Large-scale ad hoc network multipoint relay selection method based on ant colony optimization

Abstract

The invention discloses a large-scale ad hoc network multipoint relay selection method based on ant colony optimization, and the method can effectively reduce the size of a selected MPR set, thereby reducing the packet loss rate of a network, and improving the throughput of the network. Firstly, an ant colony algorithm is introduced into a selection process of an MPR set, secondly, a local three-hop neighbor database and speed measurement are added into a path selection decision of a node, and then for updating pheromones, a global pheromone updating rule is adopted to update the pheromones on a path so as to prevent the algorithm from being converged to a local optimal solution. According to the method provided by the invention, the problem that the traditional MPR selection algorithm is easy to generate redundancy and the difficulty that the traditional MPR selection algorithm cannot adapt to the existing large-scale ad hoc network are well overcome.

Description

Technical Field

[0001] The present invention relates to the fields of optimization technology in the computer field and self-organizing network technology in the mobile communication field, and in particular to a large-scale self-organizing network multi-point relay selection method based on ant colony optimization. Background Art

[0002] In recent years, the development of mobile ad hoc networks has become increasingly prominent in the civil and military fields. Unmanned Aerial Vehicles (UAVs) have been widely used in real-time monitoring, search and rescue, relay transmission, strategic strikes, etc., due to their small size, low cost, and easy deployment. However, UAV ad hoc networks have the characteristics of strong node mobility, fast network topology changes, frequent data interactions, and high energy consumption. Traditional routing algorithms can no longer meet their requirements in terms of network transmission delay, packet loss rate, and routing overhead.

[0003] The Optimized Link State Routing (OLSR) protocol is a classic link state protocol that broadcasts Hello packets between nodes to complete link detection, neighbor discovery, and selection of multi-point relay (MPR) nodes. After obtaining MPR node information, the topology control (TC) packet is used to establish and maintain the entire network topology, and finally the Dijkrastra algorithm is used to calculate the path and generate the route. The selection of MPR nodes is crucial. Each node will only send its TC packet to its corresponding MPR node to reduce the number of control packets between networks. Reference [8] mentioned that selecting the optimal MPR set is an NP-hard problem. Traditional methods often use greedy algorithms to select MPR sets, giving priority to selecting the one-hop node that covers the most two-hop neighbors of the source node. This will cause a lot of redundancy, which will increase the protocol overhead and increase the delay between networks. Many scholars have improved the MPR selection algorithm from different angles. Some scholars use the minimum-maximum algorithm to reduce the number of TC packets on each node. Some scholars have proposed a set-based MPR selection algorithm, which can effectively eliminate invalid redundant nodes by combining loops and set operations. Other scholars have proposed MPR selection based on a local database of neighboring nodes extended to three hops. The MPR selection uses the existing algorithm in OLSR, and the results show that it is superior to the standard OLSR in terms of the number of TC packets and energy efficiency.

[0004] The Ant Colony Optimization (ACO) is an abstraction and improvement of the real foraging behavior of ants in the real world. It is a heuristic algorithm for intelligent optimization. Ants transmit information to other ants by leaving pheromones when they move during the foraging process. Other ants decide their next path based on the concentration of pheromones on different paths. The advantage of the ant colony algorithm is that its search is global, while the traditional greedy algorithm cannot take the global situation into account and often falls into the local optimum. The ant colony algorithm uses random probability to select the next hop path. When pheromones accumulate and the probability increases to 1, the algorithm will degenerate into a greedy algorithm. Although the single ant colony algorithm has been improved on the basis of the greedy algorithm, it still has the problem of falling into the local optimum due to too long iteration time. Summary of the invention

[0005] The purpose of the present invention is to address the defects and shortcomings of the above-mentioned prior art and provide a large-scale self-organizing network multi-point relay selection method based on ant colony optimization. The method combines the global search capability of the ant colony algorithm, takes into account the characteristics of the unmanned aerial vehicle self-organizing network, introduces the local node three-hop neighbor database into the ant colony's path selection function, and considers the speed of the node at the same time. The original path selection and state update mechanism of the ACO algorithm are improved, and the ant colony optimization is applied to solving the MPR set problem, thereby achieving the purpose of optimizing the MPR set. Finally, it is integrated into the Qualnet network simulation software system, which well verifies the performance of the algorithm proposed by the invention.

[0006] The technical solution adopted by the present invention to solve the technical problem is: a large-scale self-organizing network multi-point relay selection method based on ant colony optimization, the method comprising the following steps:

[0007] Step 1: Initialize the source node's one-hop neighbor set S1, two-hop neighbor set S2, three-hop neighbor set S3, the number of ants num_ants, the array of nodes visited by each ant visit, the number of currently uncovered two-hop neighbor nodes Uncover_S2, and the total number of cycles of the ant colony total_iteration;

[0008] Step 2: When the current number of iterations is less than the total number of loops, do the following steps: initialize the starting node of each ant and record it in the visit array of each ant, initialize each ant's current one-hop neighbor set cur_S1, current two-hop set cur_S2, and current three-hop set cur_S3;

[0009] Step 3: For each ant, when the current Uncover_S2>0, do the following steps: Calculate the probability of the ant choosing other nodes according to the path probability selection formula proposed in the present invention, select the next hop of the current ant according to the roulette wheel selection method, update the visit array, cur_S1, cur_S2, cur_S3 array, cover the node corresponding to the selected node, and update cur_Solu and best_Solu for each ant according to the current visit array situation;

[0010] Step 4: Update the current path according to the pheromone updating method proposed in the present invention.

[0011] Furthermore, in the above step 3 of the present invention, the path probability selection formula is:

[0012]

[0013] Among them, α represents the heuristic factor of pheromone, β represents the heuristic factor of two-hop weight, γ represents the heuristic factor of three-hop weight, ε represents the heuristic factor of node relative speed, τ(i) represents the concentration of pheromone held by node i, μ(i) represents the weight of the two-hop neighbors of the source node covered by node i, η(i) represents the weight of the three-hop neighbors of the source node covered by node i, and ν(i) represents the formula for the influence of the moving speed of target node i on the probability selection.

[0014] Furthermore, in the above step 4 of the present invention, the pheromone update formula is:

[0015] τ i (t+1)=(1-ρ)·τ i (t)+Δτ i (t) (2)

[0016] Where ρ is the phased pheromone volatilization rate, τ i (t) is the pheromone increment on node i before updating, τ i (t+1) is the pheromone increment on node i after the update;

[0017]

[0018] Among them cur Represents the current iteration number, ite max Represents the maximum number of iterations set;

[0019]

[0020] in represents the pheromone released by ant k on the path to node i during the iteration process, curse kis the solution of the current iteration, and the Q value is the initial pheromone constant.

[0021] Beneficial effects:

[0022] 1. The present invention effectively reduces the size of the selected MPR set, thereby reducing the network packet loss rate and improving the network throughput. The ant colony algorithm is introduced into the selection process of the MPR set, and the local three-hop neighbor database and speed considerations are added to the node's path selection decision. For the update of pheromones, a global pheromone update rule is adopted to update the pheromones on the path to prevent the algorithm from converging to a local optimal solution.

[0023] 2. The present invention solves the problem that the traditional MPR selection algorithm is prone to redundancy, and achieves the problem that it is not suitable for current large-scale self-organizing networks.

[0024] 3. The present invention combines the global search capability of the ant colony algorithm, takes into account the characteristics of the drone self-organizing network, introduces the local node three-hop neighbor database into the ant colony's path selection function, and considers the node speed at the same time. The original path selection and state update mechanism of the ACO algorithm are improved, and the ant colony optimization is applied to solve the MPR set problem, achieving the purpose of optimizing the MPR set. Finally, it is integrated into the Qualnet network simulation software system, which well verifies the performance of the algorithm proposed by the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 The present invention is a flow chart of the method.

[0026] Figure 2 This is a schematic diagram of the OLSR protocol source code structure and functions.

[0027] Figure 3 This is a schematic diagram of the 120-node network simulation topology.

[0028] Figure 4 Schematic diagram for node mobility settings.

[0029] Figure 5 Schematic diagram of packet loss rate comparison between M-ACO and greedy algorithm under different numbers of nodes.

[0030] Figure 6 Schematic diagram of the end-to-end delay comparison between M-ACO and greedy algorithm under different numbers of nodes. DETAILED DESCRIPTION

[0031] The invention will be further described in detail below in conjunction with the accompanying drawings.

[0032] Embodiment 1

[0033] like Figure 1As shown, the present invention provides a large-scale self-organizing network multi-point relay selection method based on ant colony optimization. This method combines the global search capability of the ant colony algorithm, takes into account the characteristics of the drone self-organizing network, introduces the local node three-hop neighbor database into the ant colony path selection function, and considers the speed of the node. The original path selection and state update mechanism of the ACO algorithm are improved, and the ant colony optimization is applied to solve the MPR set problem, achieving the purpose of optimizing the MPR set. Finally, it is integrated into the Qualnet network simulation software system, which well verifies the performance of the algorithm proposed by the invention. Specifically including:

[0034] Step S1: Objective function

[0035] In the current MPR selection problem, assume that there are m nodes in total, of which the source node is A, define the one-hop set S1 of the source node, and |S1|=n, the strict two-hop neighbor set S2 of the source node, and the strict three-hop neighbor set S3 of the source node, where |S1|+|S2|+|S3|+1=m. Assume that there are k ants in total, res k Represents the number of elements in the MPR solution set obtained by the kth ant in the current iteration. j ∈{0,1}(j=1,2...n) represents whether the current ant will use s1 j Select into the MPR set. For each ant, its optimal solution For this problem, the global optimal solution bestSolu = min|res k |. The goal of this algorithm is to continuously optimize the res while considering the mobility of drone network nodes. k The value of is chosen so that the final bestSolu obtains the ideal minimum value to meet the needs of the current complex UAV self-organizing network.

[0036] Step S2: Node path selection

[0037] The global search capability of the ant colony algorithm lies in the fact that the choice of the next hop path is not absolute. During the movement of ants, they will dynamically update key information such as pheromones and weights, and select the most suitable path based on these factors. represents the probability that ant k selects node i (i = 1, 2, ..., n) as the next node selected into the MPR set, we can get:

[0038]

[0039] Among them, α represents the heuristic factor of pheromone, β represents the heuristic factor of two-hop weight, γ represents the heuristic factor of three-hop weight, and ε represents the heuristic factor of node relative speed. Among them, α∈(0,1), β∈(0,1), γ∈(0,1), ε∈(0,1). The larger the value of α, the greater the influence of pheromone left by other ants on the current ant. β, γ, ε represent the importance of the corresponding heuristic factors. When α is 1, β, γ, ε are 0, the ant colony algorithm degenerates into a traditional greedy algorithm.

[0040] τ(i) represents the concentration of pheromone held by node i, μ(i) represents the weight of the two-hop neighbors of the source node covered by node i. η(i) represents the weight of the three-hop neighbors of the source node covered by node i. ν(i) represents the formula for the influence of the moving speed of target node i on the probability selection. allowed(S1)k represents the set of nodes that ant k is allowed to join its MPR set.

[0041] For η(i), introducing the local database of three-hop neighbors into the additional decision function of MPR can further optimize MPR

[10] .

[0042] for When speedi>0, it means the node is moving towards itself, and when speedi<0, it means the node is moving away from itself. For the UAV system network, we hope to achieve a stable dynamic balance between nodes as much as possible. When a one-hop node is relatively stationary with the source node in an ideal state, its probability of being an MPR node is the highest. When a one-hop node is too fast relative to the source node, whether it is moving towards or away from the source node, this node is considered unstable, and its probability of being an MPR node decreases as the moving speed increases.

[0043] Step S3: Pheromone Update

[0044] In the ant colony algorithm, pheromones on the path will continue to accumulate as the number of ants passing through increases. Often, the pheromone concentration on a large number of non-optimal paths that have been traveled continues to accumulate, while the pheromone concentration on unselected paths continues to decrease. The truly possible optimal path may exist in the unselected path. At this time, the algorithm falls into a local optimal solution.

[0045] In order to prevent the algorithm from converging to a local optimal solution or converging too slowly, this paper adopts a global pheromone update rule to update the pheromones on the path. Compared with the traditional ACO algorithm that updates the pheromones on all paths that ants pass through to reach the destination node, this paper further accumulates pheromones for the optimal path in the current iteration process and further volatilizes the worst path to punish it. This helps to eliminate relatively poor paths faster and highlight better paths, thereby accelerating the convergence of the algorithm and reducing the probability of the algorithm falling into a local optimum.

[0046] The update formula of pheromone is as follows:

[0047] τ i (t+1)=(1-ρ)·τ i (t)+Δτ i (t) (2)

[0048] Where ρ is the phased pheromone volatilization rate, τ i (t) is the pheromone increment on node i before updating, τ i (t+1) is the pheromone increment on node i after the update, which is specifically defined as follows:

[0049]

[0050] Among them cur Represents the current iteration number, ite max Represents the maximum number of iterations set.

[0051]

[0052] in Represents the pheromone released by ant k on the path to node i during the iteration process. k is the solution of the current iteration. k For example, the smaller the value, the closer it is to the optimal solution of the current iteration, and the greater the pheromone deposition on the path it passes through. Conversely, a penalty is imposed on the path with a larger value, and the pheromone deposition on the path it passes through is significantly reduced, which is conducive to ants choosing a smaller MPR set and obtaining the optimal solution.

[0053] Since the results of each round of iteration may be different, more pheromones will be accumulated on the paths, and the initial value of the volatility rate is higher for the phased pheromone volatility coefficient, which is conducive to the ants to conduct a more comprehensive search of all paths and avoid falling into the local optimum. As the number of iterations increases, the volatility rate continues to decrease, allowing the ants to gradually gather on the optimal path.

[0054] When the pheromone is updated, in order to prevent a certain optimal node from being selected because of its low pheromone content, or because the pheromone concentration on a certain node is too high, causing the path search process to stagnate prematurely and thus miss a better path. Set an upper limit phe for the pheromone of each node max , and set the lower limit phe min , so that the pheromone of the node is always [phe min ,phe max ] within.

[0055] Embodiment 2

[0056] like Figure 2 As shown, the present invention has used the Qualnet network simulation platform to simulate in order to verify the feasibility of the algorithm described in the invention. First, the optimized protocol needs to be inherited into the Qualnet simulator.

[0057] After the protocol is added, build a network simulation scenario in QualnetGUI. Figure 3 Take the network topology of 120 nodes as an example. The green line in the figure is the CBR data service sent by the application layer. Each node uses the Random Waypoint mobile model provided by Qualnet. The scale of the entire scene is 1500m*1500m. The speed range of the node can be set as [minSpeed, maxSpeed] Figure 4 As shown in the figure, 120 nodes are randomly distributed in the network. The simulation time of the network is set to 5 minutes. The maximum transmission range of the node is 200m. The physical layer protocol and MAC layer protocol of the scene are both IEEE 802.11, and the link bandwidth is set to 2Mbps to ensure that the underlying protocol does not affect the accuracy of network layer transmission.

[0058] Under the Qualnet network protocol simulation tool, this method mainly focuses on the average end-to-end delay (Average End-to-end Delay) and the network packet loss rate (PacketLoss Ratio), which represent the following meanings:

[0059] (1) Average end-to-end delay: The end-to-end delay refers to the time it takes for a data packet sent by a source node in the network to reach the destination node. The average end-to-end delay is the average time it takes for all data packets to be sent and successfully received. The calculation formula is as follows:

[0060]

[0061] (2) Network packet loss rate: The network packet loss rate represents the percentage of the total number of data packets sent by the source node but not successfully received by the destination node during the network data transmission process to the total number of data packets sent by the source node. The calculation formula is as follows:

[0062]

[0063] The experiment compares the impact of the M-ACO algorithm proposed in this chapter and the traditional greedy algorithm for solving the MPR minimum set on the performance of the OLSR protocol under different network node scales.

[0064] like Figure 5As shown in the figure, the network topology packet loss rate performance of the OLSR protocol using the M-ACO algorithm and the greedy algorithm under 30, 60, 90, 120, and 150 nodes in the present invention. As can be seen from the figure, when the node scale in the scene is small, the MPR sets selected by the two algorithms are approximately the same, and the impact on the network packet loss rate is not much different. As the scale of the network topology increases, the packet loss rates corresponding to the scenarios using the two algorithms are constantly increasing, but in general, the packet loss rate of the network after using M-ACO is lower, indicating that the MPR set selected by it is better.

[0065] like Figure 6 As shown in the figure, the performance of the end-to-end delay of the network topology under 30, 60, 90, 120, and 150 nodes in the OLSR protocol using the M-ACO algorithm and the greedy algorithm of the present invention. As can be seen from the figure, the end-to-end delay of the protocols using the two algorithms is about 20ms when the scale is 30 nodes. As the network scale increases, the delay of both algorithms continues to increase, but the overall effect of the M-ACO algorithm is better than that of the greedy algorithm, and the effect of improving the end-to-end performance of the network is better.

[0066] The framework of the whole algorithm of the present invention is: the number of ants num_Ants, the current number of iterations ite, the total number of iterations total_iteration, the array of nodes visited by each ant visit, the number of currently uncovered two-hop neighbor nodes Uncover_S2, the current ant selection MPR set size cur_Solu, and the historical optimal MPR set size best_Solu.

[0067] Initialize the source node’s one-hop neighbor set S1, two-hop neighbor set S2, three-hop neighbor set S3, and so on.

[0068] Initialize the starting node of each ant and record it in the visit array of each ant. Initialize the current one-hop neighbor set cur_S1, the current two-hop set cur_S2, and the current three-hop set cur_S3 of each ant.

[0069] The probability of the ant choosing other nodes is calculated based on the previous path probability selection formula (2).

[0070] The next hop of the current ant is selected based on the roulette wheel selection method.

[0071] Update the visit array, cur_S1, cur_S2, and cur_S3 arrays to overwrite the nodes corresponding to the selected nodes.

[0072] For each ant, update cur_Solu and best_Solu according to the current visit array situation.

[0073] The path pheromone is updated according to the method for updating pheromone proposed above.

[0074] The above is only a preferred embodiment of the present invention, and the present invention is not limited to the above embodiment. There may be minor local structural changes during implementation. If various changes or modifications to the present invention do not deviate from the spirit and scope of the present invention and fall within the scope of the claims and equivalent technologies of the present invention, the present invention is also intended to include such changes and modifications.

Claims

1. A large-scale ad hoc network multi-point relay selection method based on ant colony optimization, characterized in that: The method comprises the following steps: Step 1: Initialize the source node's one-hop neighbor set S1, two-hop neighbor set S2, three-hop neighbor set S3, the number of ants num_ants, the array of nodes visited by each ant visit, the number of currently uncovered two-hop neighbor nodes Uncover_S2, and the total number of cycles of the ant colony total_iteration; Step 2: When the current number of iterations is less than the total number of loops, initialize the starting node of each ant and record it in the visit array of each ant, and initialize each ant's current one-hop neighbor set cur_S1, current two-hop set cur_S2, and current three-hop set cur_S3; Step 3: For each ant, when the current Uncover_S2>0, the probability of the ant choosing other nodes is calculated according to the path probability selection formula, and the next hop selection of the current ant is selected according to the roulette wheel selection method. The visit array, cur_S1, cur_S2, and cur_S3 arrays are updated to cover the nodes corresponding to the selected nodes. For each ant, cur_Solu and best_Solu are updated according to the current visit array situation. Step 4: Update the pheromone of the current path. The pheromone update formula is: t i (t+1)=(1-ρ)·τ i (t)+Δτ i (t) (1) Where ρ is the phased pheromone volatilization rate, τ i (t) is the pheromone increment on node i before updating, τ i (t+1) is the pheromone increment on node i after the update; Among them cur Represents the current iteration number, ite max Represents the maximum number of iterations set; in represents the pheromone released by ant k on the path to node i during the iteration process, curse k is the solution of the current iteration, and the Q value is the initial pheromone constant.

2. The method for selecting multi-point relays in a large-scale ad hoc network based on ant colony optimization according to claim 1, characterized in that: In step 3, the path probability selection formula is: Among them, α represents the heuristic factor of pheromone, β represents the heuristic factor of two-hop weight, γ represents the heuristic factor of three-hop weight, ε represents the heuristic factor of node relative speed, τ(i) represents the concentration of pheromone held by node i, μ(i) represents the weight of the two-hop neighbors of the source node covered by node i, η(i) represents the weight of the three-hop neighbors of the source node covered by node i, and v(i) represents the influence formula of the moving speed of target node i on the probability selection.

Images