2.1

Self-organizing network

A self-organizing network is a multi-hop AdHoc autonomous system composed of a group of (mobile) terminal nodes with wireless communication transceivers; each (mobile) terminal has both router and host functions: as a host, the terminal needs to run user-oriented applications; as a router, the terminal needs to run the corresponding routing protocol; inter-node routing usually consists of multi-hops (Hop); no network infrastructure and can be built quickly anywhere, anywhere. Because of this, AdHoc networks have the following characteristics compared to other networks [3,4].

2.2

Wireless channel fading and multipath

In the cellular mobile communication system, the base station is generally set very high (about 30 meters above the ground), while the mobile station is buried deep in the building. Take the downlink as an example, the radio wave is launched from the base station and then experiences distance loss (pathloss) and blocking loss (shadowingloss) to reach the near side of the user node. Near the user node due to the scattering and refraction of trees, buildings, vehicles, etc., the waves will become multiplexed spurious waves (we call these spurious waves as plain waves). These prime waves are superimposed on the receiving antenna of the user node, forming a fast fading phenomenon, and the signals passing through different paths arrive at the receiving end at different times, which will form a time delay extension. Suppose the power of the transmit signal is P(t)t, let the power of the receive signal be P_r (t), the transmit power is P_t(t), the distance fading is Ω_d(t), the blocking fading is Ω_s(t), and the multipath fading is g ²(t), then the power is calculated as shown below.

where the shading fading Ω_s(t) is due to the shadowing effect and fits the slow fading of the log-normal distribution. When discussing multipath fading, what must be considered is the fading rate, which is relative to the signal code duration (T). In narrowband transmission, the signal code duration (T) is much larger than the multipath delay spread [5] (DMS), and the multipath signal arriving at the antenna can be approximated by a one-way signal. The signal arriving from the base station is refracted, reflected and scattered by the objects around the mobile body, and the multiple prime waves, which reach the mobile station with a very small time difference (DLS), cause mutual interference. The fading of the received signal caused by the composite wave depends on the signal code duration (T), the number of prime waves (N), the moving speed of the mobile station (v), and the digital rate (R) of the transmitted signal. The amplitude of this fading obeys the Rayleigh distribution, and the phase obeys a uniform distribution between (0, 2π).

The radio signal in the tunnel is mainly transmitted by direct, transmitting and scattering to reach the receiver. The walls of the tunnel have shielding, absorption and scattering effects on radio waves, and theoretically, the straight tunnel is a non-ideal waveguide of oversized. The waveguide has a cutoff frequency, and only waves operating at frequencies higher than its cutoff frequency can propagate through it, otherwise they will rapidly decay. Strictly speaking, only the ideal conductive wall waveguide exists cutoff frequency f_C, frequency above the cutoff frequency electromagnetic wave without attenuation propagation, so the attenuation constant α=0; below the cutoff frequency electromagnetic wave attenuation infinity, so there is α =∞, at this time in the full frequency domain is clearly divided into cutoff and transmission areas [6-8].

2.3

Ant colony algorithm and AODV routing protocol

AODV routing protocol belongs to on-demand driven routing protocol, which is the most researched and widely used routing protocol for AdHoc network routing protocols, AODV routing protocol uses the form of broadcast for route finding, based on distance vector, and the minimum number of hops as the routing criteria, AODV routing protocol is an RFC standard [9-11].

The AODV protocol is divided into two main phases: route discovery and route maintenance, when data transmission is required, AODV initiates route discovery and propagates datagrams in the network to establish communication routes. Route discovery is when two user nodes in the network need to perform data transmission, the source node will first check whether there is an available route to the destination node from the routing table, if it exists, the source node will communicate directly through the route in the routing table, otherwise, the source node will initiate the route discovery process.

The ant colony algorithm uses ants to release pheromones to find the optimal path, which is very suitable for multi-user nodes to find the optimal routing problem, and the mathematical model is as follows.

Let there be n network nodes in the AdHoc network, C = {c₁, c₂,…, c_n} denotes the set of n nodes, the total number of ants in the colony is m, the set of neighboring nodes of node i is N₁(i), the set of pheromone values between two network nodes is denoted as Г = {τ_ij (t) | c_i ∈ C, j ∈ N₁(i)}, and at the initial moment, the pheromones on each path are the same, let τ_ij (0) = const. In the process of pathfinding, ant k decides the next direction to go based on the pheromone values on each path and the distance of the path for each network node. denotes the state transfer probability of ant k from node i to node j at moment t in the ant colony algorithm, and its formula is shown below.

where τ_ij (t) is the pheromone value on the path from node i to node j at moment t, is the heuristic function, where d_ij is the distance from node i to node j. The heuristic function represents the degree of expectation of ant k from node i to node j. denotes the pheromone heuristic factor, the larger it is, the greater the chance of selecting a route with a large pheromone value, and is the expectation heuristic factor, the larger it is, the greater the chance of selecting a route with a short path distance. Each ant needs to update the pheromone on the path after a section of the path, and the pheromone value on the path from node i to node j at moment t+1 is adjusted as:

where τ_ij(t + 1) is the pheromone value on the path from node i to node j at moment t+1, is the pheromone increment of ants passing through the path from node i to node j, and τ_ij (0) = 0, are the pheromone values of ant k remaining on the path from node i to node j at moment t. ρ is the pheromone volatility factor.

3.1

Layered structure

The network construction of tunnel self-organizing network is only borrowed from the idea of AdHoc, which is not the same as the traditional AdHoc network. The main differences are two.

One, the network levels are different. The mobile nodes of traditional AdHoc networks function as both hosts and routers, and can be used as sources or hosts, while the research model of tunnel self-organizing network proposed in this paper is a hierarchical network, i.e., each node only serves as an access point and a transmission point, as shown in Figure 1.

Figure 1.

Network hierarchy of tunneled self-organizing network

The network hierarchy of the tunneled self-organizing network can be divided into three layers: L1: the bottom layer, which refers to the terminals, either cell phones available in the existing network or other dedicated mobile terminals. The WPs are the core devices of the tunnel network, providing both wireless access to terminals and wireless transmission between nodes. Depending on the number of WPs in the L2 layer and the actual terrain, it can be decided whether L3 is needed or not. L3: Cluster management layer. When the number of WPs in L2 layer is too many, or the actual fading model of different terrains is very different, a cluster management layer can be considered, i.e., the WPs in L2 are divided into clusters by range or traffic volume, and each cluster has a cluster head. The cluster management layer manages the mobile nodes through the cluster head, and multi-hop control management, etc. It can also act as a gateway for communication with the external network.

The network structure level of the tunnel self-organizing network is affected by the actual environment of the tunnel, such as the size of the tunnel, the number of bifurcated branches, the distribution of equipment, the size of the space and many other complex factors, which determine the number of Wps that need to be cast in the tunnel and whether cluster management is required or not. For the convenience of consideration, this paper assumes that the tunnel model referred to is a flat and more uniform tunnel, and then its simple network structure is shown in Figure 2.

Figure 2.

Network structure of tunnel self-organizing network

As shown in the figure above, according to the tunnel environment can be randomly arranged some wireless access points (such as WP_i, WP_m, WP_n, etc.), the distribution of these access points can be based on the actual environment using different mathematical models, the distribution model discussed in this paper is uniform distribution. These wireless access points randomly arranged according to uniform distribution WPs (Wireless Points) have two main functions: one is to communicate with the user nodes, that is, to complete the user access and paging functions; the other is to complete the relay transmission of information from the source mobile station to the destination mobile station together with other WP_s through store-and-forward, these WP_s is equivalent to the traditional cellular These JTTs5 are equivalent to virtual extensions of traditional cellular base stations, i.e., they are used to complete comprehensive coverage of the area through multiple WP_s.

The process is as follows.

Since the multi-hop approach is used for data transmission, the distance between adjacent nodes is short, so the transmitting power can be greatly saved. At the same time, the interruption of individual nodes will only cause the reselection of routes, but not the network interruption, so it has a high degree of self-healing and anti-destructive properties. In order to enhance the self-organization of the network, MTs can be allowed to conditionally join the process of relay transmission under the condition that the WP assumes a fixed position, i.e., MTs can be certified by application, so that they have dual functions of MT and WP, which can increase the flexibility of the system and further save system power.

3.2

Ant colony algorithm based network optimization

Ant colony algorithm can prevent falling into local optimal solution, but the ant colony algorithm selects the next node according to the transfer probability, which has certain randomness and slow convergence speed, and the pheromone gap on each path in the initial stage is not large, the core nodes in multi-user node system often move, the system is unstable, and the slow convergence speed will increase the routing overhead, so the ant colony algorithm needs to be improved and then applied to multi-user node and multi-core node system. Traditional AdHoc network routing protocols focus more on how to establish the communication route from the source node to the destination node, and most of them maintain only one route, and AODV, as the most widely used routing protocol, also establishes only one route from the source node to the destination node, which may lead to network overload and increase the end-to-end delay, and if the route to the destination node fails, all links through that node will be dropped. Therefore, this paper focuses on the following three design points.

In equation 2, and referring to the literature [12] modifies dij as the user node i to core node j. The delay is used as a heuristic function factor, aiming to select the path with lower transmission delay, denoting the pheromone heuristic factor, the larger, the greater the chance of selecting a route with a large pheromone value, and the expectation heuristic factor, the larger, the greater the chance of selecting a route with a small path delay. When the core node in the system receives a route request, route response or transmits data, the pheromone on the path needs to be updated, in order to avoid the excessive concentration of data due to too much pheromone on a path and fall into the local optimal solution, the concept of maximum and minimum ant is introduced to set the maximum and minimum pheromone values on the path, define τ_min as the minimum value of pheromone, τ_max as the maximum value of pheromone, the pheromone update of the path through which the message passes, and the pheromone update on the path from user node i to core node j is shown in the following equation.

Where, ρ ∈[0,1) is the pheromone volatility factor, τ_ij(t +1) is the pheromone value on the path from user node i to core node j after the update, □τ_ij(t) is the pheromone increment on the path from user node i to core node j for this update, and the Ant-Quantity model is modified so that if the message passes from robot node i to node j, the increment is defined as ,otherwise it is 0.

where Q denotes the strength of the pheromone carried by this update and is a constant, d_ij is the delay from user node i to core node j, and τ_ij (0) = 0. Next, consider the core node load balancing problem, load balancing is based on the distribution of data transmission traffic in a multi-node system network, so that all nodes are enough to share the network transmission tasks evenly, to avoid concentrating communication tasks in individual nodes making these core nodes consume energy too fast, effective congestion control can extend the survival time of core nodes and improve the overall performance of the system. For multi-node systems, the user node data cache queue is detected to determine whether congestion occurs, and each user node monitors its current MAC layer data queue in real time and calculates the normalized remaining capacity.

where denotes the remaining capacity of the normalized data cache of user node i, L_i denotes the capacity in the current data cache queue of user node i, and L_max denotes the maximum capacity of the data cache queue of the core node. For a core node, the smaller is, the more data is cached in the current data cache queue, the higher the possibility of congestion in that node, and the lower the possibility of being selected as an intermediate node.

3.3

Self-organizing network communication optimization based on signal attenuation

Through the analysis of the relationship between attenuation and frequency, we can see that the attenuation of VLF, UHF, VHF and UHF bands is small, and the relative attenuation of IF band is large, when the working band is very low, it is easy to be subject to industrial frequency electromagnetic interference, and the channel capacity is small, requiring large transmitter power and large antenna volume. Therefore, do not choose the lower frequency band. Table 1 shows the loss of some frequencies in the flat lane [13,14]. From Table 1, it can be seen that 1000 MHz loss is always the smallest when the transmission is greater than 150 m.

Table 1.

Losses of radio waves in a flat and straight alleyway

Table 2 shows the loss of some frequencies when there is an inflection. From the table, we can see that 1000MHz loss is the smallest when transmitting more than 300m. Considering the compatibility of mobile communication systems, it is more appropriate to choose 1000Hz as the operating frequency band for tunnel communication.

Table 2.

Loss of radio waves in the presence of a bend in the tunnel

After designing the routing algorithm, the algorithm also needs to be optimized accordingly. Optimization means the ability of the routing algorithm to select the best path, based on the value of the metric and the weight value. A good routing algorithm is the result of several optimizations.

The route establishment mechanism of the routing self-organization based tunneling wireless communication system designed in this paper is based on the principle of minimum power loss, which means that the metric here is the power loss. According to the special environment of the tunnel network, the routing is required to be highly self-healing and energy efficient, so the following factors are considered for the designed routing.

First, the principle of selecting the best route is based on the least power loss between two points, rather than the shortest propagation path. The tunnel environment is complex, the size of the tunnel, the number and size of the surrounding equipment, and the influence of the tunnel facilities will affect the propagation loss of the radio waves to varying degrees. The overall loss of the radio waves, although related to the propagation distance, is not a simple one-to-one correspondence.

Second, due to the harsh environment of the tunnel, the power of each wireless access point is limited, so its workload should be minimized, so consider that a centralized routing algorithm can be used, and then sent to each WP through the guide frequency information.

Finally, as to whether the network is a virtual circuit or datagram format, as well as traffic and latency, are related to the actual amount of data transmitted and the required quality of service. Combining the actual environment of the tunnel network and the frequency allocation scheme of the core nodes.

4.1

Simulation analysis

(1) Reachability of user node packets

Reachability is one of the main metrics studied in this simulation. The experimental simulation focuses on the average reachability under the whole system, and this chapter conducts experiments in terms of node network size on message exchange.

Figure 3 shows the average reachability of packets for different network sizes. The figure compares the reachability of packets at different network sizes. A reachability node of 0.9 means that 90% of the network nodes are reachable.

Figure 3.

Reachability of packets at different network sizes

(2) Accuracy of information

Since the experiment simulates a distributed system, the propagation of information causes inaccurate information to the user nodes due to delays. The core nodes selectively forward information because of limited energy, making the information forwarded to the user nodes incomplete.

Figure 4 shows the average accuracy and completeness of the network-wide information collected during a given simulation time for different network sizes. As the graphs show, the average accuracy and completeness of the data is above 97% for all network sizes. However, the integrity of the information deteriorates in small-scale networks, especially in high-density networks, where direct communication is possible due to being in small and dense networks.

Figure 4.

Accuracy of data and integrity of data

(3) Reliability of nodes

The experiment evaluates the reliability reached by the core nodes by observing the success rate of packet forwarding (PDR) with a packet forwarding rate of 1m/s.

The Figure 5 shows the success rate of packet forwarding under different algorithms. In all the simulated algorithms, the success rate of packet forwarding of the core nodes in this paper can be improved well for networks with low or medium node density. This is because there are more distances beyond 1 hop in low-density networks, which gives opportunities to the node ant colony algorithm mechanism. The improvement in packet forwarding success rate gradually increases in larger networks, showing the scalability of the node ant colony algorithm mechanism.

Figure 5.

Success rate of packet forwarding for 2,4,6 and 8core nodes

To fully verify the effectiveness of the method, this section compares it with the original AdHoc algorithm and the original ant colony algorithm, which are used to test the optimization of self-organized networks under the same parameter conditions.

(4) Core node selection efficiency

Figure 6 shows the three self-organizing network methods. The average detection probability of 60 runs increases with the detection time, and the proposed method in this section outperforms the other two methods as the detection time progresses. This advantage is mainly due to the introduction of the hierarchical structure and the improvement of the ant colony algorithm, while considering the signal attenuation least strategy. Therefore, the method enables the core nodes to search the DEM region with a high occupancy probability of user nodes in a short period of time, effectively improving the search capability.

Figure 6.

Average detection efficiency and regional coverage

The area coverage percentage is defined as the proportion of the search area of the current user node in the whole search area. As shown in the figure, the improved AdHoc algorithm can expand the probability of the search area. Therefore, it can better coordinate the search behavior of the core nodes, which can search user nodes in a larger range.

The proposed self-organizing network optimization algorithm based on hierarchical model and ant colony algorithm is shown in Figure 7.

Figure 7.

Algorithm flow of this paper