• 沒有找到結果。

# 幾何路由網路中有能源效率的調變式Beaconing機制

N/A
N/A
Protected

Share "幾何路由網路中有能源效率的調變式Beaconing機制"

Copied!
50
0
0

(1)國立台灣師範大學資訊工程研究所碩士論文. 幾何路由網路中有能源效率的 調變式 Beaconing 機制 An Adaptive Energy-efficient Beaconing Mechanism for Geographic Routing. 研究生： Marko Durasic 杜馬克 指導教授：蔡榮宗. 中華民國. 103 年 1. 撰 博士. 7月.

(3) Contents Abstract ............................................................................................................................................................... i List of Figures .................................................................................................................................................. iii List of Tables ..................................................................................................................................................... iv 1.. Introduction ................................................................................................................................................ 1. 2.. Related work .............................................................................................................................................. 5. 3.. Adaptive Energy-efficient Beaconing Mechanism for Geographic Routing (AEE) ................................. 7 3.1 AEE .......................................................................................................................................................... 7. 4.. Performance evaluation ........................................................................................................................... 14 4.1 Results showing the impact of AEE on various node speed and various numbers of data flows with high initial energy ........................................................................................................................................ 17 4.2 Results showing the impact of AEE on various node speed and various numbers of data flows with low initial energy ......................................................................................................................................... 28. 5.. Conclusion ............................................................................................................................................... 43. References ........................................................................................................................................................ 44. ii.

(4) List of Figures Figure 1: Local Minima Problem (left) and Node x’s void with respect to destination D. (right)[4] ............... 2 Figure 2: The right-hand rule (interior of the triangle). x receives a packet from y, and forwards it to its first neighbor counterclockwise about itself.[4]........................................................................................................ 2 Figure 3. Simulation results showing effect of network mobility on Packet Delivery Fraction ..................... 18 Figure 4. Simulation results showing effect of number of data flows on Packet Delivery fraction ............. 18 Figure 5. Simulation results showing effect of network mobility on Average Hop Count ........................... 20 Figure 6. Simulation results showing effect of number of data flows on Average Hop Count .................... 20 Figure 7. Simulation results showing effect of network mobility on Average End to end Delay ............... 21 Figure 8. Simulation results showing effect of number of data flows on Average End to end Delay .......... 22 Figure 9. Simulation results showing effect of network mobility on Beacon Overhead .............................. 23 Figure 10. Simulation results showing effect of number of data flows on Beacon Overhead ..................... 23 Figure 11. Simulation results showing effect of network mobility on number of packet collisions ............ 24 Figure 12 Simulation results showing effect of number of data flows on number of packet collisions ....... 25 Figure 13. Simulation results showing effect of network mobility on Energy Consumption ...................... 26 Figure 14. Simulation results showing effect of Number of Data Flows on Energy Consumption ............. 26 Figure 15. . Simulation results showing effect of network mobility on number of received data packets ...... 29 Figure 16. Simulation results showing effect of number of data flows on number of received data packets .......................................................................................................................................................................... 30 Figure 17. Simulation results showing effect of network mobility on Packet Delivery Fraction................. 31 Figure 18. Simulation results showing effect of number of data flows on Packet Delivery Fraction .......... 31 Figure 19. Simulation results showing effect of network mobility on Average Hop Count ......................... 32 Figure 20. Simulation results showing effect of number of data flows on Average Hop Count .................. 33 Figure 21. Simulation results showing effect of network mobility on Average End to end Delay............... 34 Figure 22. Simulation results showing effect of number of data flows on Average End to end Delay ........ 34 Figure 23. Simulation results showing effect of network mobility on Beacon Overhead ............................... 35 Figure 24. Simulation results showing effect of number of data flows on Beacon Overhead ..................... 36 Figure 25. Simulation results showing effect of network mobility on number of packet collisions ............ 37 Figure 26. Simulation results showing effect of number of data flows on number of packet collisions ...... 37 Figure 27. Simulation results showing effect of network mobility on Energy Consumption ...................... 38 Figure 28. Simulation results showing effect of Number of Data Flows on Energy Consumption ............. 39 Figure 29. Simulation results showing effect of network mobility on number of shut nodes ...................... 40 Figure 30. Simulation results showing effect of number of data flows on number of shut nodes ............... 40 Figure 31. Simulation results showing effect of network mobility on number of shut nodes throughout the simulation time................................................................................................................................................. 41 Figure 32. Simulation results showing effect of number of data flows on number of shut nodes throughout the simulation time........................................................................................................................................... 42 iii.

(5) List of Tables Table 1. Variables used in the algorithms ..................................................................................................... 13 Table 2. Parameters of the first set of simulations ........................................................................................ 17 Table 3. Parameters of the second set of simulations ................................................................................... 28. iv.

(7) One of the most popular geographic routing protocols is Greedy perimeter stateless routing (GPSR) [4].In GPSR the routing decision is made using merely local topology, which means there is no global topology. This kind of characteristics makes geographic routing completely stateless. Each node keeps track of its neighboring nodes by maintaining a neighbor table [4]. Greedy forwarding of data represents the method of deciding to which node to transmit data. GPSR operates in greedy forwarding manner where a node transmits data to a neighboring node that is geographically closest to the destination. However GPSR can be stuck in local minima situation (Figure 1), when there is no any neighboring node closer to the destination than the node that has data packets to transmit. In this situation GPSR recovers by routing around the perimeter of the void where greedy forwarding is impossible (Figure 1).. Figure 1: Local Minima Problem (left) and Node x’s void with respect to destination D. (right)[4]. It uses the right hand rule for traversing a graph which is depicted in Figure 2.. Figure 2:. The right-hand rule (interior of the triangle). x receives a packet from y, and forwards it to its first neighbor counterclockwise about itself.[4]. 2.

(9) beacon interval. This makes higher mobile and popular nodes generate more frequent beacon updates than the nodes that are mostly static and non-popular. This lowers the beacon overhead and energy consumption. We run a series of simulation to evaluate our strategy and to determine what sort of impact it makes on packet delivery fraction, beacon overhead, end to end delay, energy consumption, number of packet collisions in MAC layer, hop count and other network metrics. The rest of the thesis is organized as follows: In Section 2, we discuss some related work. A thorough explanation of the AEE mechanism is given in Section 3. In Section 4 we present the results of our simulations showing performance improvements succeeded by a conclusion of our thesis in Section 5.. .. 4.

(11) messages. As well, nodes that are far away from the data flow path would still send beacon updates. In fact, it would be more suitable for those nodes to adapt their beacon intervals based on their mobility and data flows to consume less energy and decrease collision probability. As an alternative, we propose a beaconing mechanism that can reduce beacon overhead and adapt more suitably to wireless networks with high mobility and many data flows.. 6.

(13) oldest (first) entry is removed first when the number of entries in the stack becomes greater than the MAX value. The method is explained in the following pseudo code.. 1 function Initialize 2 stack_pointer = 0 3 df =0 4 set time_limit 5 Let M[0..MAX] and T[0…MAX] be new stacks 6 for i=0 to MAX 7 M[i] = 0 8 T[i] = 0 9 end for 10 return stack_pointer, df, time_limit, M, T 11 end function Function Initialize is called at the initialization of the network. This procedure just sets the value of the stack_pointer and df to zero in lines 2-3, as we are assuming that the array elements of the stack begins at 0 and no traffic has yet flowed through the node. Variable time_limit is also created in line 4. This variable will be used to remove outdated entries from the stack. It is also creating two stacks of size MAX and it fills them with zeros in lines 5-9. One is used for tracking source destination pairs, another one for storing source destination pair’s time-out intervals. We are also assuming that the maximum number of elements of the stack is MAX. Below we can see how these variables are used.. 8.

(14) 1 function maintain _df (M, newpacket,df) 2 for j =0 to MAX 3 if(newpacket != M [j]) 4 Boolean value = TRUE 5 Else 6 Boolean value = FALSE 7 Break 8 end if 9 end for 10 If (stack_pointer <MAX) 11 If (Boolean value = TRUE) 12 M[stack_pointer] = newpacket 13 T[stack_poiinter]= time_limit # decrement by one each sec 14 stack_pointer = stack_pointer + 1 15 If ( df< MAX) 16 df = df +1 17 else 18 do nothing 19 else 20 break 21 end If 22 else 23 stack_pointer = stack_pointer - MAX; 24 M[stack_pointer] = new packet 25 stack_pointer = stack_pointer +1; 26 end if 27 return df 28 end function. In detail, the functions maintain _df works as follows: The for loop of lines 2-9 checks for previous entries in the stack and sets the Boolean value to true or false depending on the presence of the packet with same source destination information as the previous entries in the stack. If the packet is coming from a different data flow, then the Boolean value is set to true, otherwise it is set to false in which case break command is invoked and the program gets out of the loop. Line 10 checks if the stack_pointer is larger than the maximum size of the stack MAX. If it’s not that means that there are still empty slots to write new SD pairs. Line 11 checks if the Boolean value is true or not. TRUE value represents that the new coming packet is from a different SD pair; therefore lines 9.

(15) 12-18 are executed. Line 12 writes the new SD pair to the stack at the position where the stack pointer is “pointing”, i.e. available empty slot. Line 13 starts the timer for that specific entry. After that the stack pointer is incremented by one in line 14. Stack pointer is incremented by one so it can point to the next empty slot; df is also incremented by 1 because the new data flow has been written to the stack; df is incremented only if the current value of df is smaller than MAX, as MAX represents the maximum size of the stack therefore it is impossible to store more information about number of data flows. This way we can keep track of the number of most recent data flows passing through each node. Line 20 is executed in case the Boolean value is set to FALSE in line 6, which means that the new coming packet is a member of an SD pair that has been written to the stack already, so we can call command break in line 20 to get out of the if statement as there is no need for inputting anything to the stack. If the value of the stack_pointer is higher than the MAX size of the stack (which means stack is full), stack_pointer value is decremented by MAX in line 23. So the stack_pointer will point to the first entry of the stack and we can write to that position. By decrementing value of the stack pointer we are making sure that the pointer is pointing to the oldest entry of the stack so that the new value of SD pair can be written in line 24. Using this technique we are providing the stack with the most relevant SD pairs. In line 25 stack_pointer is incremented by 1 so it can point to the next oldest entry of the stack. Line 26 returns variable df so it can be passed on to the next function called time_out.. 10.

(16) After inputting new source destination pairs to the stack we are calling function time_out to check the stack for the expired entries, i.e. entries that have been in the stack longer than the pre-set time_limit:. 1 function time_out (M,T) 2 for i =0 to MAX 3 if ( T[i]=0) 4 M[i]=0 5 df= df-1 6 else 7 do nothing 8 end if 9 end for 10 return df 11 end function Entries are being checked in lines 2-9. Nodes whose timer has reached zero will be removed, i.e. set to zero. After removing outdated information in line 4, df is decremented by one to reflect the changed number of relevant data flows in line 5. Function time_out returns df in line 10 so it can be passed to the next function: deltaT. Nodes can calculate their ∆T based on the algorithm below. We are assuming that the nodes know their own speed.. 1 function delta_T (df, s) 2 ΔT = (1 / ((1+df) * (1+s))) *C 3 if (ΔT < min) 4 ΔT = min 5 return ΔT 6 end function The formula in line 2 calculates beaconing period. ∆T is a function of number of Source-Destination (SD) pairs and node speed; df denotes the number of SD pairs, s denotes speed and C denotes a constant that we use to adapt the maximum value of ΔT. This means that a 11.

(17) higher number of SD pairs will result in a shorter beacon period, and vice versa. Higher speed will result in shorter beacon period. On the other hand, very low values of ΔT or value 0 would badly affect the network performance for various reasons such as increased collisions, beacon overhead, etc. For that reason the minimum value is set in lines 3-4. Functions Initialize, time_out, maintain_df and deltaT are called at the initialization of the network. After initialization, functions maintain_df, time_out and deltaT are called at the arrival of the new packet. Functions time_out and deltaT are also called every ΔT period of time, in case there are no changes in data traffic activity but there are chances in mobility. All of the used symbols are listed and defined in Table 1. It is important to adequately control the maximum value of the adapted beacon interval as excessively high value of ΔT will not contribute to improved performances of the routing mechanism. After initializations and calculation of ΔT, nodes broadcast beacons periodically, and each node’s beacon interval is set to ΔT respectively, based on the speed and number of data flows traversing through it. Depending on the mentioned factors, beaconing interval will be higher or lower.. If a node is “unpopular” or “safe”, i.e. it is moving slow, or it is not found on the data path. frequently, its value of ΔT is set to a long interval. After the expiration of this beaconing period (ΔT), ΔT is calculated again using the current speed and data flow information. This means that after broadcasting beacons for some time, in the case of speeding up or sending lot of data packets frequently, the node’s value of ΔT will become lower and it will become a “popular” and “unsafe” node. This means that its periodic beaconing interval ΔT is shorter, therefore broadcasting beacons more frequently. This kind of beaconing strategy ensures that beacon intervals for static and “unpopular” nodes stay longer thus generating less beacon packets but still sustaining an accurate local topology. On the other hand those nodes which are extremely mobile and frequently transmitting data will frequently broadcast beacon packets as well.. 12.

(18) Symbols. Initialize. stack_pointer. df. time_limit. M. Definition. Function called at the initialization of the network. Pointer variable that “points” to an empty slot in the stack. Number of data flows. Life time of each entry in the stack. Array that holds the info about source destination pairs. First column holds the source information and second one holds the destination information. T. Array that holds the information about the period of time each node has spent in the stack. maintain_df. Functions used to maintain the stack and track new data flows. newpacket. Array that holds the source and destination of the new coming data packet. Boolean Value. MAX. time_out. delta_T. s. Boolean value which represents the presence of the new source destination pair in the existing table. Maximum size of the stack. function used to remove outdated entries from the stack. function used to determine adapted beacon interval. Speed of the node. ΔT. Adapted beacon interval. C. Constant number. min. Minimum value of ΔT Table 1.. Variables used in the algorithms. 13.

(19) 4. Performance evaluation In this section we are presenting a set of simulations that we have done in a popular network simulator tool called NS-2[9]. We used the following metrics for our evaluations: Packet delivery fraction, average end to end delay, total beacon overhead, total energy consumption, average hop-count, packet collisions, number of received packets and number of shut nodes. Each network metric is explained as follows: 1. Packet Delivery Fraction: the fraction of the number of delivered data packet to the destination. This shows the proportion of delivered packets to the destination. Packet Delivery Fraction = ∑ Number of packet received / ∑ Number of packet sent. The greater value of packet delivery fraction means the better performance of the routing protocol. 2. Average End-to-end Delay: the average time that takes a data packet to arrive at the destination. This is caused due to the delay of route discovery process and the line in data packet transmission. Only the data packets that are successfully delivered to destinations are counted. ∑ (arrive time – send time) / ∑ Number of data packets received by the destination nodes The lower value of end to end delay means the better performance of the protocol.. 3. Total Beacon Overhead: the total number of beacon packets transmitted by all nodes in the network. The lower value of total beacon overhead means the better performance of the network, provided that Packet Delivery Fraction and other important metrics are not negatively affected by it.. 14.

(20) 4. Total energy Consumption: the sum of energy consumed by each node in the network. Energy is consumed during transmission and reception of packets. The lower value of total energy consumption means the better performance of the network, provided that Packet Delivery Fraction and other important metric are not negatively affected by it.. 5. Average hop-count: The hop count refers to the number of nodes through which data must pass between source and destination. Each node along the data path represents a hop. The hop count is therefore a basic measurement of distance in a network.. 6. Packet collisions: the number of packet collisions at the Medium Access Control (MAC) layer.. 7. Number of received packets: the number of delivered data packet to the destination.. 8. Number of shut nodes: the number of nodes that have lost all the available energy and are incapable of transmitting or receiving data Results are showing impact of our Adaptive Energy-Efficient Beaconing Mechanism (AEE) for Geographic Routing on the above mentioned metrics. We tested our mechanism on GPSR [4], and compared results of GPSR with AEE (GPSR-AEE) and a regular GPSR. We will be referring to GPSR with AEE as GPSR-AEE. The simulations were performed for simulation time of 900 seconds. In every simulation run, 50 nodes were positioned randomly on the field of size 1500m600m. Each node transmission range is 250m. Data packet size is 64 bytes and each source is using CBR (Constant Bit Rate) that generates 4 packets per second. In both of two sets of simulations we have chosen 15 random source-destination pairs as data-flows and we varied the speed from 1m/s to 30m/s, so we can estimate the impact of mobility on network performances. We also varied the number of data flows from 1 to 30, thus showing the effect of different number of data-flows on the network performances. In the first set of simulations we have tested our mechanism for high values of initial energy that every node has. In the second set of simulations we have tested our mechanism for low values of initial energy that every 15.

(21) node has. In the former, the energy level of each node has been set to be enormously high at 1000 Joules at which the node will supposedly never run out of energy. In the later, it has been set very low at 10 Joules, so we can observe the influence of our energy-efficient algorithm on nodes shutting down due to the energy loss.. 16.

(22) 4.1 Results showing the impact of AEE on various node speed and various numbers of data flows with high initial energy Values and parameters which are used in this simulation are presented in Table 2: Parameters. Value. Number of nodes. 50. Topology. 1500 m X 600 m. Duration of the simulation. 900s. Transmission range. 250 m. Maximum speed of the nodes. 0~30 m/s (15m/s in simulation where we tested AEE for various numbers of data flows). Number of Data Flows. 1~30 (15 data flows in simulations where we tested AEE for various node speed). Mobility model. Random Waypoint Model. Traffic Source. Constant Bit Rate (4 packets/second). Data Packet Size. 64 bytes. Initial energy level of each node. 1000 Joules. Table 2.. Parameters of the first set of simulations. 17.

(23) Figure 3. Simulation results showing effect of network mobility on Packet Delivery Fraction. Figure 4.. Simulation results showing effect of number of data flows on Packet Delivery fraction 18.

(24) Figure 3 and Figure 4 show measurement of packet delivery fraction for two scenarios of various speeds and various numbers of data flows respectively. Packet delivery fraction of GPSR-AEE is showing improvements comparing to GPSR in both scenarios. As our mechanism is of adaptive nature more accurate network topology is maintained. Figure 3 illustrates that for low mobility GPSR-AEEE achieves higher values of the packet delivery fraction. However, at higher speeds GPSR-AEE and GPSR show similar values of the packet delivery fraction. Higher speed of nodes makes them move away from their neighbors’ transmission range more frequently, thus it is harder for nodes to maintain accurate topology of their neighbors. Figure 4 illustrates that for smaller number of data flows GPSR-AEE slightly outperforms GPSR. This improvement is caused by more accurate network topology that is maintained by GPSR-AEE beaconing mechanism. More accurate topology means more alternate routes available. For higher number of data flows both GPSR-AEE and GPSR show similar results. This performance result is attributed to the fact that data packets piggyback the local sending node’s location on all data packets it forwards, and runs all nodes’ network interfaces in promiscuous mode, so that each node gets a copy of all packets for all nodes within radio range [4].Therefore, increased traffic load leads to more accurate topology of the network as location updates are transmitted more frequently... 19.

(25) Figure 5.. Figure 6.. Simulation results showing effect of network mobility on Average Hop Count. Simulation results showing effect of number of data flows on Average Hop Count 20.

(26) By maintaining a more accurate network topology, GPSR-AEE is able to route packets through fewer nodes than GPSR. Figure 5 illustrates the effect of network mobility on Average Hop Count. The reduced hop-count at low mobility is mainly due to the MAC-layer collisions illustrated in Figure 11. For higher mobility GPSR-AEE reduces the hop-count due to the fact that periodic beaconing used in GPSR is not sufficient in maintaining an accurate network topology. For example, node can often send a packet to its neighbor, which is not anymore in its transmission radius. Figure 6 depicts the effect of varying the number of data flows on Average Hop Count. For low mobility hop-count is reduced due to the lower number of MAC-layer packet collisions illustrated in Figure 12. For higher mobility, the additional number of beacons generated by GPSR-AEE is used to maintain more accurate topology. This provides more precise route for packets to traverse to the destination.. Figure 7.. Simulation results showing effect of network mobility on Average End to end Delay. 21.

(27) Figure 8.. Simulation results showing effect of number of data flows on Average End to end Delay. In figures 7 and 8 we can see that GPSR-AEE can reduce average end to end delay too. These results are proportional to average hop-count results. A higher average hop-count produces higher end to end delay.. 22.

(28) Figure 9.. Figure 10.. Simulation results showing effect of network mobility on Beacon Overhead. Simulation results showing effect of number of data flows on Beacon Overhead. 23.

(30) Figure 12. Simulation results showing effect of number of data flows on number of packet collisions. Figures 11 and 12 clearly show GPSR-AEE reduces MAC layer collisions. Generating smaller amount of beacon updates lowers the number of collisions for slow and moderately fast networks, as well as for networks with smaller numbers of data flows. For very high speeds and larger number of data flows collisions are inevitable as beacons require to be sent more often to preserve correct local topology for getting a high packet delivery fraction.. 25.

(31) Figure 13.. Figure 14.. Simulation results showing effect of network mobility on Energy Consumption. Simulation results showing effect of Number of Data Flows on Energy Consumption 26.

(32) Figures 13 and 14 show GPSR-AEE reduces the total energy consumption of the network. Beacon overhead, collision rate, energy consumption are all network metrics that are closely related to each other. So far we have noticed that generating smaller numbers of beacon updates reduces beacon overhead and collision rate. It also reduces the total energy consumption of nodes in the network. For low mobility energy consumption is reduces up to 50 %. Since energy is consumed to send packets as well as beacon updates, nodes generate fewer beacon packets and therefore use less energy. For low data load, GPSR-AEE reduces total energy consumption up to 83 %. By increasing speed and also by increasing the number of data flows, energy consumption raises because of more frequent beacon updates. However, energy consumed by GPSR-AEE is still significantly lower.. 27.

(33) 4.2 Results showing the impact of AEE on various node speed and various numbers of data flows with low initial energy Table 2 shows the parameters and values used in this set of simulations:. Parameters. Value. Number of nodes. 50. Topology. 1500 m X 600 m. Duration of the simulation. 900s. Transmission range. 250 m. Maximum speed of the nodes. 0~30 m/s (15m/s in simulation where we tested AEE for various numbers of data flows). Number of Data Flows. 1~30 (15 data flows in simulations where we tested AEE for various node speed). Mobility model. Random Waypoint Model. Traffic Source. Constant Bit Rate (4 packets/second). Data Packet Size. 64 bytes. Initial energy level of each node. 10 Joules. Table 3.. Parameters of the second set of simulations. 28.

(35) Figure 16.. Simulation results showing effect of number of data flows on number of received data packets. Figure 16 depicts the effect of number of data flows on number of received data packets. In networks with one data flow GPSR-AEE achieves higher number of received packets by the destination node. This metric is drastically higher for networks with 5 data flows because of the increased traffic load. The number of received packets of GPSR in networks of 5 to 25 data flows is constant. This is due to the number of shut nodes depicted in Figure 30. In this situation nodes are shutting almost at a constant pace, which prevents transmission and receiving of packets. The number of received packets of GPSR-AEE in networks of 5 to 25 data flows is linearly decreasing as the number of shut nodes is linearly increasing (Figure 30). For a very high number of data flows (30) the number of received data packets is increased due to 5 extra source destination pairs that are generating and receiving packets. Accurate topology is maintained while saving a lot of energy. One of the main reasons for a higher number of received packets is the reduced energy consumption (Figure 27 and 28) of GPSR-AEE as many nodes in the network that are using GPSR are shutting down due to the energy loss. Therefore they are unable to transmit data. 30.

(36) Figure 17.. Figure 18.. Simulation results showing effect of network mobility on Packet Delivery Fraction. Simulation results showing effect of number of data flows on Packet Delivery Fraction 31.

(37) Not only was the number of received packets positively influenced by GPSR-AEE, packet delivery fraction is as well. Static nodes are showing better performance judging by Figure 17 and 18. But these results can’t be looked individually. If we look at Figure 15 again, we can see that the number of received packets when the speed is equal to zero is twice larger for GPSR-AEE than GPSR. Saying this, we can conclude that in this case the number of received packets is a more vital metric than the packet delivery fraction. We can also see that GPSR-AEE is improving packet delivery fraction for mobile nodes. The average hop-count is significantly lowered comparing to plain GPSR. More accurate topology is being maintained with the AEE mechanism, so the routing path is closer to optimal. Figure 19 and Figure 20 show that the number of nodes which are shutting down due to energy losses is drastically smaller using our mechanism. This provides the possibility for multiple options when making routing decisions, leading to lower hop count metric.. Figure 19.. Simulation results showing effect of network mobility on Average Hop Count. 32.

(38) Figure 20.. Simulation results showing effect of number of data flows on Average Hop Count. Average End-to-end Delay (Figure 21 and 22) and hop count should show some similarity in results as both of the metrics are mutually dependent. As mentioned before, our mechanism is managing more accurate topology of the network, thus achieving results of lower hop count metric. Lower hop-count metric of GPSR-AEE leads to lower End-to-end Delay than GPSR as packets are traversing fewer nodes. Hop-count represents the measurement of distance in a network, therefore the longer the “distance” that packets need to take the longer the End-to-end Delay.. 33.

(39) Figure 21.. Figure 22.. Simulation results showing effect of network mobility on Average End to end Delay. Simulation results showing effect of number of data flows on Average End to end Delay. 34.

(41) Figure 24.. Simulation results showing effect of number of data flows on Beacon Overhead. A reduced amount of beacon updates generates less collision in MAC layer. For higher traffic load and higher speed of the network GPSR-AEE sets beaconing interval to shorter value in order to maintian more accurate local topology. Beacon updates are then more frequent so more collisons occur. But still, GPSR-AEE produces up to 70% fewer collisions than GPSR as we can see in Figure 25. GPSR-AEE also results in up to 77% fewer collision than GPSR as we can see in Figure 26.. 36.

(42) Figure 25.. Figure 26.. Simulation results showing effect of network mobility on number of packet collisions. Simulation results showing effect of number of data flows on number of packet collisions 37.

(43) Energy consuption is lower using GPSR-AEE in slow and moderatly fast networks as well in the networks with lower number of data flows (Figures 27 and 28). For high speed and higher number of data flows, higher energy consumpiton is inevediable because of GPSR-AEE adapting nature.This not only saves total energy consumed, it also manages to reduce the number of powered off nodes.. Figure 27.. Simulation results showing effect of network mobility on Energy Consumption. 38.

(44) Figure 28.. Simulation results showing effect of Number of Data Flows on Energy Consumption. Powered off or shut nodes can’t transmit data, which thus lowers the total number of received packets by destination nodes. The number of shut nodes for slow and moderalty fast nodes is almost around six times smaller using GPSR-AEE comparing to GPSR in Figure 29. Similar performance is shown in Figure 30 as for a low number of data flows around 13 times less nodes are powering off.. 39.

(45) Figure 29.. Figure 30.. Simulation results showing effect of network mobility on number of shut nodes. Simulation results showing effect of number of data flows on number of shut nodes 40.

(46) Figures 31 and 32 shows the nodes which are powering off in reference to simulation time. We can see that nodes using GPSR are losing energy earlier in the simulation and therfore quite a larger number of them is shutting down for both slow and fast mobility. This is happening due to an unnecessary high amount of beacon updates that GPSR generates, therefore lot of nodes consumes all of its energy earlier comparing to GPSR-AEE.. Figure 31.. Simulation results showing effect of network mobility on number of shut nodes throughout the simulation time. 41.

(47) Figure 32.. Simulation results showing effect of number of data flows on number of shut nodes throughout the simulation time. 42.

(48) 5. Conclusion We have demonstrated the necessity to adjust the beacon interval used in geographic routing depending on mobility and traffic load of the network. Our Adaptive Energy-Efficient (AEE) beaconing mechanism deals with these problems. Beaconing interval is adapted based on the speed of the node as well as on the number of data flows traversing to each individual node. This method allows nodes to generate more beacons in case of high mobility and high data transmission. Furthermore, by letting nodes generate fewer beacons in case of low mobility and low data transmission the beacon overhead, energy consumption and collision rate all become lower. We have inserted AEE within GPSR and tested its performance in comparison with regular GPSR using comprehensive ns-2 simulations. Results are showing numerous improvements to network metrics. The number of beacon updates decreases while the packet delivery fraction and number of received packets increases. Additionally, energy consumption and collision rate are drastically lowered down comparing to regular GPSR. “Life” of the network is longer with GPSR-AEE, as nodes are having longer life-time to transmit data before losing all energy. Since our mechanism (GPSR-AEE) is maintaining more accurate topology of the network hence improvement in hop-count and end to end delay is achieved.. 43.

(50) [21]. Shao Tao; Ananda, A.L.; Chan Mun Choon, "Greedy Face Routing with Face ID Support in Wireless Networks," Computer Communications and Networks, 2007. ICCCN 2007. Proceedings of 16th International Conference on , vol., no., pp.625,630, 13-16 Aug. 2007. [22]. Ming Li, "Ripple Effect: an Improved Geographic Routing with Local Connectivity Information," Telecommunication Networks and Applications Conference, 2008. ATNAC 2008. Australasian , vol., no., pp.252,257, 7-10 Dec. 2008. [23]. Biao Zhou; Lee, Y.-Z.; Gerla, M., "“Direction” assisted Geographic Routing for mobile ad hoc networks," Military Communications Conference, 2008. MILCOM 2008. IEEE , vol., no., pp.1,7, 16-19 Nov. 2008. [24]. Rong Ding; Lei Yang, "A Reactive Geographic Routing Protocol for wireless sensor networks," Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP), 2010 Sixth International Conference on , vol., no., pp.31,36, 7-10 Dec. 2010. [25]. Shengbo Yang; Chai-Kiat Yeo; Bu Sung Lee, "Robust Geographic Routing with Virtual Destination Based Void Handling for MANETs," Vehicular Technology Conference (VTC 2010-Spring), 2010 IEEE 71st , vol., no., pp.1,5, 16-19 May 2010. 45.

(51)

The average earnings for dealers grew by 6.0% and 1.2% respectively over December 2007 and June 2008 to MOP13 947, and that for employees in positions such as hard and soft

To this end, we introduce a new discrepancy measure for assessing the dimensionality assumptions applicable to multidimensional (as well as unidimensional) models in the context of

 提供 Connection Oriented (連結導向) 並達成End-to- End (兩端通訊端點對端點) Process-to-Process (程序對 程序)、Reliable Data Delivery

Hence on occupation category, total manpower requirement for managers and administrators, professionals and associate professionals taken together is projected to grow at an

Mean saving of 13% of total Lighting Cost for Corridors or Cost Saving of \$42,500/ Month Average Vertical Daylight Factor for Living Room - Block 4. Daylight Energy Saving

Is end-to-end congestion control sufficient for fair and efficient network usage. If not, what should we do

◉ These limitations of vanilla seq2seq make human-machine conversations boring and shallow.. How can we overcome these limitations and move towards deeper