Multi-hop Data Transportation with WSM Protocol in Vehicular Ad-hoc Network

Transcription

1 Multi-hop Data Transportation with WSM Protocol in Vehicular Ad-hoc Network by Qing Chen B.A.Sc, Shanghai Jiaotong University, 1989 Research Project Submitted in Partial Fulfillment of the Requirements for the Degree of Master of Engineering in the School of Engineering Science Faculty of Applied Sciences Qing Chen 2015 SIMON FRASER UNIVERSITY Spring 2015

2 Approval Name: Degree: Title: Supervisory Committee: Qing Chen Master of Engineering Multi-Hop Data Transportation with WSM Protocol in Vehicular Ad-hoc Network Chair: Andrew Rawicz Professor, School of Engineering Science Jie Liang Senior Supervisor Associate Professor School of Engineering Science Jiangchuan Liu Supervisor Professor School of Computing Science Date Defended/Approved: March 24 th, 2015 ii

3 Partial Copyright Licence iii

4 Abstract Nowadays, the wireless communication in vehicular environment is a hot topic in automotive industry. Some communication standards and protocols have been published to support the data transportation between vehicle and vehicle and between vehicle and infrastructure, which is a big challenge due to the special characteristics of vehicular ad-hoc network. In this project, we implement a hybrid stack that combines WAVE Short Message Protocol (WSMP) with a recently developed protocol the Reactive, Density-aware Timely Dissemination protocol (REACT-DIS). The project will evaluate the performance based on its throughput, goodput and packet loss ratio by using NS3 simulation tool. Furthermore, we develop an enhanced approach that combines the REACT-DIS protocol with the limited acknowledgement to improve the performance of data dissemination in the system. The improvement of new approach also minimizes the impacting range of last hopping so that only the space near the final receiver will be impacted by the last hopping and the bandwidth of other area is not interfered. The new approach is implemented and evaluated using NS3 network simulator as well. Simulation results demonstrate that the enhanced approach outperforms the REACT-DIS protocol in an environment with low or mid density but is not ideal in high density environment. Keywords: IEEE p, IEEE 1609, WAVE, WSMP, Vehicular Ad-hoc Network, iv

5 Acknowledgements First of all, I would like to express my sincere gratitude towards my senior supervisor, Professor Jie Liang for his patience and constant assistance during my research. His generous guidance is essential throughout the project. I would also like to thank the Chair, Professor Andrew Rawicz and my supervisor Professor Jiangchuan Liu for their time and advices during my project. I also thank my co-workers at Bendix CVS for providing various research options and useful materials especially at the beginning of the project. Finally, I would like to thank my wife, Haiying Zhuang, my son, Ziyi Chen and my whole family for their persistent love, support and encouragement. v

6 Table of Contents Approval... ii Partial Copyright Licence... iii Abstract... iv Acknowledgements... v Table of Contents... vi List of Tables... viii List of Figures... ix List of Acronyms... x Chapter 1. Introduction Motivation of Research Report organization... 2 Chapter 2. Background VANET WAVE protocol IEEE P IEEE Data Forwarding Mechanism Topology-based routing Geography-based routing REACT-DIS protocol Simulator NCTUns NS Research interest selection Matric Chapter 3. Implementation and Simulation Protocol Implementation Flooding protocol Packet header Design Reactive Density-aware and Timely Dissemination (REACT-DIS) protocol Protocol Packet header Design Simulation and Analysis Topology Parameters for REACT-DIS protocol vi

7 Others Experiment Result Chapter 4. REACT-DIS with Acknowledgement New Approach REACT-DIS with limited ACK Fast transmission during one hop range Queued transmission cancellation Conditional last hopping Implementation Analysis and Result Chapter 5. Conclusion and Future Work References.39 Appendix A Source Code for REACT-DIS with LIMITED-ACK vii

8 List of Tables Table 1 Header Format for Flooding Protocol Table 2 Packet header of REACT-DIS Table 3 Scenarios of Simulation Table 4 REACT-DIS Parameters Table 5 General configuration viii

9 List of Figures Figure 1 VANET Architecture... 4 Figure 2 WAVE Model vs ISO Model... 5 Figure 3 DSRC spectrum band and channels in the U.S Figure 4 LLC header and SNAP header... 7 Figure 5 Architecture of NCTUns Figure 6 NS3 Conceptual Structure Figure 7 Packet Processing Logic for Flooding Protocol Figure 8 WSM Packet Processing Logic Figure 9 WSM Stack Periodical Task Logic Figure 10 Topology of Simulation Figure 11 Packet Transportation Status in Space for Flooding Protocol Figure 12 Packet Transportation Status in Space for REACT-DIS Protocol Figure 13 Throughput - Flooding vs. REACT-DIS Figure 14 Goodput - Flooding vs. REACT-DIS Figure 15 Packet Loss Ratio - Flooding vs. REACT-DIS Figure 16 Header Format of REACT-DIS with Limited ACK Figure 17 Throughput - REACT-DIS vs. REACT-DIS with Limited ACK Figure 18 Goodput - REACT-DIS vs. REACT-DIS with Limited ACK Figure 19 Packet Loss Ratio - REACT-DIS vs. REACT-DIS with Limited ACK ix

10 List of Acronyms AODV BSS BSSID CCH DSDV DSR DSRC IEEE ITS LLC MANET MFR OBU OLSR PSID RSU SAE SCH SNAP TPMS VANET WAVE WSMP ZRP Ad-Hoc On-Demand Distance Vector Routing Basic service set Basic service set identification Control channel Destination Sequenced Distance-Vector Routing Dynamic Source Routing Dedicated short range communications Institute of Electrical and Electronics Engineers Intelligent Transportation System Logical Link Control Mobile Ad-hoc Network Most Forward within Radius On-board unit Optimized Link State Routing Service provider identifier Road side unit Society of Automotive Engineers Service channel Sub-network Access Protocol Tire Pressure Monitoring System Vehicular Ad-hoc Network Wireless Access in Vehicular Environment WAVE Short Message Protocol Zone Routing Protocol x

11 Chapter 1. Introduction 1.1. Motivation of Research The rapid developing wireless techniques are now being widely used in the vehicle industry. The applications are not limited to satisfy local system monitoring such as tire pressure monitoring system (TPMS) but are also developed to assist road safety, traffic control, and transportation analysis, increase comfort of travel and support invehicle entertainment. Vehicular communication may use different types of architectures such as cellular network architecture or ad-hoc network architecture. Cellular network has a wide range of coverage. As far as the vehicle is in the range of a valid base station, the cellular network is able to provide a reliable connection between vehicles or between vehicle and service center for data transportation. The vehicles may also communicate over ad-hoc network, calling Vehicular ad-hoc network (VANET). VANET is a special case of mobile ad-hoc networks (MANET), in which each vehicle acts as a mobile node. As an ad-hoc network, it does not need a base station or router to link mobile nodes together in advance. It provides a more flexible and dynamic network which makes it more suitable for high mobility traffic environment. [6, 11] Because of the high mobility of the vehicles in the system, VANET has its own special characteristics. First of all, due to various speeds and directions, the wireless network frequently connects and disconnects. It is difficult to establish a reliable routing path before a multi-hop communication starts. Secondly, as the traffic condition varies in different places, the scale of the network changes. The high network scalability in 1

12 vehicular environment is a challenge for a routing protocol to be flexible enough to fit different network sizes. At the same time, to protect the privacy of the user, the nodes in the network are also secured to avoid being traced. The characteristics of VANET makes it a big challenge to transport large amount of data, especially data streaming such as video from point to point with multiple hops, where a stable routing path is normally required in traditional network environment. After many years of design and research, official standards have been gradually published recently. IEEE introduced a series of standards called WAVE protocols (Wireless Access in Vehicular Environment) to support vehicular ad-hoc network. The WAVE protocol consists of IEEE p [25] and IEEE 1609 protocol [ ]. It simplifies the network establishment processes to reduce the latency and introduce a special approach call WAVE Short Message Protocol (WSMP) that provides a reasonable small overhead. Sponsored by Intelligent Transportation Systems (ITS), applications in vehicle industry are developed based on the WAVE protocol and its associated application level specification such as SAE J Dedicated Short Range Communications (DSRC) [21] Message Set Dictionary. In order to achieve reliable video streaming in VANET, this project will introduce, based on WSMP, a data dissemination protocol called reactive, density-aware and timely dissemination protocol (REACT-DIS). We will also discuss a potential enhancement of the protocol. Simulation is used to analyze and compare the performances of the protocols Report organization In the next chapter, the report will provide some background information, including VANET and its characteristics, WAVE protocol and its content, forwarding mechanisms, simulation tool, and research interest as well matric used for analysis. Chapter 3 focuses on simulation experiments to implement the enhanced flooding protocol and REACT-DIS protocol. It is separated into two sections, the 2

13 implementation section and simulation section. In implementation section, we introduce the packet header designed for each protocol and describe the logic of the major functions using block diagrams. The simulation section describes the network topology, the experimental scenarios and general parameter settings. Finally, simulation results of two protocols are compared. The entire Chapter 4 introduces a new approach that combines REACT-DIS protocol with a limited acknowledgement. The approach is implemented in NS3 simulator. The simulation is run and its result is analysed. The last chapter summarizes the report and provides some potential directions for further research. 3

14 Chapter 2. Background 2.1. VANET In VANET, the communication devices can be separated into two types: onboard unit (OBU) and road-side unit (RSU). The on-board unit is usually mounted on vehicles and the road-side unit is usually stationary and installed alongside the road. Figure 1 shows the general architecture of VANET. Figure 1. VANET Architecture [8] Due to unstable topology and frequent disconnection, video streaming becomes a challenge in VANET. The major issue is the decrease of video quality caused by increased delay, loss and overhead during data transportation in high dynamic 4

15 environment. Therefore, the traditional data routing mechanisms are no longer completely suitable of video streaming in VANET. The new protocol stack shall have following properties to adapt to the requirements of VANET: Low latency Small overhead High scalability 2.2. WAVE protocol WAVE protocol contains a group of protocols mcovering from physical layer up to transport layer. The general structure of the protocol is shown in Figure 2 [13]. ISO Model WAVE Model Application Layer Safety Non-Safety Presentation Layer Session Layer Transport Layer Network Layer IEEE (Security) Application (Not part of WAVE protocol) IEEE (WSMP) Application (Not part of WAVE protocol) TCP/UDP IPv6 Data Link Layer IEEE (LLC sublayer) IEEE (MAC sublayer extension) IEEE P (MAC Sublayer) Physical Layer IEEE P (PHY Layer) Figure 2 WAVE Model vs ISO Model 5

16 IEEE P IEEE802.11p standard [25], finished in 2010, described the physical layer as well as the medium access control protocols required in VANET. IEEE802.11p uses 5.9 GHz band, from 5.85 GHz to GHz. The 75 MHz bandwidth is divided into seven channels as shown in Figure 3 [21]. Each channel uses 10 MHz bandwidth. Channel 178 is reserved for control information. Channel 172 and 184 are assigned for life safety and public safety information. The rest four channels are assigned for other services. Channel pair 174/176 and 180/182 can also individually combine into a 20 MHz channel 175 and 181 respectively. Figure 3 DSRC spectrum band and channels in the U.S. In order to suit for the fast changing topology of the vehicular environment, an IEEE802.11p WAVE device is allowed to send and receive data with wildcard BSSID without joining a BSS. The authentication process and channel scanning process that is used by traditional protocols are eliminated in order to reduce the time for network establishment. As a trade, p MAC doesn t provide data confidentiality service. The relative service is provided by higher layer of WAVE protocol. Theoretically, the communication distance of IEEE802.11p WAVE device is up to 1000m. The expected bitrate ranges from 3Mbps to 27 Mbps for 10 MHz bandwidth channels. 6

17 IEEE 1609 IEEE 1609 is made up of a series of sub-specifications, mainly [22], [23] and [24]. IEEE provides security function for application and management messages. IEEE is a specification that covers network service and transport layer functions. IEEE is an extension of MAC sublayer function to support multi-channel operation. As shown in Figure 2, IEEE1609 supports both IPv6 TCP/UDP and WSMP. WSMP is a special protocol designed for WAVE. It allows applications to control physical layer characteristics such as channel used and transmission power directly. The WSMP protocol is mainly design for safety application. But it also supports non-safety applications. The WSMP header contains multiple fields such as PSID, extended fields and etc. The PSID is used to determine the application entry. The extended fields are used for optional lower layer control parameters such as channel number, data rate and transmit power. The header of WSMP is about 10 bytes, variable because of the length of PSID field and optional extended fields, while the minimum head of IPv6 standard is 52 bytes. So the WSMP header is much shorter. In order to identify between WSMP message and IPv6 message, a two-byte ether-type field is added into SNAP header (refer to Figure 4), of which a 0x88DC indicates a WSMP message and a 0x88DD for IPv6 message. LLC Header SNAP Header Protocol ID Ether-type Figure 4 LLC header and SNAP header 2.3. Data Forwarding Mechanism In general, ad-hoc routing protocol can be classified into topology-based approach and geography-based approach. 7

18 Topology-based routing The topology-based routing can be divided into three types: proactive, reactive and hybrid. The proactive routing, for example DSDV (Destination Sequenced Distance-Vector Routing) and OLSR (Optimized Link State Routing), continuously discovers the environment and keeps a table of the nearby nodes. It causes high cost of overhead and latency for route maintenance. The reactive routing will create the route path when necessary. An initial delay to construct the route is needed but it avoids maintaining the routing table. AODV and DSR are the examples of reactive routing. It partially solves the problem of proactive routing but still not efficient enough when facing the dynamic vehicular environment. The hybrid routing protocol such as ZRP is to combine above two approaches to improve the efficiency and scalability. The proactive routing and reactive routing are also classified as flat routing, and hybrid routing as Hierarchical routing Geography-based routing Geography-based routing depends on physical position information of one-hop neighbours and/or destination to decide the route. Since the position information is easy to obtain due to the wide usage of GPS technology, the geography based routing is one of the research directions in VANET. Geography-based routing usually uses greedy strategy. For example, the Most Forward within Radius protocol (MFR) [28] selects the forwarding node that has the most progress (the distance between the source node and the projector of an intermediate node on the line that connects source node and destination node) in the range of source node. The geography-based routing may also take least distance between intermediate node and destination node or the closest direction of intermediate node comparing the direction of destination (called compass routing) as the strategy to determine the forwarding nodes. Most of the strategies use destination location as a reference. 8

19 Geography-based routing may have sender-based forwarding or receiver-based forwarding. In sender-based forwarding, the sender will select the relay node that will be responsible of forwarding the message. In order to select the proper node, the sender has to keep tracking information of its neighbours. Therefore, it is necessary for each node to distribute its information such as position, speed, and direction, which is quite difficult as these information keeps changes all the time in vehicular environment. The receiver-based forwarding, on the other hand, selects the relay node by receiver. It doesn t have to track its neighbours nor have to understand the topology of the network system, and hence considerably reduces the latency and bandwidth cost to construct and maintain the route while keeps better scalability REACT-DIS protocol The routing protocol used in the project is called Reactive, Density-aware and Timely Dissemination introduced by Cristiano Rezende, Richard W. Pazzi and Azzedine Boukerche [7]. As a receiver-based protocol, the sender will include its position information in the packet. The forwarding node will calculate the distance between the sender and itself, and then use greedy strategy to decide if it shall relay the packet or not. The further the forwarding node is, the higher the priority to reply the packets. To fulfill this algorithm, a forwarding node won t relay the packet immediately when it receives a packet. Instead, when a packet is received the first time, the forwarding node will assign a short delay for the received packet depending on the distance with the sender and relay the packet after the delay time expires. In order to prevent the problem of connection disruption in low density traffic area and broadcast storming in high density traffic area, the protocol include a density awareness mechanism. The forwarding node will increment a counter every time the same packet is received during the delay. When the delay expires, it will calculate a forwarding probability based on the number of the same packet received during the delay and decide the relay operation accordingly. It also includes other algorithms to solve the problem such as long end-to-end propagation delay due to the wait time of each forwarding node, especially when multiple hopping is required between the data source and final user. The protocol has following characteristics: 9

20 Delivery guaranteed by redundant retransmission during data dissemination. Storming issue prevented by involving a set of algorithms that will be introduced in details in Chapter 3 Overhead limited by working with WAVE protocol Propagation delay limited by including a relay mode in which the received packet will be forwarded with short delay Scalability enhanced by including density awareness algorithm 2.4. Simulator Due to the high research cost in real VANET environment, network simulators are widely used to design and analyze the performance of various protocols, for example NS2, NS3, OMNet, and NCTUns. In this research, NCTUns and NS3 are two options selected at the beginning of the research preparation because both of them have WAVE protocols implemented NCTUns NCTUns is a simulation tool that was developed by Prof. S.Y. Wang in National Chiao Tung University. It can work both as a network simulator or an emulator. NCTUns provides GUI for user to easily create network topology, add, remove or change the protocols at different layer to create user s own communication stack, run and replay the simulation in graphical interface. Figure 5 shows the architecture of the tool. The distributed architecture of the tool also enables remote simulation and simultaneous simulation. [15, 16, 17, 27] 10

21 Figure 5 Architecture of NCTUns The integrated traffic simulation function of NCTUns tool allows user to define the road map, the vehicle nodes and the traffics in GUI interface, and modify communication stack for each of the nodes and mobility parameters, which makes it a handy tool for VANET simulation. NCTUns was open-source and available for free up to version 6.0. NCTUns was commercialized since 2012 and renamed as EstiNet. The source code and the technical support are now available depending on the license purchased NS3 NS3 tool is a new discrete-event simulator that is to replace NS2 [18]. A discreteevent simulator means that a point in simulation time is assigned to every event, events are initiated and triggered consecutively and simulation time moves in discrete jumps from event to event [9]. NS3 is aimed for research and education. NS3 is developed in C++ and incompatible of NS2. NS3 has five general elements: node, application, net device, stack 11

22 and channel. To make the simulator easy to use, helper class is developed in most of the modules to assist creating and configuring simulator application. Node Application Stack Channel Network Net Device Figure 6 NS3 Conceptual Structure NS3 is open-source and is well-documented. A WAVE module is developed, which includes IEEE p and WSM protocol. But, there is no traffic simulator developed in NS3. Instead, NS3 uses mobility module to add and configure mobile parameters for the nodes Research interest selection In order to achieve data streaming in VANET, the first thing is to design an appropriate stack that is suitable for high dynamic VANET environment. This project focuses on integrating a stack and analyzing the performance using the simulator. In the stack, IEEE p is selected to form the PHY and MAC layer. Network layer uses WSMP protocol of IEEE-1609 protocol to provide low latency and small overhead. Above them, REACT-DIS protocol is implemented to provide forwarding algorithm that satisfy the dynamic VANET environment. At the end of the research project, a new improved REACT-DIS protocol called REACT-DIS with limited ACK is designed and implemented, and its performance is analyzed as well. 12

23 When selecting network simulator, NCTuns was first investigated but NS3 tool is finally selected because, based on the research interest of this project, mobility module of free open-source NS3 tool provides sufficient needs in the simulation Matric To compare the performance, an enhanced flooding protocol is implemented and is used as starting point. Then, the reactive, density-awareness and timely dissemination protocol is implemented and the performance is compared in throughput, goodput and packet loss ratio. In communication network, throughput is usually used to measure the performance. Throughput of a communication network is the rate of successful message delivery over a communication channel [6]. In the research, the throughput is calculated based on data received by the receiver in follow method: Throughput = total number of bits received time However, since the packets are transported in broadcast manner and redundant transmission is used to increase the reliability of data transportation, there are a lot of retransmissions and overhead during communication in simulation. From application point of view at receiver side, the redundant data are useless. As a comparison, the goodput is calculated to measure the number of useful data transmitted to receiver. Goodput = number of useful bits received time Packet loss is also used to measure the reliability of the protocol for data transportation. Due to the nature of broadcast during multiple hopping, the collision will happen even though only one service is running in the simulation. Although redundant transmission will help to avoid packet loss, the packet loss may happen still during 13

24 transmission. In this project, the packet loss ratio is used to measure the packet loss status in the network. Packet loss ratio = number of packets missed number of useful packet received 14

25 Chapter 3. Implementation and Simulation During the research, groups of simulation experiments are run using NS3 simulator to simulate different protocols and the new approach, and validate the performance of various designs in VANET. In this chapter, we will introduce the implementation of the enhanced flooding protocol and REACT-DIS protocol, and their simulation details Protocol Implementation Flooding protocol Flooding protocol is a traditional routing protocol that doesn t need route discovery and maintenance. It uses broadcast to flood the network by continuously sending each packet it receives to its neighbours. Because of redundant packet broadcast and all route paths are used, the delivery quality and the shortest path is guaranteed Packet header In the process of flooding protocol, two restrictions are included to avoid duplicated packets sent from one node and prevent broadcast storming in the network: Tracking each packet received and its timestamp to guarantee that the same packet is delivered only once. Setting maximum hop for each packet so that the broadcast stops if the maximum hop reaches 15

26 In the experiment, a header is designed to contain essential information for the intermediate node to make a forwarding decision. The header has the format as shown in Table 1. Sender Seq No. Hop Name Length Description Sender 4 bytes The ID of initial data source Seq 1 byte Sequence no of the packet sent by the data source Hop 1 byte Maximum hops that next node shall carry on Table 1 Header Format for Flooding Protocol Design The following diagram shows the processing logic of flooding protocol when a packet is received. 16

27 Retrieve packet header Is the a receiver? No Yes Is the packet received before? Transmit message to application to process (Ignored in simulation) Yes Record message No Get max. hops from header Is max. hops zero? No Max. hops - 1 Yes Build header with Max. hops Send packet with new header Figure 7 Packet Processing Logic for Flooding Protocol 17

28 Reactive Density-aware and Timely Dissemination (REACT- DIS) protocol Protocol As introduced in section 2.3.3, the REACT-DIS protocol contains three main algorithms to control the dissemination operation. They are forwarding delay selection, density awareness and maximum relay. [7 10] 1. Forwarding delay selection Each node that receives a message will choose a delay time based on the distance between the sending node and itself. The delay time is inversely proportional to the distance. Therefore, the node that is further away from the sender will have higher priority to relay the message. t = [(1 distance ) (T R max T min )] + T min 2. Density awareness In order to avoid the high packet collision situation in high node density environment, the protocol includes a forwarding probability. Each node will count how many times the same packet is received during the delay time and calculate the forwarding probability based on the count. With the increase of duplication of one packet, the probability of retransmission will reduce. The relationship is shown as follow: ρ = 1 r count 1 [7]. where r is forwarding probability reducer. Based on test, the ideal value of r is Maximized relay 18

29 An issue of reactive approach is the end-to-end delay accumulated from the waiting time described in the first algorithm during multiple hops. To avoid this issue, once a node decides to relay a packet, it will retain in relay mode for a certain time. As a result, the node in relay mode won t compete for forwarding based on distance in this time period. It will just wait for a very short delay before it forwards new received packet. However, there could be a chance that at one certain time period, there are excessive numbers of nodes are in relay mode. This will again cause more than enough redundant packet transmissions and raise the possibility of collision. To solve the problem, another probability is included in the protocol as follow: δ = 1 c k 1 (c k) τ c > k where c is the number of the redundant packets transmitted, k is the expected number of forwarding and τ is the forward reducing factor. Based on the evaluation, the ideal k is 4 [7] Packet header Based on the protocol, the forwarding nodes that receive the packets need the position information of the previous hop node. Therefore, the packet header is extended to include two fields for X position and Y position, as shown below Sender Seq No. Hop X Position Y Position Name Length Description Sender 4 bytes The ID of initial data source Seq 2 bytes Sequence no of the packet sent by the data source Hop 1 byte Maximum hops that next node shall carry on X position 8 bytes X position of the transmitter when a packet is sent Y position 8 bytes Y position of the transmitter when a packet is sent Table 2 Packet header of REACT-DIS 19

30 Design In the experiment, REACT-DIS protocol is implemented in a MyWSMStack class. The major functions of this class include 1) processing received packets; 2) maintaining queued packets and packets history; and 3) transmitting the packet when the forwarding condition matches. When a WSM packet is received, the WSM packet processing function is triggered. It disassembles the packet header and, based on the logic shown in Figure 8, assigns either SCHEDULE_MODE or RELAY_MODE or NORELAY_MODE for each packet. A periodical handler then scans through queued packets to decide if a packet in SCHEDULE_MODE or RELAY_MODE shall be sent or not when time is due. When a queued packet is sent, the packet is assigned with RELAYED_MODE. Otherwise, NORELAY_MODE is assigned. The handler also checks if the relay mode of the forwarding node is time-out or not. Figure 9 shows the control logic of the periodical task based on REACT-DIS protocol. In order to track the received packets, the stack contains a member called m_msghistory. A set of methods are also implemented to maintain this message queue and remove the message when it expires. A WSMProtocolHeader class is implemented to handle the packet header as mentioned in section To simply the configuration in simulation application, MyWSMStackHelper is implemented to help installing the MYWSMStack to one node or a group of nodes (called NodeContainer in NS3 tool). The following diagrams show the packet processing logic and the periodical task control logic implemented for REACT-DIS protocol 20

31 Retrieve packet header Is the node a receiver? No Yes Is the packet received before? No Yes Record message Get max. hops from header Transmit message to application to process (Ignored here) Increment count of message No Calculate distance Set a random delay based on distance No Max. hops - 1 Is in Relay mode? Yes Set random delay Is max. hops zero? Yes Set SCHEDULE_MODE Set RELAYING_MODE Set NORELAY_MODE Figure 8 WSM Packet Processing Logic 21

32 Reset NodeRelayMode if NodeRelayMode times out For each queued message Evaluate Mode RELAYIN G MODE SCHEDUL E MODE Others IF delay time expires IF (condition 3 of section ) Send packet Set to RELAYED_MODE ELSE Set to NORELAY_MODE IF delay time expires IF (condition 2 of section ) Send packet Set to RELAYED_MODE Set NodeRelayMode flag Set NodeRelayMode timer ELSE Set to NORELAY_MODE Figure 9 WSM Stack Periodical Task Logic 22

33 3.2. Simulation and Analysis Topology In this research, we assume that all nodes know their positions in a 750mx750m 2D space. There is no obstacle between nodes to block the wireless communication. All nodes have the same transmission range. To simply the experiment, we assign a node as a packet generator and another node as a receiving node. Both nodes start from the center of the map and move in opposite direction at a constant velocity. At the same time, there are hundreds of fixed nodes evenly distributed in the 2D space and act as forwarding nodes. Figure 10 shows the general topology of the simulation. d v R G v n Figure 10 Topology of Simulation There are three main scenarios designed based on number of forwarding nodes at the background, 256 nodes, 961 nodes and 2601 nodes (refer to Table 3). From 23

34 scenario 1 to scenario 3, each of them individually represents the low density environment, the mid density environment and the high density environment. Scenario Nodes per row Number of rows Distance in meters (d) Number of nodes Table 3 Scenarios of Simulation The velocities of the packet receiver and the packet generator are both m/s in all simulations. The packet receiving node goes towards node 0 and the packet generating node towards node n in the simulation Parameters for REACT-DIS protocol In REACT-DIS protocol, the values used for parameters of three algorithms mentioned in section are shown in Table 4 Symbol Description Value Algorithm R Redius 180 m T max Maximum delay time 100 ms T min Minimum delay time 10 ms r forwarding probability reducer 10 k Expected number of forwarding 4 τ forward reducing factor 0.1 Table 4 REACT-DIS Parameters t = [(1 distance ) (T R max T min )] + T min 1 ρ = r count 1 δ = 1 c k 1 (c k) τ c > k Besides above parameters, the maximum time for a node to stay in relay mode (time ) is 100ms. 24

35 Others Although our experiment is developed based on WSMP, WSMP is not the major focus of this research. The research is concentrated on the performance of the forwarding protocol with continuous data transportation under scenarios mentioned in section As WSMP control message is not part of this research, the simulation experiment is simplified by excluding the 100-ms control message on CCH and all WAVE management messages required by WSMP protocol. Except specifically defined in the simulation application, the default configurations of WAVE module are used in the simulation. The data transportation will start one second after the beginning of the simulation to create an initial distance between packet generator and packet receiver but still keep them in one hop range. The entire simulation will run for fifteen seconds. Table 5 shows the general configuration used in the simulation. Sending period (ms) 100 Maximum transmission range of WIFI 250 m Maximum allowed hops 4 History packet timeout (ms) 2000 Packet size (bytes) 1000 Table 5 General configuration. To trace the simulation, a packet capture function in NS3 simulator is enabled at physical layer to capture all transmitted packets or received packets and save them into.pcap file in the simulation. WireShark network protocol analyzer is used for data analysis after the simulation Experiment Result Comparing the results of the simulations with 963-node scenario, the packet transportation status of flooding protocol is much worse than that of REACT-DIS protocol. Although theoretically the network with flooding protocol shall transport more 25

36 packets, the peak number of the packets transported is only 498 packets with flooding protocol, while 794 packets are transported in the network with REACT-DIS. This indicates that a severe packet collision and data loss situation happens in the network with flooding protocol. At the same time, from Figure 11 and Figure 12, it is obvious that the real network covering range of REACT-DIS protocol is much further than that of flooding protocol even though both of them have four hops and the configurations at the physical layer are the same. 500 Number of Packets Y (m) X (m) Figure 11 Packet Transportation Status in Space for Flooding Protocol 26

37 800 Number of Packets Y (m) X (m) Figure 12 Packet Transportation Status in Space for REACT-DIS Protocol From the throughput point of view, the data rates of the receiver in all three simulations remain low at about 135Kbit/s with flooding protocol, which are resulted from poor network performance such as high network congestion, packet collision and small network cover range. Comparatively, the throughput with REACT-DIS protocol is much better. With the increase of the density, the throughput also increases because more retransmissions occur due to more nodes in the area. In other words, the REACT-DIS protocol is much more flexible and has much better scalability in a dynamic vehicular environment. 27

38 0.6 Throughput Mbit/s Flooding REACT-DIS Number of Nodes in 750mx750m area Figure 13 Throughput - Flooding vs. REACT-DIS Neglecting retransmission and overhead, the goodputs are quite even in three scenarios either in the simulation with flooding protocol or in the simulation with REACT- DIS protocol. It is 74Kbit/s in the network with REACT-DIS protocol, comparing a 51Kbit/s goodput in flooding protocol network. From packet loss ratio diagram, the ratio of REACT-DIS protocol network is 7.3%, comparing that of 36.6% for enhanced flooding protocol. Therefore, the overall performance of REACT-DIS protocol is much better than that of enhanced flooding protocol. 28

39 Goodput Mbit/s Flooding REACT-DIS Number of Nodes in 750mx750m area Figure 14 Goodput - Flooding vs. REACT-DIS Packet loss ratio (%) Flooding REACT-DIS Number of Nodes in 750mx750m area Figure 15 Packet Loss Ratio - Flooding vs. REACT-DIS 29

40 Chapter 4. REACT-DIS with Acknowledgement 4.1. New Approach Based on REACT-DIS protocol, each packet will be transported in all directions for the maximum allowable hops. From Figure 12, though the network using REACT-DIS protocol covers wider range and may reach further to the receiver node, it also means that each transmission will impact more nodes in the simulation space. No matter how close the data source and the receiver are, any transmission will cause a waste of bandwidth in outer range of entire hopping area, resulting redundant retransmission and causing packet collision. The larger the number of the allowed hops, the wider the impacted range is and the worse the potential packet collision is. At the same time, due to lack of the receiving status of each packet, the data source has to broadcast the packet at a default rate even though the previous packets are already received. At the same time, since there is no way for the data source to detect whether the receiving node exceeds its maximum hopping range, it will keep broadcasting the data until the data is completely transmitted. To avoid these issues, the data transportation shall involve a backward acknowledgement to inform either the sender or the forwarder the receiving status. There are two types of acknowledgement mechanisms that could be integrated here: complete acknowledgement and limited acknowledgement. For complete acknowledgement, the system will try to transmit the ACK packet from the receiver all the way back to the data source within the network. The advantage of the mechanism is that the data source is able to detect the receiving status and then transfer new packet as quickly as possible and terminate a transmission service in case the 30

41 acknowledgement packet is no longer received. Theoretically, since both the receiving node and the forwarding nodes know the position where the packet is sent from, the ACK packet is able to be transferred backward. However, this also means that, more packet transportations will happen within the hopping range and will result in high probability of packet collision. The sender may also terminate the service mistakenly because of the loss of the acknowledgement packet during forwarding. Therefore, the entire control algorithm of the complete backward acknowledge will be very complex. Due to the time limitation, a limited acknowledgement is selected to be integrated with REACT-DIS in this research, which we call it REACT-DIS with limited ACK, to improve the performance of the data transportation. In REACT-DIS with limited ACK protocol, only the receiver node will broadcast an acknowledgement packet to its neighbour nodes after it receives a packet. The forwarding nodes that receive the ACK packet won t relay the ACK packet any further. With limited acknowledgement, the sender is able to immediately launch new packet transmission if it receives the acknowledgement from the receiver. This mechanism will accelerate the data transportation when the receiver is within one hop range of the data source. At the same time, instead of transporting the packet with maximum allowed hops, the last hopping will be conditional so that, the forwarding nodes cancel last hopping in case it doesn t receive the acknowledgement from the receiver before. In this way, the overall transmission range of a service is reduced and the bandwidth at the most outside hop is saved. However, the REACT-DIS with limited ACK won t guarantee receiving of acknowledgement. The sender will continue data transmission until all the data are sent if the ACK packet is lost or the receiver is out of one hop range of the sender REACT-DIS with limited ACK Besides the major forwarding node selection criteria of REACT-DIS, the REACT- DIS with limited ACK protocol has three differences comparing REACT-DIS protocol after involving the acknowledgement packet: 1) one-hop fast transmission, 31

42 2) queued transmission cancellation, and 3) conditional last hopping Fast transmission during one hop range When the receiver is close enough to the sender, the sender is able to receive the acknowledgement of the previous transmission and launch a new transmission immediately instead of waiting for a fixed interval. This will accelerate the data transportation whenever possible. The fast transmission will be terminated automatically when the sender no longer receives the ACK packet from the receiver Queued transmission cancellation Based on the acknowledgement packet, the nodes nearby are able to realize the receiving status of the receiver. It will cancel that queued transmissions if those packets are already received and acknowledged by the receiver. This will save the bandwidth and again reduce the probability of collision Conditional last hopping Obviously the last hop will cover much broader area than that of the inner hop. The proposed approach attempts to cut down multiple hopping dissemination range by one hop. The nodes will cease the last hop if they don t receive the acknowledgement packet from receiver for last several packets. Otherwise, they will complete the last hop by sending the new packet if they receive the ACK of previous packets. Due to the possibility of the packet loss, the receiver node is allowed to transmit multiple acknowledge packets for a same data packet in the approach to increase the reliability of transmitting the acknowledgement information. 32

43 4.3. Implementation To identify between regular data packet and the acknowledgement packet, one byte is included in packet header, called packet type. A value of 0 means data type packet and a 1 refers to ACK type. Figure 16 shows the header that is used for REACT- DIS with limited ACK. Sender Seq No. Hop X Position Y Position Packet-Type Figure 16 Header Format of REACT-DIS with Limited ACK The sender will generate new packet in following method: IF (receiving ACK of previous packet) Transmit new packet ELSE Transmit new packet when time due Set next transmission time At the same time, for the final receiver, it will operate as follow: IF (new data packet received) Transmit ACK packet ELSE IF (retransmission of latest data packet received && ACK count is less than 3) Transmit ACK The forwarding node will have to record both data packets and their ACK statuses in the simulation. In case the ACK packet is received before a data packet, which is possible because of multi-hop operation, the forwarding node shall also record the ACK packet. When a new packet is received, besides using processing logic as shown in Figure 8 to determine the mode for the packets that have two or more hops left, the following control mechanism is applied for the packets with one hop left: IF (a packet is received with only one hop left && the ACK of any of last two packets is received) 33

44 ELSE Apply packet processing logic of REACT-DIS (Figure 8) Set packet to NORELAY_MODE Right before relaying a queued packet, the forwarding node will check the ACK status of the relative packet and will stop the transmission if the acknowledgement associated with this packet is received, as required in section The Appendix A lists the code implemented for REACT-DIS with limited ACK Analysis and Result Similar as before, three simulations are run individually with 258 nodes, 963 nodes and 2603 nodes. Comparing with REACT-DIS protocol, the throughput of new approach is better when the node density is low but is worse when node density increases. The goodput performance is greatly improved when there are 258 nodes in the simulation range. With 14% increase of throughput, the goodput reaches 146Kbit/s, double the goodput of REACT-DIS protocol. The packet loss ratio decreases to 3.42% which is 11.5% with REACT-DIS protocol. In the simulation with 963 nodes, the goodput is 98.6Kbit/s, which is not as good as that with 258 nodes. But it is still 29% higher than that of REACT-DIS protocol. The packet loss ratio reduces to 1.22%, which is 4.48% with REACT-DIS protocol. With smaller hopping range, the performances of new approach with 258 nodes and 963 nodes are better than that of REACT-DIS protocol. 34

45 0.6 Throughput Mbit/s REACT-DIS REACT-DIS-ACK Number of Nodes in 750mx750m area Figure 17 Throughput - REACT-DIS vs. REACT-DIS with Limited ACK Goodput Mbit/s REACT-DIS REACT-DIS-ACK Number of Nodes in 750mx750m area Figure 18 Goodput - REACT-DIS vs. REACT-DIS with Limited ACK 35

46 25 Packet loss ratio (%) REACT-DIS REACT-DIS-ACK Number of Nodes in 750mx750m area Figure 19 Packet Loss Ratio - REACT-DIS vs. REACT-DIS with Limited ACK However, when the density is high, the performance of REACT-DIS with limited ACK is not ideal. Although the goodput looks a bit better than REACT-DIS, but the actual connection time between sender and receiver drops down to 4.47 seconds, while it is 6.71 seconds in REACT-DIS. In other words, the valid communication time decreases, which causes the drop of total data transported during simulation. At the same time, the packet loss ratio increases significantly from 5.97% (with REACT-DIS protocol) to 21.7%. This is because, even though the REACT-DIS protocol has density-awareness functionality, the packet retransmission increases still when the number of nodes in the range increases. This situation becomes dramatically worse due to fast transportation algorithm and causes critical collision issue and blocks the packet transportation in the network. Even after the fast transportation stops, this situation last within the network for a period of time because of multiple hopping. Therefore, stricter retransmission limitation criteria shall be designed in order to make REACT-DIS with limited ACK protocol work in high density environment. 36

47 Chapter 5. Conclusion and Future Work This project implemented the recently developed REACT-DIS protocol for multihop data transportation in VANET environment. A number of simulations are designed and run in NS3 to compare the performance of REACT-DIS protocol with that of the improved flooding protocol. The simulation result shows that the REACT-DIS protocol is able to forward data by intelligently selecting the forwarding nodes in a high dynamic environment without creating a route path in advance. Its density awareness algorithm helps to control the duplicated transmissions in different environment with various node densities. However, the fixed transmission rate of REACT-DIS limits the speed of the data transportation. Since the lack of the packet receiving status, the forwarding node will continue to relay the data to its neighbours until the maximum allowed hops reaches. Without extra information, the sender will not terminate the data transportation until all data are transmitted. To partially solve the problem, this project proposes a new design which includes a limited acknowledgement into REACT-DIS protocol, called REACT-DIS with limited ACK. With this new approach, the sender in one hop range or the intermediate nodes near the receiver are able to realize the receiving status and thus decide its packet generation rate or relay behavior as describe in section 4.2.1, and The protocol is implemented and the simulations are run again in NS3 simulator. The result shows that when in the low or mid density environment, the new approach out-performs the REACT-DIS protocol. But it is not suitable for high density environment without a strict retransmission criterion. 37