CHAPTER 5 PROPAGATION DELAY

98 CHAPTER 5 PROPAGATION DELAY Underwater wireless sensor networks deployed of sensor nodes with sensing, forwarding and processing abilities that operate in underwater. In this environment brought challenges, such as signal attenuation and propagation delay. Technology brings a broad range of applications for underwater sensor networks. UWSNs have new research paradigm that poses exciting challenges since intrinsic properties of the underwater environments but they suffer, (a) Large propagation delay The propagation speed of acoustic signals in under water has about 1.5 x 10 3 m/s. (b) Node mobility Underwater sensor network moves with the sea water current (up to 1 to 3 m/sec). (c) High error probability of acoustic underwater channels The underwater acoustic channels have very low bandwidth capacity (of the order of KHz) that depends on frequency and transmission range, suffers high bit rate error due to involving propagation delay. As the terrestrial communication based on Radio Frequency (RF) communication, all of these technologies assumed that propagation latency is negligible and have effectively factored out of design consideration. Under water communication generally used in acoustic technology communication. There are lot of basic differences between RF frequency and acoustic waves, such propagation delay as large and sensor node mobility due to current water. Thus, the protocol for terrestrial sensor network could not be directly applied to underwater acoustic sensor networks. Because underwater circumference is much varied from terrestrial and some of challenges need to address, while using

99 sensor networks as an effective communication technology for underwater level. Due to salty water and have high density, electromagnetic and optical waves cannot be transmitted for long distances in the ocean because of scattering, high attenuation and absorption effect. Acoustic communication can be used to dilute this problem, better means of data transfer which provides such a circumference. Hence, available propagation speed is compared from the speed of light to speed of sound that is five orders of magnitude slower, i.e. 1500 m/sec, which brings long propagation latency and propagation delay. Also, they consume more power and replacement of sensor node s battery is not so easy since the underwater sensor nodes larger in size. Underwater applications require multihop network where sensor nodes forward data to one of more sink node has located at the terrestrial level. Moreover, facing many challenges in under water wireless sensor network for the applicable of emerging application, necessary to solve the propagation delay in under water wireless sensor networks. The aim of this research work have to understand the impact of propagation delay on medium access with ALOHA protocols as the case study. First, consider that the area-dependent propagation delay has a fundamental impact on the slotted ALOHA since a packet receiving time at the receiver depends not only on its forward time but also on its relative propagation delay to the receiver. Moreover, that both dimensions of uncertainty need to handled at the same time. Then, investigate the propagation delay tolerant ALOHA protocol, to improve the performance of the slotted ALOHA by adding guard bands. That the performance of classical slotted ALOHA deteriorate in an underwater setting where transmissions from one slot can overlap with future slots. Referred to their scheme as the propagation delay tolerant ALOHA protocol. In this chapter, discussed about ALOHA protocol, the guard-band in each slot is designed, to overcome collisions between packets of different senders forwarded in consecutive time slots. It is necessary to reducing collisions between the data packets while increasing the guard-band but it also increases

100 the length of each time slot and generally reducing the achievable throughput of the sensor node. To get the maximum achievable throughput has equated an optimization problem and thus it is essential to selection of the appropriate guard-band length. Moreover, analyzed analytically, the throughput of the ALOHA protocol given a guard band, a congested load and the maximum propagation delay. And also achieved a number of interesting and useful properties concerning the performance of the ALOHA protocol with guard band. 5.1 ALOHA with Space-time Basic class of ALOHA is the MAC protocols that do not try to prevent of data packet collision. The protocol works as follows, every network that has sent the data. If the collision occurred while two sensor nodes forwarding the data packets at the same time slot. ALOHA have improved by having discrete timeslots, in case of retransmission happening at the sensor nodes. The sensor network have no long send packets at anytime but just at the beginning of a timeslot and thus chances of collision reduced. In this version referenced as Slotted ALOHA. This chapter, focused on the impact of the long propagation delay of acoustic signals on those protocols. ALOHA in ground communication, that a packet sending time t will collide with other packets are sending in time [t- 1, t+1]. This can visualize in Figure (5.1). The figure shows three stations that have to send the packets. The first packet after time t - 1, the second at time t and the third before time t + 1. Overlap packet since all the part of another transmission failures, in underwater acoustic communication have looked at ALOHA using a similar approach. Then forwarding of a packet is successful and if the packet does not collide with another packet at the sink node. The difference that propagation delay to emphasize can cause look at an intuitive example. It is based on the similar transmission explained in Figure (5.1). And took the time t- 1, the first node forwards the message. At time t, the second node again forwards the

101 message. Moreover, also the first message keeps propagating. Both signals have eventually met, form constructive and destructive waves. After that, the signals have kept going in that direction. At time t + 1, the signals send by node 2 will continue to propagate and the signal from message sent by node 1 have reached node 2. Since the message reaches node 2 without collision, the message is considered to be successful. Note that if it radio signals, this would result in a collision. Figure 5.1 Collision of ALOHA 5.1.1 ALOHA with Time Uncertainty First analyzed of pure and slotted ALOHA protocol, where sensor nodes immediately forwarding message packets. Thus only considering temporal uncertainty, assuming collisions with respect on the transmission time alone. As the collisions takes place among the sensor nodes, then forwards the messages simultaneously, and that matter only at an intended receiver, this analysis has an implied assumption of no propagation delay. If propagation delay is significant, this analysis then requires assuming transmitter as equal distance at the receiver end. And then further assumptions made the large number of sensor nodes, with all arriving packets served at a new node and packets forwarded immediately into the sensor network. The data packets that collide and are filled with buffered block, making sensor nodes backlogged. Such backlogged sensor nodes

102 retransmit after an exponential propagation delay. The total offered load to the network has the combination of the Poisson arrival and backlogged exponential retransmissions. In this results have the combined Poisson packet arrival process (with mean ) to the sensor network having normalized throughput (expected packet /unit time) where represents the number of backlogged nodes in the network. The vulnerability interval (VI) defined as the time interval relative to a sender s transmission within the sensor nodes transmission made the collision. Assuming packet transmission time, Figure (5.2) shows that the is equal to. On the other hand, the transmission allows slotted ALOHA transmission only at the start of synchronized slots of length. As shown in the Figure (5.3) ensured synchronization that only hindering packets that arrive in slot 0 have result in a packet collision. Thus reduces the from to and preventing any cross-slot overlap. As this analysis using the concept of vulnerability interval shows that slotted ALOHA achieved maximum normalized throughput of 1/e with of 1 packet per slot, while pure ALOHA achieves its maximum of 1/2e at 0.5 packets / slot. Figure 5.2 ALOHA

103 5.1.2 Slotted ALOHA Slotted ALOHA is a minor modified version of ALOHA, discrete time slots introduced and then time slots divided into equal length slots. The data message is sent only at the initial stage of the time slot. In case of giving a formal, to present an intuitive explanation of the performance of Slotted ALOHA in under water area. To synchronize the sensor nodes so that they could implement Slotted ALOHA and assumed that there is ways again have to consider the propagation delay. Instead of looking at the time to send the packet and have to look at the time the packet received at each node. There is no guarantee that they will arrive in time slots although the sensor nodes send the messages in pre-defined time slots. Therefore, Slotted ALOHA in underwater has no effect different from ALOHA except the cases where the propagation delay has the multiple of time slot interval. When divided by the sound speed, in other words, the distance between sensor nodes, results in an integer of time slots. All the sensor nodes randomly placed that the probability, in this scenario is zero. Figure 5.3 Slotted ALOHA

104 5.1.3 Space time Acoustic underwater wireless sensor networks characterized by increased uncertainty in medium access, due not only when data sent, but also due to significantly different propagation delays from spatially diverse transmitters, this is known as space time uncertainty. In this chapter summarized the concept of space-time uncertainty with regards to medium access. Propagation delay is negligible while channel states in short-range RF frequency networks have estimated quickly. The area of the receiver and potential interferers makes the large propagation delay of acoustic media and it is essential for also considered. The distance between sensor node forwards into uncertainty of current global channel status of Space-Time Uncertainty. Presents the systematic description of this principle and its impact on contention based medium access. Consider the Figures (5.4&5.5) the two concurrent transmissions from A and E is received separately at nodes B and D have collided at node C. Figure 5.4 Space-time uncertainty - without collision

105 Figure 5.5 Space-time uncertainty - collision at node B In this scenario, that packet dropping collision and packet reception in slow sensor networks depends on both transmission time and receiver location area. When the similar collision of sensor nodes where this space-time uncertainty have also viewed as a duality. Although, in principle, this uncertainty occurs in all communication, it is only significant where propagation delay is very high. 5.2 Analytical Analysis of Propagation Delay Analyzed the normalized throughput of the ALOHA protocol and addition of additional guard band beyond the transmission time in time slots, so the Space-time uncertainty have handled and minimizing the propagation delay in the underwater level. It has needed to tolerate the large propagation delay, these additional guard band added to ensure a single slot overlap at the receiver end. First look at the time slot, hence each time slot contains the transmission time and a guard band following the former. Since the guard band of the size of maximum propagation time should eliminate all the collision between different time slots, it does not make sense to have the guard band whose size is more than only decreasing the throughput without any gain. In this

106 modification to slotted ALOHA, sensor nodes still transmit only at the start of globally synchronized slots. Underwater time synchronization protocol have achieved using Global time synchronization. Hence, using the normalized factor in expressing the size of guard band. The slot duration, however, has increased from, where represents the fraction of maximum propagation delay ( ) that sensor nodes wait after finishing their transmission as shown in the Figure (5.6). Hence, is the guard band and considered as the normalized guard band. Choosing ensures that no overlap at the receiver, unless data packets forwarded in the same slot, the guaranteed that slotted ALOHA is generally designed to achieve when the delay is not important. However, this value of results in a long wait time after each packet that has increase the packet transmission latency and bandwidth overhead. When there remains the possibility that some sensor node pairs still have the vulnerability interval of duration of two slots. Therefore, the value of reducing lowers the bandwidth overhead but increasing probability of collision among the sensor nodes. Based on the distance between two sensor node pairs is generally smaller than the maximum propagation delay and vary to evaluate the trade-off between bandwidth overhead and probability of collision. The speed of propagation, the communication has a positive finite constant regardless of the area in the sensor network, so that the maximum propagation time is from the receiver to the farthest transmitter. The transmission rate is constant for every transmitter and also the data packet size is constant so that the transmission time for a packet has constant. Only a proper scaling is needed for some notations, particularly to cope with the general transmission time. Hence, the normalized maximum propagation delay a to the transmission time is.

107 5.2.1 Slotted ALOHA Figure 5.6 Timing diagram of Packet transmission To analyzed the normalized throughput, and derived the expected number of successful packet receptions in a particular time slot. Moreover, then using the linearity of expectations and probabilities of conditional to calculate the expected number of data packets delivered from the source node. Let the indicator variable i i denote whether or not the receiver receives the packet from i-th transmitter successfully in the time slot. { Where r denotes receiving packets { } { } Where N denotes random variable of the successful reception. denotes the event that no collision occurs given that i- th sender transmits.

108 5.2.2 Without Collision Probability The packet collision depends not only on the temporal uncertainly, but also on the special uncertainty. If more than one sensor node forwarding data packets in the same time slot, the data packets collide with each other regardless of the area of their transmitters (when a < 1). If two data packets are transmitted in various time slots collisions have occurred and depending on their sender s area. The former collision has intra-timeslot collision and the latter inter-timeslot collision. It turns out of the system have at most three collision regions for each and every transmitter, one for the intra-timeslot collision and two for the inter-timeslot collision. Let us suppose an arbitrary transmitter which has the normalized time distance of and have identified the regions. The region such that a data packet sent from would collide with a data packet if it has sent in the previous consecutive time slot by the node in the region, known as the first collision region. Similarly, the data packet from would collide with a data packet sent in the same time slot by another sensor node in the second region, and the third region has for the collision with a packet in the next consecutive time slot. Moreover, then denote the three collision region by respectively, generally these three region depends on the distance of the source transmitter from the receiver. Moreover, they also depends on the normalized guard band (β). Consider a simple sensor network with two sensor nodes A and B and one receiver R. The node A locates right next to R while node B is very far from R and the size of guard band has small enough. Then, if node B transmits in the i-th time slot, R would receive last part of the packet in the beginning of the (i + 1)-th slot, produced collision with the packet transmitted in the (i + 1)- th slot by node A although the two data packets are sent in different time slots. The time diagram in Figure (5.3) visually shows this situation, where is the normalized propagation time distance of { } from R defined. The normalized guard band size β is less than 1 and. However, if

109, without collision between data packets in different time slots. Therefore, that the collision depends on sensor nodes area and two packets transmitted in different time slots have experience collision between each other. Definition 1 The propagation delay of sender X from the receiver is the propagation time from the receiver to X divided by the maximum propagation time in the network. i.e., where d is the distance between sources and sink node. The probability of without collision given a data packet sent by an arbitrary i-th sender is then as follows conditioning on the s normalized time distance, after calculation note that the regions have expressed in terms of normalized time distance and guard band. The region is where the normalized time distance from the receiver has at least but no more than 1. That is, { }, where denotes the normalized time distance of the point P from the receiver. Similarly, have { }, and { }. Figure (5.7) visually presents the regions. { } { } { } Where is the area of, is the probability density function of the normalized time distance of from the receiver given that transmits, and is the noted representation of the event that without collision occurs given transmits and its normalized distance from the receiver is. The last equation holds because the area of the sensor node independent of the data packet transmission and the probability of without collision of a specific

110 data packet depend on the area of its transmitter. The three collision regions as shown in the Figure (5.7) areas affect the probability. Figure 5.7 Network Collision { } {{ }} Where Nn, Nc, and Np denote the number of other transmitters in the collision regions respectively. Note that (i) since the intra timeslot collision region is the whole area assumption and (ii) there are (n 1) other transmitters in total since focused on one specific transmitter s success. The probability of conditional turns, without collision turns out of involve the binomial series as follows. { } { } } { } ( ) { } Substituting equation (5.10) in the equation (5.6) and can obtained the equation for the probability of no collision which have evaluated easily with the numerical method. Thus, the equation for probability of without collision does

111 not involve the maximum propagation delay implying the probability is independent of iso that the success rate is also independent of. Suppose a network of sensor nodes with fixed spatial area of sensor nodes, a fixed transmission probability in a time slot for each node i, and a transmission time T for a packet. Then, maximum propagation time is independent of the success rate f of the network as long as. In other words, propagation speed is independent. The number of sensor nodes in each of, and is constant regardless of the propagation speed, and so the probability of without collision of the i-th transmitter is constant. Therefore, { } 5.3 Performance Analysis of UWSNs The network has one receiver and 200 nodes of transmitters which are deployed in the 2-D disk area with the measurements of 5x5m 2. Table 5.1 Characterization of UWSN model Symbols a p N Nc β T F S Definition Propagation time Normalized maximum propagation delay Probability in each time slot Number of successful reception Event no collision Guard band Normalized time distance Transmission time Success rate Throughput Propagation speed

112 Table 5.2 Parameters list of UWSNs Symbols Definition Value N Number of nodes 200 - Transmitting power -20dbM - Transmission rate 250kbps - Packet length 125bytes T Transmission time 1s/kbps The receiver locates at the center of the disk area and the transmitters deployed uniformly at random in the area, maximum transmission range is 30m long. The speed of propagation in communication has positive finite fixed regardless of the area in the sensor network, so that the maximum propagation time from the receiver to the farthest transmitter has a positive finite constant. The rate transmission has fixed for each transmitter i.e. 250Kbps. The packet size is fixed so that, along with the fixed rate of transmission, the transmission time for a packet is constant T. The traffic in the network has I.I.D. Bernoulli, so that a transmitter forwards the packet to the receiver with probability p in each time slot. The transmission time T is less than the maximum propagation time. And also consider the throughput S in packets per packet length. Since the expected number of packet receptions are successful in a time slot, independent of the propagation time as long as it is positive finite, S have expressed after introducing a new variable as follows as the ratio of the maximum propagation delay to the transmission time of a packet. { } Using the numerical evaluation of Equation (5.12). Characteristics of the throughput depending on the size of guard band in the Figure (5.8)) the relative maximum propagation delay t is fixed, but the number of nodes N is

113 varying from 20 to 100 with the interval of 20. In the Figure (5.9) implies, fixed N but t is varying from 0.2 to 1 sec with the interval of 0.2 sec. These graphs shows that the throughput responds to the variables, the optimizer β values are similar of the case but different in the other case. Analyzed the results of optimal throughput of propagation delay- ALOHA obtained in previous section to observe the effect of guard band and network delay regime. Furthermore, for comparison and simulate propagation delay-aloha to verify the correctness of analysis. Moreover, then first introduce the notations of the simulation used for comparison and then focus on the results. Figure 5.8 Throughput of packets Simulation using a custom-built, packet-level simulator designed for UWSN MAC model. Simulation scenario consists of a single receiver that does not transmit with nodes randomly deployed in a circular region with a radius equal to the maximum propagation delay. Sensor nodes and the single packet buffer transmit based on an offered load to the network modeled as a Poisson

114 process, with mean ranging from 0 to 3 packets/transmission time, packets successfully received at the designated receiver and it has observed. Moreover, choose a single receiver to parallel, analysis of protocol behavior but have verified that the results holding with packets reception at other nodes in the network. Figure 5.9 Number of Successful packets Protocol performance evaluated through throughput normalized to the channel bandwidth. Simulations were run with 200 nodes unless otherwise noted. Moreover, analysis, then used a packet length of 125 bytes, resulting in a transmission time of 1 second (at 1kb/s) to normalized throughput. Moreover, also assumed a constant speed of sound as 1500m/s. After altering the maximum range to simulate different delay regimes. Each simulation data point is the averaged result of 25 simulation runs with error bars showing 95% confidence intervals. Look at the maximum achievable throughput (throughput capacity) that propagation delay-aloha have achieved (at an optimal offered load) as a

115 function of guard band length, which is a fraction of the maximum propagation delay. Figure (5.10(a)) represents the function of as the throughput capacity for a fixed N (200 nodes) using numerical methods for maximizing over p. Moreover, the graph, response for different delay regimes characterized by different values of a. Figure (5.10(b)) shows the graph for the same notations. However, here instead of using analysis and derive the results from empirical data collected from simulations. As the results from both simulation and analysis complement each other. Both results shows that throughput capacity of a network have increased by using propagation delay- ALOHA and that the benefit of the guard band highly correlated to its size and the delay regime in which the network is operating. Figure 5.10a Throughput (Analytical results)

116 Figure 5.10b Throughput (simulation results) Then observed two trends as increases. First, with very small a = (e.g. 0.01 in simulation results) and shows the throughput increases (approaching the optimum) as larger guard band used due to a decreased inter-timeslot collision probability. Conversely, with large a (e.g. equal to 1 when the propagation delay equals transmission time), the throughput becomes insensitive to the use of guard band. Furthermore, simulation result shows that for any value of a beyond 1, the benefit of choosing additional guard time diminishes. Thus, choosing a packet length that normalizes the propagation delay to an appropriate value is essential to yield the benefits of propagation delay-aloha. Varying the value of a, to observe, how the throughput capacity is affected by propagation delay in propagation delay-aloha. As per the result of Figure (5.11a) gives, that throughput capacity as a function of the normalized maximum propagation delay a when the guard band β is given and fixed. They are obtained for N =50 maximizing in equation (5.7) over p with given β and a.

117 Comparing the results, and have the same graph generated from simulation results in the Figure (5.11b). The response using different values of β. It has seen that a fixed value of β might lead to a suboptimal throughput. When β=0.5, propagation delay-aloha is closest to the β-optimal curve when a is near 1 but the gap increases as a goes to 0. Conversely, for β=1 propagation delay-aloha is closest to the β -optimal curve for smaller values of a but becomes inefficient as a approaches 1. Although the throughput decreases monotonically with increasing values of a, it is observed very little sensitivity to a with smaller β values. This insensitivity is due to limited collision prevention provided by the shorter guard band. Also, the monotonically decreasing slope increases with β causing throughput to become more sensitive to a. Figure 5.11a Maximum throughput (Analytical results) Simulation and analytical results from the Figure (5.11a & 5.11b) shows that propagation delay-aloha achieved about 19% (when a= 1) to 90%. When a = 0, improvement on throughput over slotted ALOHA in an underwater

118 environment. Figure (5.11a&b) represents the normalized throughput regarding the maximum propagation delay a. Moreover, the maximum throughput changes in terms of the guard band. From the numerical analysis that, given a guard band in [0, 1], the maximum throughput have been obtained with (hence, a = ). Figure 5.11b Maximum throughput (Simulation results) Table 5.3 Consolidated results for propagation delay Propagation delay in sec PD Analytical Normalized Throughput PD Simulation Normalized Throughput β =1s β =.5s β =.1s β =1s β =.5s β =.1s 0.2 0.38 0.32 0.25 0.2 0.39 0.33 0.25 0.4 0.34 0.30 0.25 0.4 0.33 0.30 0.24 0.6 0.29 0.28 0.25 0.6 0.30 0.29 0.23 0.8 0.26 0.26 0.25 0.8 0.28 0.28 0.23 1.0 0.25 0.25 0.25 1.0 0.26 0.26 0.24 1.2 0.24 0.24 0.25 1.2 0.25 0.25 0.24

119 5.4 Summary In this chapter, have exploring the impact of spatio-temporal uncertainty on under water wireless Sensor Network MAC protocols. For such networks, shown that the location- dependent acoustic propagation delay significantly affects MAC protocols such as slotted ALOHA. Thus, it is necessary to consider both space and time uncertainties while designing MAC protocols under varying latency environment of an acoustic UWSN. And modified slotted ALOHA by adding extra guard band in each slots. Then have investigated different metrics of performances, expected number of successful packet receptions in a guard band slot, throughput and maximum throughput. After that obtained exact equations for the number of receptions and throughput in terms of well-known functions using both analytical and protocol simulations. These results given that the throughput capacity of propagation-aloha is 19-90% better than that of simple slotted ALOHA in an underwater environment. Then shows that for the optimal throughput capacity the value of optimal changes based on operating delay regime. This argues for deploying dense, short range, multi-hop networks as opposed to sparse and long range networks currently used in underwater networks. However, this chapter of research work focused on capturing the impact of latency on ALOHA-like protocols and understanding the mechanics of underwater medium access.