Sage Journals: Discover world-class research

Abstract

High-speed nodes in Aeronautical Ad Hoc Network lead to dynamic changes in the network topology, thus increasing the frequency of re-routing and decreasing the performance of network communication. Selecting a routing path with higher stability can effectively reduce the re-routing frequency and extend the survival time of routes. This article presents a new prediction model of link stability utilizing the receiving signal intensity of data packets. Furthermore, Link Power Aware–Based Routing protocol is designed with the model. The simulation results verify the effectiveness of the prediction mechanism which can significantly increase the delivery ratio and decrease the average packet end to avoid transmission delay and lower control overhead.

Keywords

Aeronautical Ad Hoc network stable path routing protocol mobile computing Mobile Ad Hoc Network

Introduction

With the development of global economy and technology, the global air transport industry has achieved rapid development. In fact, there are more and more types of aircrafts whose performance keeps improving and number has been surging, which brings great pressure to the current civil aviation communication system and makes the capacity of the system close to saturation. In order to better carry out aviation management and provide passengers with better service, there is an urgent need to develop a real-time, efficient, and reliable aviation communication system. The system needs to meet the future needs of aviation communication, thereby improving the efficiency of aviation management, flight safety, and passenger air experience.

In the exploration of the solutions of future civil aviation communication, there are problems of high cost, long delay, and limited bandwidth in the solution of space-based network according to the satellites. Meanwhile, there are problems of high construction cost, limited coverage, and no support for transoceanic flight in the solution of ground-based network based on the ground stations. Therefore, Aeronautical Ad Hoc Network (AANET) can be developed to serve as a significant complement to the future civil aviation network communication system. AANET and traditional Mobile Ad Hoc Network (MANET) share a common feature that they require self-organizing nodes. However, AANET has its own features. (1) Most of AANETs are heterogeneous networks, of which different types of links are disparate and the air spectrum resources are limited. (2) In AANET, nodes move regularly and predictably. (3) Cyberspace span and node communication radius are large. (4) Nodes move fast and the topology is highly dynamic. (5) Nodes have a strong processing ability with adequate energy. (6) Design of network routing protocol is effectively supported by a large amount of airborne equipment. In response to these features, a corresponding routing protocol needs to be designed to maximize the performance of AANET.

Related work

After AANET is applied in civilian areas, a large number of organizations and individuals have put efforts into the study. Advanced Technologies for Networking in Avionic Applications (ATENAA) project was jointly carried out in 2004 by Thales Communications (THC), Technological Education Institute of Piraeus (TEI), and other research institutes from Germany, Greece, Italy, France and other nations.¹ AANET is a project² conducted by the University of Sydney in 2006. The project proposes the concept of AANET that provides air voice communications and Internet access services to the passengers. Networking the Sky for Aeronautical Communication (NEWSKY)^3,4 is a large-scale research project of AANET, jointly conducted by the German Aerospace Center and six other institutions of European Union (EU) countries in 2007. The purpose of the project is to accommodate the needs of future aeronautical communication and to explore the possibilities and ways of integrating a variety of aeronautical communication systems into a global heterogeneous network.

High dynamic topology of AANET is an important reason why the protocols of ordinary MANET cannot be used in AANET. In addition, since the routing protocols are designed for meeting the needs of network in a particular scene, the network scenes and requirements are different. Thus, the design of AANET routing protocol is more challenging than the ordinary MANET. The current research of AANET which is being built among civil aircrafts is still in theory. The following part is a brief overview of the research results of AANET routing algorithm in recent years.

In 2006, to address the problem of poor routing stability caused by high dynamic topology in AANET, Sakhaee and Jamalipour⁵ raised a routing algorithm utilizing the Doppler frequency shift to achieve the stable routing, named MUDOR. MUDOR algorithm effectively improves AANET routing stability, reduces routing switching frequency to some extent, improves network communication quality, and meanwhile reduces the routing overhead of network operation. However, the algorithm has no quantitative calculation of the routing stability. In addition, it does not consider the load balancing and Quality of Service (QoS) problems of network, which is prone to cause local network congestion. On the basis of MUDOR, E Sakhaee et al.⁶ proposed the QoS-MUSDOR routing algorithm. In QoS-MUSDOR, aside from Doppler frequency shift, routing remaining bandwidth, routing delay, and other QoS factors are added into the conditions of routing determination. In addition, the algorithm adopts a policy that forwards the optimal packet and discards the rest when responding to the request, transmitting the packet, and thereby preventing route request flooding. This is the first time that the QoS is introduced in the AANET routing algorithm. However, the algorithm only gives a theoretical analysis without specific any QoS metric or analysis in specific application scenarios. Under the future aviation concept of free flight, the literature⁷ presents a routing protocol ARPAM (Ad Hoc Routing Protocol for Aeronautical Mobile Ad Hoc Networks) based on AODV (Ad Hoc On-Demand Distance Vector). Combined with the positioning technology, the protocol utilizes the geographic information of nodes to control directional antennas. Furthermore, the protocol adopts active routing mechanism and the route maintenance strategy is also improved. However, the routing protocol refers only to the distance or hops between nodes as parameters to find the shortest route, which is difficult to ensure the stability of routing and end-to-end delay. In addition, since GGR (Greedy Geometric Routing) and GPSR (Greedy Perimeter Stateless Routing) have been proposed, a variety of improved greedy forwarding algorithms based on location have been adopted in AANET, such as GLSR (Geographic Load Share Routing), AeroRP, and automatic dependent surveillance-broadcast (ADS-B) system aided geographic routing protocol (A-GR). GLSR is proposed for air Internet access scenarios in North Atlantic.⁸ GLSR routing algorithm takes the geographical factors and the load of the next hop node into account when choosing the next hop node forwarding packets. In addition, the algorithm can balance the load of each base station by controlling the base station selection. However, the algorithm ignores the dynamic feature of nodes. AeroRP⁹ is proposed for the applications of aircraft remote sensing data acquisition. With the node position and trajectory, AeroRP forwards packets to the neighbor node which is supposed to be the fastest forwarding packet to the destination node, in the way of storing–carrying–forwarding. AeroRP has a favorable effect on the applications of remote sensing data acquisition. However, the protocol cannot meet the requirements of massive data transmission and real-time data communication due to network delay, congestion, and other issues in AANET. A-GR¹⁰ is proposed for high dynamic topology in AANET under conventional flight scenario. A-GR utilizes ADS-B system to obtain real-time information of the location and movement of aircrafts for routing scheme. The main feature of A-GR is that it can reduce the routing overhead with the combination of ADS-B system. However, the improvement makes no difference and leads to poor independence and complicated implementation process when the transmission data are massive. Furthermore, based on greedy forwarding, these routing algorithms belong to a local optimum routing strategy. Due to the blindness of the forwarding strategy, there are problems of wormhole and routing without quality guarantee.

Link stability prediction model

This section presents a simple and effective link stability prediction model. The core idea of the model is to calculate the variance of signal strength of packets received. The new model can accurately predict the stability of the link with low network control overhead.

Precondition

Modeling and simulations are based on the free-space propagation model, and all the nodes correspond to the clock synchronization mechanisms (clock synchronization of aircraft nodes having been well solved in AANET). The received signal strength of packets depends on the distance between the sender and the receiver. It is assumed that the receiver can still obtain a sufficiently strong signal by interference. Node sending power is a constant P_T, and the distance between the radio nodes is d. Ideally, the receiving power P_R = λ²P_S/(4πd)².

The minimum packets received power is defined as P_min to ensure that the packets are successfully received. When the received power is lower than P_min, the higher bit error rate (BER) results in retransmission of packets. If the retransmission fails for several times, the node will discard the packets. At that time, the received power is set as the minimum packet received power. When a node moves out of the signal coverage area of one neighbor, the last successful received power can be considered as P_min. Under the circumstances, when the received power of packets is less than P_min, the BER is high, and the packets are likely to be received incorrectly. Each node in the protocol is supposed to store P_min in the current environment. Since the signal-to-noise ratio (SNR) is influenced by the sending signal of the surrounding nodes and ambient sound, P_min is not a fixed value. For example, when the surrounding nodes are sending a large amount of data, the corresponding value P_min will be higher. When the link is broken, the P′ (the last received power when successfully receiving the neighbor packets) is utilized to update the value of P_min that the node stores. v_min is the minimum flight speed of the aircraft, and v_r is the current flight speed of the aircraft

$P_{min} \leftarrow \frac{v_{min}}{v_{r}} P' + (1 - \frac{v_{min}}{v_{r}}) P_{min}$ (1)

The above formula can make the value of P_min update under the surrounding environment. If the received power is larger than P_min, the packets can be successfully received. Otherwise, the link is considered broken.

Definition 1

Link stability prediction value

Link stability prediction value (LSPV) is a prediction value that represents the degradation of wireless link stability. In this article, the value is determined by two decision factors, including the S_LSPV (Single Link Stability Prediction Value) and M_LSPV (Multilink Stability Prediction Value).

LSPV

Stability vector function

Nodes monitor the broadcast packets of neighbors through the link layer (such as Hello packets) to record the received power. As shown in Figure 1, the receiving (monitoring) node is R and the sending (monitored) node is S. The stability vector function can be defined as f_R(S)

$f_{R} (S) = 10 \log_{10} \frac{P_{R}^{Now} (S \to R)}{P_{R}^{Pro} (S \to R)}$ (2)

The logarithm used in the formula is aimed at calculating the variation of powers intuitively. If $P_{R}^{Now} (S \to R) < P_{R}^{Pro} (S \to R)$ , f_R(S) < 0, the two radio nodes are away from each other. If $P_{R}^{Now} (S \to R) > P_{R}^{Pro} (S \to R)$ , f_R(S) > 0, they are close to each other.

Figure 1.

Single link stability prediction value calculation.

S_LSPV

As shown in Figure 1, the receiver (monitoring) R needs to continuously sample the received power through the link layer to ensure the sample value of the received power changes linearly. The nth received power value of R from S (one neighbor of R) as $P_{R}^{n} (S \to R)$ is sampled, and then the (n − 1)th stability vector value can be calculated as follows (collecting N times from a neighbor)

$f_{R}^{n - 1} (S) = 10 \log_{10} \frac{P_{R}^{n} (S \to R)}{P_{R}^{n - 1} (S \to R)}$ (3)

The S_LSPV can be worked out by calculating the variance of the stability vector values $f_{R}^{1} (S), f_{R}^{2} (S), \dots, f_{R}^{n - 1} (S)$ , which are calculated by power values $P_{R}^{1}, P_{R}^{2}, \dots, P_{R}^{n}$ acquired at n times from the receiver R and R’s neighbor sender S (collecting n received power values of R sent by one sender). The S_LSPV is calculated as follows

$S_LSP V_{RS} = \frac{{[f_{R}^{1} (S)]}^{2} + {[f_{R}^{2} (S)]}^{2} + \dots + {[f_{R}^{n - 1} (S)]}^{2}}{n - 1}$ (4)

Smaller S_LSPV_RS indicates the lower relative movement of the two nodes that create a wireless link; on the contrary, larger S_LSPV_RS shows larger value and higher relative movement. Based on this approach, the direct stability prediction value of wireless link between a receiver and any of its neighbor senders can be calculated.

M_LSPV

As shown in Figure 2, it is assumed that a receiver R creates n wireless links with n neighbor senders S₁, S₂, …, S_n, at same time. The stability vector value f_R(S_k) of S_k (one of the neighbor senders) can be calculated by collecting received power of R twice at time t1 and t2 (t2 > t1) from each neighbor sender. The formula is as follows

$f_{R} (S_{k}) = 10 \log_{10} \frac{P_{Rk - t 2} (S_{k} \to R)}{P_{Rk - t 1} (S_{k} \to R)}$ (5)

Figure 2.

Multilink stability prediction value calculation.

The variance of the stability vector value between the receiver R and all its neighbor senders S₁, S₂, …, S_n is worked out in order to calculate the M_LSPV as follows

$M_LSP V_{RS} = \frac{{[f_{R} (S_{1})]}^{2} + {[f_{R} (S_{2})]}^{2} + \dots + {[f_{R} (S_{n})]}^{2}}{n}$ (6)

Smaller M_LSPV_RS indicates weaker movement of receiver in local area; on the contrary, larger M_LSPV_RS indicates larger value and stronger movement.

LSPV calculation

The LSPV can be calculated by the following formula

$\begin{array}{l} L S P V (S) = ω_{1} \times S_L S P V_{R S} + ω_{2} \\ \times M_L S P V_{R S}, (ω_{1} + ω_{2} = 1) \end{array}$ (7)

ω ₁ and ω₂ represent the weight of the two decision factors of the link stability value. Node S acts as the sender in the calculation of S_LSPV and serves as a receiver in the calculation of M_LSPV. Smaller value of the LSPV(S) indicates that the link stability is stronger.

Link lifetime

In order to predict the lifetime¹¹ of a link, the node needs to periodically sample the values of the received power. The sampling interval is set as Δt. Also, the change rate of sample value is calculated as follows

$Δ P_{R} = \frac{P_{R}^{Now} - P_{R}^{Pro}}{Δ t}$ (8)

ΔP_R_(ave) represents the average change rate. If ΔP_R_(ave) is a positive number, the link lifetime T_live becomes longer, and the nodes that create the link get close to each other. Otherwise, equation (9) is used to periodically calculate the link lifetime T_live

$T_{live} = \frac{P_{min} - P_{R}^{n}}{Δ P_{R (ave)}}$ (9)

In order to reduce packet loss due to link break, the critical time T_cri can be calculated by formula (10) to ensure the execution repairing link, and switching route operation before the critical time is 0 (i.e. before the link breaks)

${T_{c}}_{ri} = T_{live} - T_{s}$ (10)

T_s represents the time interval between the two continuous packets received successfully. T_s changes according to the change of the network load. T_cri used in the routing mechanism can adapt to the changes in the network environment and minimize the packet loss rate.

Model analysis

The functions of the two decision-making factors in the link stability prediction model are as follows: S_LSPV is an effective measure model of relative motion between one sender and one receiver. Nonetheless, M_LSPV is an effective measuring model of the motion state which changes one node itself. M_LSPV fully considers the impact on the link stability caused by the motion state change of one node. In the calculation of S_LSPV, the receiver needs to obtain a set of consecutive packets (periodically sending Hello packets by a sender). The number of packets is defined as sample space, denoted by n. Sending interval is defined as the sampling interval denoted by Δt. Only when the sample space is n, it is possible to calculate the S_LSPV, namely, the sender and the receiver in order to maintain the neighboring relationship for a period of time.

Supposing that the sample space is 6 and the sampling interval is 1 s, only the duration of the neighboring relationship between the receiver and the sender lasts over 5 s perceived by the receiver, the S_LSPV value can be calculated. In view of the characteristics that the network environment will not change too much in the short term, link stability prediction model proposed in this article not only can accurately assess the relative mobility of each node but also can consider the effects of the node motion state change on the link stability. In the route selection process, the possibility that links with stronger stability should increase (as written in “LSPV,” lower link stability value indicates that the link stability is stronger). The link stability prediction is simple and effective based on the above model, which can be implemented in existing hardware platforms in a distributed form with lower network overhead.

Link Power Aware–Based Routing protocol

This section improves the traditional multicast routing protocol MAODV (Multicast Operation of the Ad Hoc On-Demand Distance Vector) based on the characteristic of AANET and proposes the Link Power Aware–Based Routing (LPAR) protocol. The factor of path stability value is introduced in the protocol, which can effectively enhance the stability of the route.

Routing policy

Routing policy of the new protocol is that the most stable path in reverse paths set will finally be chosen as the route.

Definition 2

Path stability value

It indicates the Prediction Value of Path Stability (PVPS). When the source node needs to transmit information to the destination nodes, it needs to first assess the stability of all the reverse paths that have been found. Path stability value, as a parameter of the routing decision, is calculated with the stability prediction value of all the links in the path. Path stability value is calculated as follows

$PVPS = max {Link_S_{ij} \ i, j \in Path & i \to j}$ (11)

Nodes i and j are parts of a directed wireless path. i → j represents that the node j is the next hop node of node i. Link_S_ij indicates the stability value of the link from i to j. Formula (11) indicates that the stability of a path depends on the most unstable link of the path along the direction of data flow. If the value of PVPS is smaller, the path is more stable. If several paths have the same stability value, the path with the minimum hop will be chosen to be the route.

A simple example is given to illustrate this routing policy. As shown in Figure 3, the directed edge between nodes indicates the stability value of the wireless link as the stability value between node A and node C: Link_S_AC = 0.09. The stable value of path P (S-A-C-G-R) Path_S_p = 0.15 (i.e. PVPS = max {Link_S_SA, Link_S_AC, Link_S_CG, Link_S_GR} = max {0.12, 0.09, 0.16, 0.08} = 0.16). There are three paths between node S and node R in Figure 3. According to the calculation method of path stability value, paths P(S-A-D-G-R) (PVPS = 0.12) and P(S-B-E-F-H-R) (PVPS = 0.12) are more stable while the number of hops of the path P(S-A-D-G-R) is 4 and the number of hops of the path P(S-B-E-F-H-R) is 5. Thus, the path P(S-A-D-G-R) can be determined as the final route.

Figure 3.

Path stability value calculation.

The existing routing strategy of routing protocol based on the link stability prediction assesses the stability of several paths between the source node and the destination node by the single-hop unit. Also, the model we adopt is characterized by two-hop unit with the introduction of decision factor M_LSPV; hence, the calculation is more accurate.

Route selection and activation

LPAR is the extension of MAODV multicast routing protocol; therefore, the format of multicast routing table and control packets is basically the same with MAODV. A new field, path stability value that is the stability prediction value of the path, is added to the routing entries of multicast routing table in LPAR protocol. Meanwhile, a new field link stability value is added to control packets (routing request packet (RREQ) and routing reply packet (RREP)) as well, representing the stability prediction value of the links that these packets go through.

Routing request

When a node wants to join a multicast group or have packets to send, without routing path in its routing table, it will broadcast the RREQ packets. After the nodes that do not belong to the multicast tree receive the packets, they continue to broadcast until the nodes that belong to the multicast tree receive the packets. When a node receives a RREQ packet, the packet which is the same as the received RREQ packet will be discarded. If the RREQ packet is new, the node compares its link stability value and the stability prediction value of the previous hop node (obtained from the previous hop node) in the link. Whichever is bigger will be updated to link stability value, and the number of hop value hop_cnt increases by 1. If the node is in an active state with multicast route entry and its multicast serial number is no less than the Dest_Seq value of the latest received RREQ packet, the node belongs to the multicast tree and the multicast route is the latest 1. Subsequently, the node waits for a fixed period of time. During this time, if other RREQ packet with more stable path passes the node, the path with the highest stability will be packaged in the RREP reply packet.

Routing reply

If a node belongs to the multicast tree, or can connect to the multicast tree with a route, after receiving the RREQ message, the node will unicast the RREP to the source node in accordance with the well-established reverse path. The information encapsulated in RREP stems from RREQ. When the intermediate node receives a RREP, an entry can be created in the routing table based on the node routing information in RREP packets. When a node receives several RREPs responding to the same RREQ, RREP which satisfies the following conditions is forwarded, and the rest are discarded: (1) the newly received RREP contains a larger sequence number than the RREPs received before and (2) when the sequence number is the same, the newly received RREP contains a smaller link stability value. When the source node receives the RREP, it will cancel the sending timing. A while later, the source node chooses the most stable route from several received RREPs as the candidate route sends the MACT (multicast activation) packet to activate this route.

Routing activation

After receiving the RREP, the source node sends an MACT packet. Then, other nodes receive MACT packet and search for the routing table, inserting the previous hop address in the next hop’s domain of routing table entries. If the routing table entry has been activated, MACT should not be sent anymore. If not, MACT packet should be sent according to the address in Pre_Hop domain of routing table entry. After the above steps, the nodes establish routes to the multicast tree.

Multicast tree maintenance

The root node of multicast tree sends Group Hello packets in every regular time interval to maintain or rebuild the multicast tree.

Link repair

When the link of multicast tree breaks, the downstream nodes are responsible for route reconstruction and the upstream nodes delete invalid nodes from Next_Hops. If the upstream node has no other descendant node, it will wait for a while. If the upstream node still has no descendant node after a designated period, it will execute the prune operation. When the multicast tree nodes receive RREQ from the downstream nodes, only the nodes that are closer to the root node than the initiating node can reply RREP, avoiding the repaired route connecting with a descendant node and forming into a routing loop.

Route switching

LPAR protocol maintains only those routes being used, while the movement of the nodes which are not on the active route do not affect the routing. The maintenance mechanism of predictive routes implements the soft handover which is usually adopted in the wireless cellular network. Also, it can calculate the link stability value and predict link lifetime at the same time. In order to accurately estimate the lifetime, the value should be periodically calculated to reduce the deviation caused by unforeseen circumstances. Based on the proposed link stability prediction model, each node in the broadcast tree which is constructed well will monitor the link states between itself and the surrounding neighbor nodes. When the received power is lower than the critical threshold, monitored by a node, the link is considered broken. Therefore, it needs to be ensured that the advance repair operation starts before the value of link available time is 0 and the old route is switched into the new one in advance before the route is interrupted. After the new route is established, the node will delete the relevant entries of the old route in the routing table and utilize the new route for packet transmission.

Simulation analysis

In this article, the simulation tool NS-2 is used to evaluate the performance of the protocol under different experimental environments. In the wireless simulation environment, the network nodes are randomly distributed in a 500 km × 500 km network, and each node uses the same radio equipment. The wireless communication radius is 75 km, and the wireless channel capacity is 1 Mbit/s. In the simulation, source and destination nodes utilize the constant bit rate (CBR) to simulate the flow of data traffic. The packet length is 512 KB, and the transmission rate is 5 packets/s. The aircraft mobile model is simulated by self-developed flight flow simulator AVMSim. Mobile nodes select the next destination coordinates according to the flight scheduling and then move to this destination at a fixed rate, ranging randomly between the minimum speed and the maximum speed. When nodes reach the destination, they continue to set off to the next destination by scheduling after some time, repeating the above steps. The simulation time is 24 h. In order to reduce the random error, the final experiment result is the average of experiment results for 20 times.

Experiment settings

To verify the efficiency of LPAR, its performance is compared with the following three protocols, including the traditional MAODV protocol¹² and another two improved MAODV protocols, which are route stability based QoS routing (RSQR)¹³ and distributed maximum lifetime multicast (DMLM)¹⁴ based on different link stability prediction mechanisms.Table 1 shows the simulation parameters. The weight of the link stability decision factor is assigned by the following two settings: (1) ω₁ = 0.8 and ω₂ = 0.2; (2) ω₁ = 0.5 and ω₂ = 0.5. Setting 1 indicates that the S_LSPV is more significant than the M_LSPV for link stability prediction. Setting 2 shows that two decision factors have the same importance. The two settings of LPAR in the simulation are abbreviated as LPAR-U and LPAR-B.

Table 1.

Experiment parameters.

Experiment number	Network size	Source node number	Destination node number	Node maximum speed (km h⁻¹)
1	60	9	15	300–1200
2	60–120	15%	25%	900

Experiment result analysis

In order to fully analyze the performance of LPAR protocol, the article analyzes and compares the performance of four multicast protocols through the below six performance metrics:^15,16 (1) delivery ratio, (2) average end-to-end delay, (3) network throughput, (4) sending amount of management packets, (5) total of transmission packets, and (6) multicast route lifetime.

Experiment 1: changing maximum movement speed of node

Through randomly deploying 60 nodes in simulation, the maximum movement speed of node increases from 300 to 1200 km/h. Figure 4(a) shows the relationship between the delivery ratio and the maximum movement speed of node. With the increase in the maximum movement speed of nodes, the delivery ratios of four protocols are declining. The MAODV protocol declines fastest while the LPAR declines slowest. In addition, the delivery ratio of MAODV protocol is the lowest while the delivery ratio of LPAR protocol is the highest. RSQR and DMLM are somewhere in between above two. This difference becomes more apparent with the increase in the maximum movement speed, as the MAODV protocol is vulnerable to nodes’ mobility and increases the probability of link break and packet loss. However, the three other protocols are added with link stability prediction module to predict the link status between nodes, thus avoiding frequent link break. Especially in the case of nodes moving at high speed with severe network topology changes, the three protocols can quickly adapt to topology changes and choose a route with higher stability. In addition, before the network topology changes, link is repaired and route is switched to increase the probability of successful packet transmission. When maximum speed reaches 1200 km/h, the delivery ratio of RSQR, DMLM, and LPAR protocols is still more than 70%. LPAR protocol adopts a mechanism through calculating the variance of reception signal strength, with the optimization of stability decision factors selection. Therefore, the obtained value of link stability prediction is more accurate. The LPAR protocol further reduces the negative impact of network topology changes on the multicast tree, in comparison with the DMLM and RSQR protocols.

Figure 4.

Comparison of protocol performance with the change of maximum movement speed: (a) packet delivery ratio, (b) average end-to-end delay (ms), (c) network throughput (packet number/s), (d) management packets sending volume, (e) total transmission packets, and (f) multicast route lifetime(s).

Figure 4(b) describes the relationship between the average end-to-end delay and the maximum movement speed. With the increase in the maximum speed of the node, the delay of four protocols is increasing. Hence, the delay of RSQR, DMLM, and LPAR increases slightly, while the delay of MAODV protocol increases obviously. The reason for that is the delay comes mainly from packets waiting and retransmission. With the increase in the maximum movement speed, network topology changes dramatically, which results in more broken links, route switching, and the increase in delay. As can be seen from the figure, the delay of RSQR and LPAR protocols is lower, while the MAODV protocol has the maximum delay. Especially, when maximum movement speed is larger and network topology change is more dramatic, the LPAR protocol has a lower delay. RSQR, DMLM, and LPAR are added to link stability prediction mechanism. With the increase in node speed, the delay caused by broken link and re-routing operations is decreased with the stability prediction mechanism. When the maximum movement speed is lower than 630 km/h, the delay of the LPAR protocol is higher than the RSQR protocol slightly, as the route established by LPAR is more stable and increases the delay in stable route establishment process. As the speed continues to increase, the LPAR protocol has the minimum delay. As can be seen from the results, LPAR is more adaptive and timely for high-speed mobile network.

Figure 4(c) shows the relationship between the network throughput and the maximum movement speed of node. Compared to the three other protocols, the network throughput of LPAR is the highest. When the maximum speed of node is 750 km/h, the network throughput of LPAR is to send 0.263 packets/s, and network throughput of MAODV protocol is to send 0.221 packets/s, indicating an improvement of 19%. A lower delivery ratio in Figure 4(a) means smaller network throughput.

Figure 4(d) shows the trends of management packet transmission with the maximum movement speed changing. When the maximum movement speed of node is less than 650 km/h, the management packet sending volume of MAODV is lower than the RSQR, DMLM, and LPAR protocols. However, as the speed continues to increase, the performance is the opposite. Reasons why management packet sending volume shows different results at different stages are as follows. (1) In the RSQR, DMLM, and LPAR protocols, in order to predict the link stability, more RREQ and RREP packets need to be sent. In addition, some extra route updates and maintenance control packets also increase the number of management packets; however, MAODV does not consider the link stability factors. (2) With the increase in the maximum movement speed of nodes, the opposite results indicate that MAODV can significantly reduce control overhead compared with the three other modified protocols. In terms of MAODV, as the maximum speed increases, the network topology changes more intensely and the network continuously sends route request and maintenance packets in order to repair broken links and execute re-routing operations. However, benefiting from the link stability prediction mechanism, RSQR, DMLM, and LPAR can send fewer control packets for multicast tree maintenance. The management packets sending volume of LPAR is the lowest since it can select the most stable route.

Figure 4(e) indicates the total of transmission packets with the movement speed changing. According to the figure, the total of transmission packets increases with the growth of the maximum movement speed. RSQR, DMLM, and LPAR reduce the probability of frequent link break by link stability prediction mechanism, avoiding forwarding a large number of packets caused by retransmission. In addition, as the node speed increases, more control packets need to be sent to adapt to topology changes and re-routing operations. LPAR shows the best performance.

Figure 4(f) presents the relationship between multicast route lifetime and maximum movement speed of nodes. RSQR, DMLM, and LPAR take consideration of link stability factors in the process of constructing multicast route. LPAR takes advantage of RSQR and DMLM for further optimization. LPAR not only can accurately predict the level of link stability but also can repair the link and switch the route ahead of time based on the periodic calculation of link critical time. As can be seen from the figure, with the increase in maximum movement speed of nodes, multicast route lifetime of every protocol is declining. The main reason is that the wireless link becomes fragile with the increase in node movement speed. The lifetime of DMLM is longer than RSQR protocol, and the LPAR protocol is the longest.

Experiment 2: changing the size of network topology

The size of network topology in the simulation increases from 60 to 120 nodes, and the maximum movement speed of nodes is 900 km/s. As presented in Figure 5, the delivery ratio of LPAR is the highest and that of MAODV is the lowest. The gap becomes larger with growing size of network. With the increasing size of the network topology, the delivery ratio of RSQR, DMLM, and LPAR declines slowly, yet decline rate of MAODV is obvious. With the growth of network size, it becomes more difficult to maintain the multicast tree, and the probability of link break increases. RSQR, DMLM, and LPAR adopt link stability factor, increasing the probability of successful packet transmission and reducing the negative impact of node random movement on the route discovery. In contrast, nodes in MAODV only perform traditional multicast routing protocol, which maintains only a short multicast route from the source node to the destination node. As a result, the delivery ratio cannot be improved when link breaks.

Figure 5.

Comparison of protocol performance with the change of network topology size: (a) packet delivery ratio, (b) average end-to-end delay (ms), (c) network throughput (packet number/s), (d) management packets sending volume, (e) total transmission packets, and (f) multicast route lifetime(s).

According to Figure 5(b), average end-to-end delay in multicast routing protocol increases with the growing size of the network topology. The delay of RSQR, DMLM, and LPAR increases slowly, while the delay of MAODV is obviously increasing. The main reason is that end-to-end delay comes from waiting and retransmission of packet queue. With growing size of network topology, RSQR, DMLM, and LPAR introduce link stability factor to reduce the probability of link break; therefore, the delay rises slowly. However, the MAODV protocol takes no consideration of this mechanism. As a result, frequent link break leads to greater packet queue waiting and retransmission delay. As seen from Figure 5(c), compared to the three other protocols, LPAR can achieve the highest network throughput, and this advantage becomes more apparent with the growing size of the topology. In addition, corresponding to Figure 5(a), lower delivery ratio reflects smaller network throughput.

Figure 5(d) shows that with the growing size of the network topology, the sending amount of management packets of four multicast routing protocols increase, which is mainly caused by the size of network topology. The sending amount of management packets of RSQR, DMLM, and LPAR is lower than MAODV, and RSQR has the minimum. There are two main reasons:

With the increasing size of network topology, multicast route construction needs to send more route requests and maintenance packets.

In the route discovery process, three protocols with link stability prediction function can select a multicast route with higher stability to reduce the control overhead of repairing link break.

As shown in Figure 5(e), the total of transmission packets increases with the growing size of network topology; LPAR has the minimum, while MAODV has the maximum. Link stability factor is introduced in RSQR, DMLM, and LPAR to choose a stability route, reduce the control and maintenance overhead caused by link breaks, and lower the probability of packet retransmission.

As shown in Figure 5(f), multicast route lifetime of LPAR is the longest and MAODV is the worst. The number of intermediate forwarding nodes of multicast tree is growing with the increase in the member nodes in the multicast group. The duration of the multicast route is proportional to the number of intermediate forwarding nodes. Therefore, as shown in Figure 5(f), as the size of network topology is becoming larger, multicast route lifetime increases accordingly. Experiment results show that the link stability prediction mechanism proposed in this article is simple and efficient. In addition, the performance of LPAR outperforms that of MAODV, RSQR, and DMLM. LPAR can enhance the stability of the route effectively, improve the delivery ratio significantly, lower average end-to-end delay, and prolong the lifetime of multicast route under the premise of lowering network control overhead. Furthermore, the performance of LPAR-U outperforms that of LPAR-B, indicating that the S_LSPV is more important than the M_LSPV during the process of weighted synthesis of link stability value.

Conclusion

On the basis of the analysis and summary of the existing link stability prediction mechanism, this article introduces two link stability decision factors (S_LSPV and M_LSPV) to assess the link stability and add the link stability prediction function to multicast the routing protocol MAODV. The new protocol LPAR considers route stability value as route selection parameter in order to find out a forwarding route with better stability. Based on the reaction mechanism, LPAR offers a flexible and viable way to construct a more efficient and stable multicast tree that can better adapt to network topology changes. Experiment results show that the link stability prediction model proposed in this article is simple and efficient with low cost so that it can combine with the existing routing mechanisms flexibly. In future study, the link stability prediction model will be further improved. The influence of power sampling results on this model in different time periods is considered. For instance, the sampling results within a certain period is divided into several time windows to deal with, and the time window with the latest date will be given a larger weight during the calculation. Meanwhile, the link stability model can be improved based on the scheduled flight trajectory of the aircraft. In addition, the performance of LPAR protocol will be compared with other multicast routing protocols.

Footnotes

Handling Editor: Sergio Toral

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research,authorship,and/or publication of this article.

Funding

The author(s) received no financial support for the research,authorship,and/or publication of this article.

References

Fuchs

Henniger

Giggenbach

et al . Broadband communications for aeronautical networks: the ATENAA outer optical link validation. In: Proceedings of the 1st CEAS European air and space conference, Berlin, 10–13 September 2007. Nein.

Sakhaee

Jamalipour

Kato

Aeronautical ad hoc networks. In: Proceedings of the wireless communications and networking conference, Las Vegas, NV, 3–6 April 2006. New York: IEEE.

Schreckenbach

Schnell

Scalise

et al . NEWSKY—networking the sky for aeronautical communications. In: Proceedings of the 1st CEAS European air and space conference, Berlin, 10–13 September 2007. Nein.

Schreckenbach

Schnell

Riera Diaz

et al . NEWSKY—a concept for networking the sky for civil aeronautical communications. Space Commun 2008; 21(3): 157–166.

Sakhaee

Jamalipour

The global in-flight Internet. IEEE J Sel Area Comm 2006; 24(9): 1748–1757.

Sakhaee

Jamalipour

Kato

Multipath Doppler routing with QoS support in pseudo-linear highly mobile ad hoc networks. In: Proceedings of the IEEE international conference on communications, Istanbul, 11–15 June 2006. New York: IEEE.

Iordanakis

Yannis

Karras

et al . Ad-hoc routing protocol for aeronautical mobile ad-hoc networks. In: Proceedings of the 5th international symposium on communication systems, Patras, 19–21 July 2006. New York: ACM.

Medina

Hoffmann

Rossetto

et al . A geographic routing strategy for North Atlantic in-flight Internet access via airborne mesh networking. IEEE ACM T Network 2012; 20(4): 1231–1244.

Jabbar

Sterbenz

. AeroRP: a geolocation assisted aeronautical routing protocol for highly dynamic telemetry environments. In: Proceedings of the international telemetering conference, Las Vegas, NV, 26–29 October 2009. New York: IEEE.

10.

Wang

Fan

Deng

et al . A-GR: a novel geographical routing protocol for AANETs. J Syst Architect 2013; 59(10): 931–937.

11.

Hong

Huang

T-P

Zou

W-X

et al . Research of AODV routing protocol based on link availability prediction. J Commun 2008; 29(7): 118–123.

12.

Royer

Perkins

. Multicast operation of the ad-hoc on-demand distance vector routing protocol. In: Proceedings of the 5th annual ACM/IEEE international conference on mobile computing and networking, Seattle, WA, 15–19 August 1999, pp.207–218. New York: ACM.

13.

Sarma

Nandi

Route stability based QoS routing in mobile ad hoc networks. Wireless Pers Commun 2009; 54(1): 203–224.

14.

Guo

Yang

Maximizing multicast communication lifetime in wireless mobile ad hoc networks. IEEE T Veh Technol 2008; 57(4): 2414–2425.

15.

Caleffi

Akyildiz

Paura

OPERA: optimal routing metric for cognitive radio ad hoc networks. IEEE T Wirel Commun 2012; 11(8): 2884–2894.

16.

Cacciapuoti

Caleffi

Paura

et al . Link quality estimators for multi-hop mesh network. In: Proceedings of the Euro Med Telco Conference, Naples, 12–15 November 2014. New York: IEEE.