Signal Strength Sensitivity and Its Effects on Routing in Maritime Wireless Networks

Transcription

1 Signal Strength Sensitivity and Its Effects on Routing in Maritime Wireless Networks Chee-Wei Ang and Su Wen Institute for Infocomm Research, A*STAR (Agency for Science, Technology and Research) 1, Fusionopolis Way, #21-1, Connexis, Singapore {angcw, Abstract Unlike terrestrial networks, the received signal strength in maritime wireless networks are subjected to perturbations due to the sea movement. Surface motions at sea cause variations in antenna heights and orientations of the communicating node (e.g. ships and buoys), thus affecting the received signal strength. In order to minimize such undesirable effects, we carried out a detailed study of the relationship between sea waves and the received signal strength. The effect of antenna gain variations due to tilting of antenna masts were studied by modeling sea waves. The channel path-loss variations were studied using a two-ray path-loss model for the maritime environment. Our analysis showed that the sea wave movements affect both short and longer links. We proposed a simple scheme to avoid the use of unstable links and implemented it on the QualNet network simulator. We carried out IEEE based mesh network simulations with sea terrain and path loss models. We observed performance improvements in packet delay and throughput when the stable link selection scheme was used. I. INTRODUCTION While high-speed networks have brought Internet connectivity to our daily lives, such ease of communication has not been extended to oceanic ships traveling world s waters. Authorities along major water ways have shown interests in extending terrestrial networks to the sea, e.g. along coasts or from ports. The WISEPORT [1] service, in which the network provider extends the network coverage to ships along coasts and ports in Singapore, is one example. The TRITON project [2], which investigates the design of a WiMax-based wireless mesh networking between ships (especially in ports and along straits where there are many ships), is another example of many maritime wireless communication efforts. The ships can connect to the terrestrial network through the multihop network to gain access to the Internet. The ships can also use this network to do ship-to-port communications, such as checking in and out on a port, or carry AIS data which are currently transmitted via VHF. Currently maritime communications have largely relied on the VHF band. The main reason for using this band is that path loss is smaller at this frequency, thus signal can reach further with the same power (i.e. EIRP, the Effective Isotropically Radiated Power), as compared to a node using a higher frequency band. A corollary is that a node can cover the same distance with a smaller EIRP. Smaller EIRP can be translated to the use of a lower gain antenna. And lower antenna gain in turn means wider antenna beamwidth. Wider antenna beamwidth is advantageous in the sea environment because sea motion can cause significant fluctuations in directed antenna gain if a narrow beam is used (to be discussed in Section II). There are mechanical and electronic antenna stablization solutions designed for ships to counter antenna tilt caused by sea motion [3]. However such stablization systems are bulky and expensive. In this paper we investigate the case where no antenna stablization system is used along with the use of 2.4GHz frequency band in maritime environment. The reason for using a higher frequency band is that bigger bandwidth can be allocated for high-speed communications. Maritime wireless communication environment is different from the land-based environment in that nodes (e.g. ships or buoys fitted with communication equipments) are subjected to movements caused by sea waves. The sea movement causes variations to the communication antenna s height and orientation, which affects the antenna gain and the quality of the link. Variations in antenna height affect the angle of reflection on the sea surface, which in turn affect the pathloss experienced by the transmission. These effects significantly reduce the stability of wireless links in the maritime environment. It is thus useful to study how significant each of these parameters affect the received signal strength. Instability in received signal strength can cause frequent link breakages. If the mesh network is formed using a reactive routing protocol such as AODV, link breakages will trigger route repairs or route re-discoveries, which generally involve additional control messages flooding the network. Broadcast of control messages is expensive as it takes up a lot of channel bandwidth. Frequent route discoveries can thus reduce system throughput significantly. Route repair and re-discovery also increase end-to-end packet delay as it takes time to set up alternative routes 1. Note that proactive routing schemes such as OLSR also do not work well under frequent link breakages because network information (which are periodically distributed) can be made stale by intermittent link breakages. Clearly it is important to understand the stability of shipto-ship communications in order to construct stable multi-hop networks at sea. This paper attempts to identify the maritime link characteristics that have significant effects on network 1 Although our discussions are in the context of mesh networks, the study is also applicable to point-to-multipoint networks, since providing link quality stability is essential in fulfilling service level agreements /8/$ IEEE 192

2 routing performance. We then propose a simple algorithm to minimize these effects. The IEEE mesh MAC [4] was implemented with link filtering to present only stable links to the AODV routing layer 2. With the effects of sea motion minimized or removed, we can then address other channel impairments using existing techniques. Our simulations showed that these efforts led to increase in packet delivery ratio and reduced average end-to-end packet delays. The main contributions of this paper include the study of sensitivity of the received signal strength with respect to antenna tilt and effective antenna height, and the design of a simple link selection scheme in selecting stable links in maritime wireless networks. II. EFFECTS OF SEA WAVES ON RECEIVED SIGNAL STRENGTH Sea waves caused ships and buoys (and thus the antennas placed on them) to move in various ways. Ships can be tilted in all three dimensions: pitch and roll; and rise or fall to the waves and tides. Figure 1 illustrates these movements, which cause variations in the received signal strength. To study the various factors affecting maritime communication, we adopt several models that are representatives of key features in the actual systems. A. Characterization of the Sea Motion The condition of the sea surface movement can be characterized by the Pierson-Moskowitz sea states [6], which classify the sea conditions in 1 different levels. Table I shows the wave values associated with the ten sea states. Each level is further divided into two to four sublevels, which is not included in the table. Sea state levels -3 indicate generally calm sea conditions, with waves less than 2 meters high. Sea states 4-5 are moderate sea conditions, with wave heights around 2 to 4 meters. Sea state level 6 and above generally indicate severe sea conditions, with wave heights going up to near 3 meters. In our simulation study, we used the values associated with each sea state level and sublevel to model the wave s effect on the movement of the antenna. TABLE I THE PIERSON-MOSKOWITZ SEA STATE TABLE Sea State Significant Avg. Period Avg. Wave Level height (m) (sec) Length (m) Fig. 1. Variations in received signal strength cause by sea waves Note that because distances between ships are long relative to the antenna heights, changes in antenna gains caused by variations in antenna heights is small. Variations in received signal strength are more significantly affected by the antenna tilt. There are work done on characterizing maritime channel fading. Reference [5] studied the effect of reflections cause by sea waves on fading. They carried out actual measurements and found that the reflection had a coherent component with variations caused by the random sea surface. They proceeded to characterize the propagation characteristics with Ricean fading model. While their work focused on the computation of the received power at the propagation level using a sine-based wave model, our work centers on the effect of sea wave to MAC and routing protocols in the communication network. 2 Although AODV is used as the routing protocol in this paper, the work is applicable to other routing protocols since it is always good to use stable links to form routes. B. Pathloss model The sea surface presents itself as a large reflector of radio waves. As such the two-ray pathloss model seems to be an appropriate model for our study. Our experimental measurements carried out at sea around Singapore indeed showed that the two-ray pathloss model accurately represents what was observed in the field [7]. The two-ray pathloss model is given by: ( L(h t,h r,d) = 1 log λ 2 (4πd) 2 ( 2 sin( 2π λ ) ) 2 h t h r d ) db (1) where L is the pathloss in db; d is the distance between transmitter t and receiver r; h t and h r are the effective heights of transmitter t and receiver r respectively; λ is the wavelength of the radio transmission in the same dimension as the antenna height and distance. Note that as the distance becomes large, the angle in the sine term become less than π ht hr 2, thus d>4 λ, the value 193

3 of the sine term becomes monotonic, and the pathloss value does not fluctuate any more. At a distance much greater than 4 ht hr, the sine term can be estimated to its inner term. λ An example plot of two-ray pathloss and free-space pathloss versus distance is shown in Figure 2. path-loss (db) comparison of path-loss models two-ray free-space strength. The received signal strength of a node can be expressed by: p r = p t + G t (θ tr ) l t L(h t,h r,d)+g r (θ rt ) l r (3) where p r and p t are received and transmitted power in db; G t and G r are antenna gains in db, which are functions of orientation θ tr and θ rt respectively; θ tr represents the elevation angle of transmitter t antenna towards receiver r; likewise, θ rt represents the elevation angle at receiver r towards transmitter t. Note that antenna gain with respect to azimuth is unchanged in horizontal omnidirectional antennas, l t and l r are other losses such as cable loss, mismatch loss and antenna efficiency expressed in db distance (m) Fig. 2. Two-ray pathloss model (h t = h r =18m, λ =.125m) versus free-space pathloss. A. Antenna Gain Sensitivity Since sea movements affect the antenna gains (equation (2)) and pathloss (equation (3)), we can determine the sensitivity of received power p r with respect to sea state parameters by taking first derivatives: C. Antenna model In this study, we assume that all antenna used are omnidirectional in the horizontal plane. The gain of the main lobe of the antenna pattern can be approximated by [8]: G(θ) =G 12(θ/α) 2 (2) where G is the boresight gain, α is the beamwidth (the angle between -3dB points of the main lobe), θ is the antenna angle from the normal. Figure 3 plots the main lobe for an antenna of G =, and beamwidth of 45 o. For an omnidirectional antenna in the horizontal plane, equation (2) can be used to approximate antenna gain in the vertical axis. normalized gain (db) Main lobe approximation beamwidth = 45 degrees angle from normal (degrees) Fig. 3. Main lobe approximation with α =45 o. III. SENSITIVITY ANALYSIS The received signal power is affected by the transmission power, the gains of the transmission and reception antennas, and the loss experienced on the transmission path which includes path loss and other factors attenuating the signal p r (s) G t θ tr (s)+ G r θ rt (s) θ tr θ rt L (h t h r ) (h th r )(s) where s represents the sea-state. Note that sea state s {h s,p s,l s } where h s is mean wave-height in m, p s is mean wave period in seconds, l s is mean wavelength in m. θ tr and θ rt are functions of ship dimensions and weights, and instantaneous values of l s, h s. h t and h r depend on h s at transmitter t and receiver r. Sensitivity of received power with respect to antenna tilt can be expressed as: G t θ = 24(θ/α2 ) db/degree (4) Equation (4) shows that the variation in received power within beamwidth is within ±.267 db/degree. As long as the wave movements do not cause the antennas to tilt more than the beamwidths, the antenna gains G t and G r do not change much (i.e. within 3dB), as evident from the antenna plot in Figure 3. For example, with sea state 6, a distance of 1m, ship dimensions of 64m by 8m, antenna height of 16m, the maximum antenna tilt can be estimated to be about 1 o. The maximum variation in antenna gain is about 1dB. Therefore the maximum variation in received power p r caused by antenna tilt at both transmitter t and receiver r is about 2dB. However, if higher antenna gain is used to increase rangeper-hop, this gain-variation component will have bigger effect. This is because higher gain corresponds to smaller beamwidth α; and smaller α in turn increases sensitivity. Note that since it is desirable to have omni-directional antennas (in the horizontal plane) in mobile mesh nodes, the straightforward way to increase antenna gain is to focus the beam vertically. 194

4 B. Pathloss Sensitivity Next we study the sensitivity of the two-ray pathloss with respect to sea movement. Differentiating equation (1) with respect to antenna heights, we get: L (h t h r ) = A B (h t + h r ) cot(b h r h t ) (5) where: A = 2 2π,B = ln 1 λd path-loss (db) path-loss shift with respect to antenna heights variation dB antenna heights = 18m antenna heights = 19m distance (m) sensitivity of path-loss at effective antenna heights = 18.5m 8 7 sensitivity (db/m) path-loss (db) 2 Fig. 5. Shifting of loss spikes when antenna heights change from 18m to 19m at distance = 57m. 6 1 δ path-loss / δ height (db/m) Fig distance (m) Sensitivity of two-ray pathloss versus height, h t = h r =18.5m. A plot of equation (5) is shown in Figure 4, with h t = h r. With transmitter and receiver effective antenna heights at 18.5m, and distance of 56m, pathloss/ height is about 13dB/m. In other words, a variation of 1m in antenna height in both transmitter and receiver will lead to variation of 13dB in pathloss. Note that the variation in pathloss is caused by the shifting of the loss spikes and not by the change of magnitude in the spikes. A zoomed-in view of the spike shifting is shown in Figure 5, at distance of 57m pathloss changes by about 16.52dB with just 1m increase in effective antenna heights at both transmitter and receiver nodes. In comparison, the pathloss variation with respect to effective antenna heights is about 1dB at further distances. Note that at sea state 6, wave heights can go up to 4m. The change in pathloss at this region around the spikes are thus significant and links that are of these lengths should not be used. IV. A SIMPLE LINK SELECTION SCHEME Analysis in Section III showed that there exists a region where received signal strength fluctuates significantly when a slight change is experienced in antenna heights. These regions occur when the distances between two nodes is below 4 ht hr λ (see section II-B). We call these regions null regions where loss spikes can be experienced even under calm sea conditions (e.g. wave heights fluctuate around 1 meter). Another region where a link connection may fluctuate frequently (i.e. linkage goes on and off) is when the link length is at the edge of the transmission coverage (i.e. where SINR is just above the threshold). Even though our analysis shows that antenna tilt caused by sea waves caused relatively smaller fluctuations in 1 path-loss (db) received signal strength, the small variation of received power can cause the received SINR to fall below the threshold near the edge-of-coverage; which would render the link unusable. Since frequent breakage of links cause control signaling overheads and decrease network performance in terms of throughput and packet delay, it is thus beneficial to avoid the use of links in the null and edge-of-coverage regions when constructing wireless networks in the sea environment. These links can be identified and filtered out before presenting them to the routing layer. A. AODV Characteristics AODV [9] is a reactive ad hoc network routing protocol, applying a shortest path routing algorithm. The route length is measured by the number of hops. AODV discovers the route to a destination by sending route requests (AODV RREQ) towards the destination, multiple route replies (AODV RREPs) may be received from the nodes with route already established. The AODV RREP with the smallest hop count is chosen as the candidate route. As a result, AODV will choose a route with fewer, but longer links than a route with more numerous but shorter links. Unfortunately, longer links are more likely to be the links within the edge-of-coverage region. Therefore, our link selection scheme helps AODV to choose more stable routes by not presenting to the routing layer such long links that are close to the edge of coverage. B. Link Selection Algorithm The proposed link selection scheme works as follows. Algorithm 1 is invoked to filter out the unstable links in the null and edge-of-coverage regions at the MAC layer. Only AODV route discovery packets (i.e. AODV RREQ and AODV RREP messages) are subjected to this filtering. Other AODV control messages, such as RERR, which is an urgent message sent only once to inform other nodes that a link has broken, is not filtered. Algorithm 1 calls Algorithm 2 just before the received packet is passed to the routing layer for processing. By discarding AODV messages at MAC layer, AODV will be oblivious to the existence of these links, thus the routes chosen will only be formed by a set of more stable links. 195

5 SINR threshold is the minimum SINR required to attain a given BER (bit error rate) for the modulation and coding used for the transmission, and is determined by the physical layer. margin is the peak-to-peak fluctuation in antenna gains due to tilting of antennas caused by sea waves. Algorithm 1 receive packet from PHY(src, sinr) 1: in: transmitter node src, received sinr sinr, packet pkt 2: if (packet type(pkt) == AODV RREQ or AODV RREP) then 3: if (to discard packet(src, sinr) == TRUE) then 4: free(pkt) 5: return 6: end if 7: end if 8: pass packet to upper layer(src, pkt) 9: return Algorithm 2 to discard packet(src, sinr) 1: in: transmitter node src, received sinr sinr 2: out: TRUE: to discard; FALSE: do not discard 3: if (distance(src, self) < 4 ht hr λ ) then 4: return TRUE 5: end if 6: if (sinr < (SINR threshold + margin)) then 7: return TRUE 8: end if 9: return FALSE In practical implementations, we can assume some value for the effective antenna height, e.g. 2m. The variations in effective height caused by sea motion do not significantly affect the size of the null region. We also assume that the ships are equipped with the Automatic Identification System (AIS)[1], which is required on all large ships. AIS provides a way for ships to exchange their GPS location information, from which the inter-ship distance can be computed. We would like to qualify that the link selection algorithm does not have to be implemented at the MAC layer. For the MAC layer to filter packets, it needs to probe into the received packets to identify the route discovery packets (in this case it is AODV specific). It is also possible for the routing layer to identify the signaling packets to the MAC layer before transmission, so the transmitting MAC can tag routing layer signaling packets for them to be easily identified at the receiving MAC. One way to implement the algorithm without MAC layer filtering could be to export the received SINR of the routing layer message to routing layer as a metric for link selection. C. Additional application Raising the SINR threshold in the link selection process can also be used to mitigate interference from concurrent transmissions by hidden terminals. For example, when margin is set to SINR threshold, links that can be interfered from transmissions two or more hops away will not be passed to the routing layer. Consider the case where nodes T and R are neighbors (i.e. they can receive each other s transmission when no interference is present), but node I is not a neighbor of node R or T, i.e., the received power from node I at R or T is below SINR threshold. Since at node R, the weakest received power from node T could be just above SINR threshold, node I may cause transmissions from node T to R to fail if I is transmitting at the same time. Since at node R, the strongest interference power from node I can be just below SINR threshold (else node I will be a neighbor to node R), transmissions from node T can be received by node R if margin is set to SINR threshold. Algorithm 2 will only make the link {T,R} available for the routing layer if the received power from node T is above 2 SINR threshold. As a result, weak links that can be interfered by neighbors two or more hops away (such as I) will be automatically eliminated. V. SIMULATION AND DISCUSSION To evaluate the performance improvement gained from our algorithm, we set up a simulation using QualNet [13]. We implemented new modules in the simulator to simulate the effect of sea movement on antenna gains. Fig. 6. Orientation in QualNet is defined by the azimuth angle (A) and the elevation angle(e). A. Modeling the Ocean and Ship Movement In the QualNet simulator, the orientation of an antenna is specified by two angles, the azimuth and the elevation (see Fig. 6). The azimuth, which corresponds to the antenna degree in the horizontal direction, can be easily specified based on the direction of the wave simulated. The elevation, which indicates the degree of the tilt, has more impact on the change of the antenna gains. We developed a model to calculate the degree of tilt, i.e. the elevation angle, based on the wave characteristics (such as maximum wave height, wave period and wave length), as well as the size of the ship. cean waves are generally described as trochoid waves (see Fig 7), which are formulated as: 196

6 Fig. 7. Shape of a trochoid is formed by line traced by a smaller radius b inside a larger circle with radius a rolling on a fixed line. x = aθ b sin θ, y = a b cos θ (6) where a > b. Based on the definition of a trochoid, the constants a and b can be defined as: a = λ/2π, b = H/2 (7) where H is the wave height (magnitude) and λ is the wave length. We also simplified the model by defining θ with the standard traveling wave formula θ =2π( x λ t T ). T is the period of the wave, x is the relative position of the ship from a fixed origin, and t is the time instant of the calculation. Fig. 8. Ships relatively smaller compared to the wave have larger degree of rocking in our model Given the wave formula, we can calculate the tangent at any point on the wave to find the corresponding the degree of tilt caused by the wave. This is a simplistic view without taking the ship s size into consideration. Since a larger ship will experience less rocking than a smaller ship under the same sea conditions, the relative size of ship and wave determines the degree of rocking. We approximate this relationship in the following way: we identify the relative location of the ship s front and end, labeled as x head and x tail, respectively, and calculate the corresponding y values on the wave form. The tangent line formed by connecting the points of head and tail of the ship on the wave gives the degree of the tilt of the ship. The relative effect of this calculation is shown in Fig 8. The size of the ship is defined as the input parameter in our maritime simulations. B. Simulation Setup The sea surface movement, node mobility and two-ray pathloss models were coded into the QualNet simulator. The mobility model is based on actual shipping lane monitoring data provided by the Maritime Port Authority of Singapore. The lane is divided into two parts. The ships in upper part which is closer to land (and the basestation) are moving in the eastern direction. The ships in the lower part of the region are moving in the western direction. Node 4, a node in our study, is in the lower part of the region. The starting node placement is shown in Figure 9. A line is drawn between two nodes if they are within 1-hop distance to each other (in other words they can receive each other s control messages, e.g MSH NCFG and MSH DSCH). Node 1 is the base station and is not subjected to sea wave movements. The details of the simulation settings are shown in Table II. TABLE II SIMULATION SETTINGS Parameters Settings Application CBR, 4 bytes packet,.2 sec interarrival time Routing AODV Transport/Network UDP/IP MAC IEEE mesh (coordinated distributed scheduling) PHY OFDM, QPSK 1/2 rate, FFT size = 512, channel bandwidth = 16 MHz, frequency = 2.4 GHz pathloss model Two-ray pathloss model Transmission power 19 dbm at both base station and ships Antennas Omnidirectional; 45 o beam width; db at ship, db at the base station (node 1) Ship 64m(L) x 8m(W) Sea States level 3. and 6. Number of nodes 52; node 4 transmitting to node 1 Simulation time CBR traffic from 1 to 1 seconds latitude.3.25 Transmission graph at simulation time =. seconds longitude Fig. 9. Starting node placement. The x-coordinates are in longitude degrees, and the y-coordinates are in latitude degrees. C. Results and Discussions We ran simulations for three cases: no filtering; filtering by SINR; filtering by both SINR and avoidance of null regions. Since the sets of links presented to the routing layer are different in all cases, the routes are different in these simulations. The routes chosen by various cases are tabulated in Table III. The corresponding performances for different cases are shown in Figures 1, 11, 12 and 13. Note that the 1 197

7 x-axes labels methods in the graphs correspond to the cases listed in Table III. The simulations presented here were done with only one flow, i.e. flow from node 4 to node 1. This allows us to carefully trace and analyze the flow of events. If we have multiple flows, then it is more difficult to trace and analyze because there are many factors that can affect performance. These factors include: interference between flows (which is a function of link scheduling), queue management policy, slot allocation policy, node placement etc. Of course we can conduct experiments with multiple flows and do averaging but the results only reflect the percentage improvement with respect to specific routing protocol (AODV), MAC (82.16-based mesh MAC), node placement, mobility, and other settings. But what we want to show in this paper is how the sea wave movement affects various performance metrics, using AODV and based mesh MAC as an example platform. The first observation made is that performance figures with sea-state 3. are generally better than that with sea-state 6.. This is consistent with the analysis in Section III where smaller magnitude of sea motion results in smaller variations in received signal strength, which then translates to stable routes. number of hops in route from node 4 to node 1 (BS) Fig Methods Number of hops in the route from node 4 to node 1 (BS). seen in Figure 11. It is also observed that the inclusion of null-region avoidance does not affect much in the number of hops in the route. This is because AODV tends to choose long hops and the links that were chosen are mostly outside of the null region. With margin =1.5dB, the links in the route are above the peak-to-peak signal strength fluctuations caused by sea motion. As a result, there was no link breakage in the simulation. ss3 ss6 1 9 ss3 ss6.18 ss3 ss packet delivery ratio (%) minimum packet delay (seconds) Methods Fig. 1. CBR packet delivery ratio (%). Without filtering (Algorithm 1), there were link breakages e.g. link (4, 43). This link is at the edge-of-coverage from node 4. As a result, variations in received power produced by sea waves easily cause the received power to fall below reception threshold, resulting in link breakage that calls for route rediscovery. Since AODV tends to choose long hops, the rediscovered route can break again. Because packets cannot be transmitted during the route discovery, they are dropped from the buffer. This explains the low packet delivery ratio when no filtering is done (Figure 1). With margin =3.dB, fewer links are available for AODV to choose from. In fact, there was no link available that satisfies the SINR requirement at the initial 5 to 25 seconds of the simulation with or without null region avoidance. This results in lower packet delivery as compared to the cases with margin =1.5dB. As a result of using a higher SINR margin, only shorter links are chosen. This in turn results in more hops in the route, as Methods Fig. 12. Minimum CBR packet delay. A tradeoff for having stable route with more hops is that the packet delay is longer since a packet will have to traverse more nodes to reach the destination. In every hop/node, each packet will be queued and have to wait for its next transmission opportunity subject to scheduling. This results in longer minimum packet delay as shown in Figure 12. Without filtering, we observe that average packet delays are much higher than that with filtering, as shown in Figure 13. It is also interesting to observe that although the cases with margin =1.5dB and margin =3.dB result in the same number of hops in the routes, the case with margin =3.dB has higher average delays. This is caused by the fact that it takes longer time for route discovery to complete in the latter case. It actually took about 5 to 25 seconds before a route that met the requirements can be found; that is, after some movement of the ships. As a result some packets that were not dropped experienced longer delays. 198

8 TABLE III ROUTES CHOSEN UNDER DIFFERENT CASES (SEA STATE 6.) Cases Routes 1) No filtering (link 4 43 broke) 2) Filtering with margin = 1.5dB ) Filtering with margin = 3.dB (after seconds) 4) With only null region avoidance (1km) (link 4 43 broke) 5) Null region avoidance and margin = 1.5dB ) Null region avoidance and margin = 3.dB (after seconds) average packet delay (seconds) Fig. 13. Methods Average CBR packet delay. To see the effect of null region avoidance, we focus on the scenario with sea-state 6. and threshold margin = 1.5 db. Threshold margin of 1.5dB is used to ensure that there is no link breakage within simulation time (because a link breakage will trigger route re-discovery which might cause a different route to be selected). We then ran two cases: with and without null-region avoidance. We set the null region radius to 1km. The routes are shown in Table III. The number of packets delivered without null region avoidance is 2285 (out of 2 packets) while that with null-region avoidance is The difference of 79 packets are traced to be caused by packet losses due to intermittent reception of packets at node 28 from node 16. Node 28 is within the null region of node 16, and is observed to experience periodic fluctuations in received signal strength, as depicted in Equation 5. VI. CONCLUSIONS Our analysis shows that as a result of sea movements, longer links are subjected to antenna gain fluctuations while shorter links are subjected to high sensitivity of pathloss caused by effective height variations. We use AODV routing protocol as an example to show how these factors affect performance. AODV tends to choose links that are long as a result of using hop-counts as its path selection metric. The chosen links are thus prone to instability due to sea surface movements. We proposed and simulated a scheme that the MAC layer was made to present only stable links to the routing layer by filtering out links that were too long or too short. As a result, ss3 ss6 routes chosen by the routing layer only comprised of stable links which were not subjected to big fluctuations in received SINR caused by sea wave movements. Our simulation results showed that the choice of links by routing layer has an effect on performance. REFERENCES [1] First in the world: Wireless mobile wimax access in singapore s seaport now a reality, Mar. 28, aspx?getPagetype=2. [2] Su Wen, Peng-Yong Kong, Jaya Shankar, Haiguang Wang, Yu Ge, and Chee-Wei Ang, A novel framework to simulate wireless maritime communication networks, in Proceedings of MTS/IEEE Oceans Conference, Vancouver, Canada, Oct. 27. [3] Won Mooncheol and Kim Sung-Soo, Design and control of a marine satellite antenna, in Journal of mechanical science and technology, 25, vol. 19 of 1, pp [4] 82.16, IEEE Standard for Local and metropolitan area networks Part 16: Air Interface for Fixed Broadband Wireless Access Systems, 24, IEEE Computer Society and the IEEE Microwave Theory and Techniques Society. [5] J. An, Empirical analyses on maritime radio propagation, in Proceedings of the 59th IEEE Vehicular Technology Conference (VTC 24-Spring, May 24, vol. 1, pp [6] Pierson-moskowitz sea spectrum, Sept. 27, Products/seastate.htm. [7] J. Joe, S.K. Hazra, S.H. Toh, M.W. Tan, J. Shankar, V.D. Hoang, and M. Fujise, Path loss measurements in sea port for WiMAX, in Proceedings of IEEE Wireless Communications and Networking Conference (WCNC), Mar. 27, pp [8] Robert A. Nelson, Antennas, antennas tutorial.htm. [9] C. Perkins, E. Belding-Royer, and S. Das, RFC3561: Ad hoc On-Demand Distance Vector (AODV) Routing, 23, [1] International Maritime Organization, Recommendation on performance standards for an universal shipborne automatic identification systems (AIS), in IMO Resolution MSC.74(69), Annex 3, [11] C.F. Ball, E. Humburg, and F. Treml, Different OFDM Link Level Performance under the Presence of Co-channel Interference and Noise, in Proceedings of the 17th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC 6), Helsinki, Finland, Sept. 26. [12] David Tse and Pramod Viswanath,, in Fundamentals of Wireless Communication. Cambridge University Press, 25. [13] Qualnet network simulator 3.9.5, 199