Link availability Prediction based on capacity optimization in MANET

Mobile Ad hoc Network (MANET) is a wireless network composed of multiple wireless nodes without fixed infrastructure support, which is expected to play an important role in future commerce and military, especially in marine and aerospace communications systems. In this paper, for link failure caused by node mobility, the prediction of link availability is given according to the dynamic characteristics of the link, and transmission modes and relay nodes are selected to optimize the link capacity and reduce the interference. Simulation results show that the proposed routing metric method can select stable routing paths with less interference, and reduce routing overhead caused by node movement based on link availability analysis.


Introduction
The integrated satellite, sky, terrestrial and ocean networks has become the focus of state, military and aerospace scholars, due to its capability of time-space continuity, high reliability, large regional capacity and high mobility. However, the integrated network carries a variety of services, which are characterized by large time and space, long time delay, interruption, multi-track and multi-type. MANET can not only exist as an independent network, but also serve as an effective supplement to the integrated network structure based on its independence and flexibility.
MANET is a multi-hop temporary autonomous system which has no fixed base station and basic network, and all nodes in the network have the functions of host and router at the same time, and can be moved at will. Therefore, dynamic topology is one of the most remarkable characteristics of MANET. Due to the rapid topology changes, established routes can be broken frequently, which brings severe challenges to routing design. In order to reduce reroute operations, the most stable path must be chosen. In [1], the interference in large wireless networks influence on the performance in the system is analyzed. In [2], the influence of co-channel interference on the interruption rate of wireless Ad hoc network is defined. In [3], the link expiration time is predicted, assuming that the velocity and direction of the two nodes have not changed since the beginning. In [4], the current motion information of nodes (position, velocity and direction) is used to predict the respective positions of two nodes and whether they are still connected at time t. In [5], A. Bruce McDonald et al. proposed the concepts of link availability and path availability, and applied them to the design of routing protocol. In [6], proposed link models for predicting the availability, the probability that the link will last to a certain time is predicted based on the possible changes in motion. In [7], the transmission reliability is defined to measure the transmission performance of the MANET, the impact of interference is considered, and genetic algorithm is used for optimization. In [8], proposed a learning automata-based topology control method within a cognitive approach, which deals with adding cognition to entire network protocol stack to achieve stack-wide and network-wide performance goals. In [9], proposed a Bayesian pursuit algorithm, which predicts the node movement parameters by learning environment to predict the length of the link. In [10], artificial neural networks were applied to classify the set of cooperative relays under DF-relaying constraints, to maximize the achievable secrecy rates. However, above literature did not consider the dynamic change of network topology caused by mobile node, and the topology structure with more than two hops is not suitable for fast moving network. In addition, control of network topology requires acquisition of certain network state information, so it is not applicable to mobile AD hoc network without central control infrastructure. Therefore, a two-hop link model based on mobility is established, and a capacity optimization link predictive (COLP) topology control scheme with cooperative communication is proposed in this paper.
The paper is organized as follows. Section II gives a detailed description of the dynamic system model in MANET. In Section III, link availability is calculated according to the relative motion of nodes to predict the link life, and the optimal topology is selected based on network capacity and interference. In Section IV, numerical results from the model simulation are presented. Finally, conclusions are drawn in Section V.

System model
The prediction method in this paper is based on the modified Random Walk-based Mobility model. Due to the fast change of topology, the limited storage space and energy of nodes, it is difficult to predict the future motion state of nodes based on historical data. However, link prediction requires that the motion of nodes must conform to a certain habitant rather than be completely random. In MANET, nodes can obtain their position through some devices, such as GPS, the topology state of the network can be predicted effectively according to the motion habitant of nodes [11]. Assuming that the nodes obey the Poisson distribution and move independently, thus the links break independently, which is fits the actual scene of transportation and ship navigation.

/ 19
In this model, it is assumed that the maximum radius of a mobile's radio coverage is R, if a node is within the transmission range of another node, then the link composed of them is available, and one-way communication is not considered. In wireless communication, signal attenuation occurs in the transmission process, the received signal power is rt  It is assumed that nodes M and N move according to movement parameters ( Capacity optimization link predictive topology control is proposed to improve the network. The broken of any link will lead to the failure of the path, so the stability of the path depends on the stability of all the links in this path. During the non-breaking time, the most stable path refers to the path with the maximum capacity and minimum interference. Transmission mode is one of the main factors affecting outage capacity, interference is another major factor affecting network capacity. Therefore, link availability and interference needs to be analyzed to improve the network reliability, and objective function needs to be given to measure the topology performance.

Link availability analysis
In order to predict network topology changes, it is necessary to analyze link availability between nodes. Link availability refers to the probability that link exists at time 0 t and still exists at time 0 + tt [13]. In order to describe the link connectivity of two nodes in a time period, the motion distribution of individual nodes should be analyzed first, then according to the magnitude and the direction of one moving node relative to another, the connection probability of node is calculated, which is the influence of motion on link availability [14].
At time t, relates the distance between two nodes to the link availability. Raleigh vector and a constant, the complete derivation can be found in Reference [15]. Break down    Rt into the X and Y components of the normal distribution, add them to C x and C y components of the initial distance, and the probability distribution function of the link available time is given by the following expression: When node N moves to the limit of the communication range of node M, in the triangle, according to the law of cosines, we can figure out the angle between line M' and N' and the X-axis is Dt is greater than or equal to 0. When R , link availability reaches its maximum value.

Objective Function
The objective function is the capacity while network is connected. In the transmission, signal is attenuated from one terminal to another, therefore each terminal can only communicate directly with the others closed to it.
If the nodes are far apart, middle terminal is required as a router to forward the information, until the destination receives the message. Thus, direct transmission, two-hop transmission and cooperative transmission modes will be considered [16], the analysis of topological properties in this paper is divided into three aspects.

1) Direct Transmission
According to the initial position, when the distance between source S and destination D is less than effective communication radius R, the two nodes become adjacent nodes. The source S is located in the center of the circular, the destination D should be located in the effective communication range of source, and traditional pointto-point direct transmission can be used. Signal-to-Noise Ratio (SNR) of the received signal is Outage capacity is used to reflect the link rate, which is the link capacity obtained under a small interrupt link probability ɛ, the outage capacity of direct transmission is [17]: To ensure the quality of transmission, interference of the path should be the sum of all participating nodes covered in the transmission. Since only two nodes participate in direct transmission process, the interfered nodes include the nodes covered by source and destination: Due to the mobility of nodes, motion state of nodes will affect link connectivity and system stability, which

2) Two-hop Transmission
Two -hop transmission attempts to improve the link quality by replacing its long distance direct transmission with two-hop transmission, it takes two time slots. In the first slot, data is transmitted from the source to the relay, in the second slot, it is transmitted to the destination, as shown in Figure 3. The SNR of the signals received by relay from the source, and by the destination from the relay are respectively. The maximum instantaneous mutual information of the two-hop transmission link is:    (7) According to the interference model, all covered nodes must remain silent while the node is transmitting

3) Cooperative Transmission
The traditional direct transmission only uses signals from source to destination, while cooperative transmission is able to decode the combined signals from source and relay received at destination to obtain the better SNR. The DF scheme with only one relay is adopted for cooperative transmission, that is, the relay decodes the received source signals, then recodes and transmit them to the destination. Maximum ratio combining technique (MRC) [18] could be used to decode the two signals. Its maximum instantaneous mutual information is: The interference of cooperative transmission is more complex than the other two transmission modes, to ensure the success of transmission, the interference of path should be the sum of coverage of all participating nodes [19]. In the broadcast stage, not only the neighbors covered by source, but also the neighbors covered by relay and destination need to be silent. Thus the interference of cooperative transmission may be more serious, which will reduce the number of transmission nodes and capacity. The interfered nodes include the coverage of source, relay and destination, the interference set of cooperative transmission is: Since cooperative transmission requires constant communication, the predicted communication time is . Therefore, the objective function is: (12) This system choose the appropriate transmission mode and relay with minimal interference to optimize link capacity [20]. Cooperative transmission does not have a higher rate than direct and two-hop transmission, because the performance of transmission also depends on the relative position of relay. If there is no relay capable of achieving higher interrupt capacity, direct or two-hop transmission should be used. In MANET, the objective function of the path is: Therefore, when the relay set

Results and discussion
In this section, MATLAB is used to simulate the dynamic topology reconstruction performance of random topology model. Wireless channels follow the Rayleigh distribution of slow fading, and only the effect of network topology on capacity is concerned. In the simulation, 30 nodes randomly distributed over 1000x1000 square meter area. Nodes move according to the random-walk-based mobility model, and generates a velocity for its next movement randomly. Nodes periodically exchanges information with neighbors, the effective transmission area is circle, the transmission range of each mobile node is R. Figure 4(a) shows the initial position of each nodes, Figure   4(b) is the position of nodes when time is 100s. Figure 5 compares the topological performance of the COLP with reconstruction mechanism with that without reconstruction mechanism and direct transmission. The maximum speed of movement is 1m/s, and the path loss index is set as 3.5. Simulation results show that topology performance of COLP is higher than other systems, and performance optimization comes from the joint design of reconstruction mechanism and link prediction in topology algorithm. Simulation results show that when the coverage radius is smaller than 100m, the larger the coverage radius is, the more adjacent nodes are in the simulation area, and the larger the outage capacity is. When the coverage radius is greater than 100m, topological properties deteriorates instead with the increase of coverage radius, this is because the large coverage radius not only increase the capacity, but also increase the interference accordingly. In this case, direct and two-hop transmission perform better than cooperative schemes, and reduce the energy consumption. When the coverage is 100m, the topological performance of COLP system is better, and improved by 17% over that of direct transmission system.
Due to the mobility, the link availability is less than 1, thus there is a mismatch between the predicted curve and the practical curve, see Figure 6. Although prediction error is inevitable, they have the same change trend, which means that the prediction model can predict communication duration and capacity. Since node is not allowed to move out of the given space, in the simulation, the higher the speed, the easier the link is to break. However, low node activity can increase the accuracy of link prediction, and prediction error is smaller when angle of motion was π than when it was π/2, in the direction of motion, the larger the motion angle fluctuation is, the closer it is to the random walking movement model, and the relative motion speed of the two nodes is small, the link will connected for a long time.  Table 1. The prediction accuracy of the communication time can reach 99%, average error of less than one second, because when there are multiple relay available in the system, it tends to choose the relay with a low speed relative to source and the destination, 7 nodes with low relative motion are selected, while only one node with high relative motion is selected, which increases the stability of the link and the accuracy of the prediction.
When the interrupt probability increases, the topological performance of the system is improved, as shown in Figure 8. When path loss index is 3.5, the optimal transmission range increases from 79m to 113m, as shown in Table 2. Due to the interference, the greater path loss index is, the difference value of optimal transmission range is smaller when the interrupt probability are 10 -4 and 10 -3 respectively.

Conclusion
In MANET, random movement of nodes results in dynamic change of network topology, this paper propose a predictive topology control mechanism of capacity optimization for highly dynamic wireless self-organizing networks, and the connectivity probability of the whole route can be obtained by link availability. Theoretically, cooperative transmission can achieve higher network capacity than direct transmission, however, it does not necessarily have a higher rate than direct and two-hop transmission, because the performance of cooperative transmission also depends on the relative location of the relay. If there is no relay that can reach the higher interrupt capacity, direct transmission should be used. Interference of cooperative transmission is more serious, which will reduce the number of nodes transmitted simultaneously in the network. Therefore, the performance gain obtained in this algorithm is actually derived from the joint design of retaining optimal transmission mode, more reliable and stable relay node, and minimum interference path in topology control. This algorithm significantly reduces data storage capacity by obtaining the location information of nodes within a certain time interval, computational time and complexity are reduced to adapt to the rapidly changing network structure by abandoning the multi-hop mode and iterative algorithm, network capacity is improved by optimizing topology for mobile AD hoc networks without central fixed infrastructure, and unnecessary overhead of frequent information exchange is avoided by predicting communication time of link through the motion law and trajectory of nodes. The theory proposed in this paper not only provides a reference for MANET, but also can be used as an effective complement to the integrated satellite, sky, terrestrial and ocean network structure.
Error of this prediction model is affected by mobility and wireless communication conditions, so there is no perfect prediction technique with 100% accuracy. Although the prediction error is inevitable, this system can accurately predict the existence time and the capacity of link, generate a reliable and stable network topology, achieve higher network capacity compared with fixed topology configuration. However, due to lack of global information, this paper use interference of the initial path, learning algorithm can be used to predict the degree of interference in future research. Communication quality and network coverage will be affected by obstacles in wireless transmission path, so multi-hop topology can be extended to enhance network connectivity. Message exchange can run out of nodes' power, and a more energy-efficient topology model should be considered in the future.

List of Abbreviations
MANET: Mobile Ad hoc Network; COLP: Capacity optimization link predictive.
MRC: Maximum ratio combining technique.
SNR: Signal-to-Noise Ratio

Availability of data and materials
The datasets used and analysed during the current study are available from the corresponding author on reasonable request.   Two-hop transmission attempts to improve the link quality by replacing its long distance direct transmission with two hop transmission. Fix relay at the point R which is the center of circle, the source and destination should be located in this circular area. When the source moves ZS,R along from S to S ' or destination moves ZR,D along from D to D ' with respect to the relay, the path will be broken. Therefore, available communication time can be predicted with the relative moving distance obtained by the relative moving speed and angle between the two pairs of mobile nodes. (a) t=0s: The initial position of each nodes. In the simulation, 30 nodes randomly distributed over 1000×1000 square meter area.

Figure Legends
(b) t=100s: The position of nodes when time is 100s. Nodes move according to the random-walk-based mobility model, at each interval, they generate velocity and angle for their next movement randomly in the range of maximum speed and angle, respectively. Outage capacity topological performance of the reconstruction COLP mechanism, system without reconstruction, and direct transmission are all influenced by communication radius. 30 nodes distributed randomly, move with the maximum speed 1m/s, and the path loss index is set as 3.5. The objective function of COLP is higher than other mechanism, and performance of cooperative schemes are not always better than direct or two-hop transmission schemes, due to the affection of interference radius. Objective function of the COLP mechanism is effected by maximum movement speed and motion angle. With the same motion angle, the higher the speed, the easier the link is to break. In the direction of motion, the larger the motion angle fluctuation is, the small the relative motion speed of the two nodes is, and the smaller the prediction error is. Although there is a mismatch between the predicted curve and practical curve, they have the same change trend, and they are close.

Fig 7. Objective function versus path loss index.
The topology performance of the COLP mechanism is influenced by path loss index under different communication radius, and the loss factor also affects the optimal transmission range of nodes. There are multiple relay available in the system with a large communication radius, means the system has more options and tends to choose a relay with appropriate position and low speed relative to the source and the destination, which can increases the stability of the link and the accuracy of the prediction. The topology performance of the COLP mechanism is influenced by interrupt probability under different communication radius. The topological performance of the system is improved with the increase of interrupt probability. And predicted curve and practical curve are close to each other in both the interrupt probability are 10 -4 and 10 -3 .