay Triangulation Floor Acquisition Multiple Access Frequency Hopping Spread Spectrum Fisheye State Routing Group Allocation Multiple Access Graph Based Global Positioning System Global State Routing Institute of Electrical and Electronics Engineers Infra Red Local Area Network Location Aided Routing Location Information Link State Topology Location Information No Topology Localized Minimum Spanning Tree Link Redundancy

24 List of Abbreviations xxii LR-MST LR-RNG MAC MACA MACAW MANETs MG minr MPT MST MST+1 NAP NAV ND ND-GB ND-PCB Non-GB NonND Non-PCB NRP NTC OFDM OSLR PC PCB PCF PCL PHY PIFS PL Link Redundancy Algorithm Applied to Minimum Spanning Tree Link Redundancy Algorithm Applied to Relative Neighbourhood Graph Medium Access Control Multiple Access Collision Avoidance Multiple Access Collision Avoidance For Wireless LAN Mobile Ad-hoc Networks Mobile Grid Minimum Radius Graph Maximum Power Topology Minimum Spanning Tree Minimum Spanning Tree Graph with One Additional Link Per Node Neighbour Addition Protocol Network Allocation Vector Node Degree Node Degree Graph Based Node Degree Power Control Based Non Node Degree Graph Based Non Node Degree Non Node Degree Power Control Based Neighbour Reduction Protocol Novel Topology Control Orthogonal Frequency Division Multiplexing Optimised Link State Routing Point Coordinator Power Control Based Point Coordination Function Probable Critical Link Physical Layer Point Coordinate Function Inter Frame Space Pathloss

25 List of Abbreviations xxiii PL RDMAR RNG RNG+1 ROAM RP RTS SE SIFS SLURP SSA STAR TB TBRPF TORA VCS WAN WLANS WRP ZHLS ZRP Power Level Relative Distance Micro-discovery Ad-hoc Routing Relative Neighbourhood Graph Relative Neighbourhood Graph with One Additional Link Per Node Routing On-demand Acyclic Multipath Routing Protocol Request To Send Side Effect Edges Short Inter Frame Space Scalable Location Update Routing Protocol Signal Stability Adaptive Source Tree Adaptive Routing Tree Based Topology Broadcast Reverse Path Forwarding Temporally Ordered Routing Algorithms Virtual Carrier Sensing Wide Area Network Wireless Local Area Networks Wireless Routing Protocol Zone Based Hierarchical Link State Zone Routing Protocol