Wireless ad hoc network resource scheduling simulation method based on position information

Abstract

The invention discloses a wireless ad hoc network resource scheduling simulation method based on position information, link establishment in subnets is carried out according to a fixed allocation time slot strategy, after initial network establishment in the subnets is completed, a DQN-based beam scanning algorithm is adopted in the link establishment stage between the subnets, and the link establishment stage is applied in an actual scene based on an offline training mode. Model parameters after offline training are stored in nodes, a local airspace resource adjacency state table and a received neighbor node adjacency state table serve as input in the actual flight process, beam pointing of the nodes in the next time slot is guided, the neighbor discovery process is accelerated, and finally an airspace resource adjacency state table of the nodes between subnets is formed. And dynamic updating is carried out. According to the invention, the beam pointing is updated by using the node position information in the multi-node flight process in the network so as to improve the neighbor discovery efficiency, and meanwhile, the efficient transmission of the link is maintained in the node flight process.

Description

Technical Field

[0001] The invention relates to a wireless self-organizing network resource scheduling simulation method based on location information, belonging to the field of wireless networks and communications. Background Art

[0002] In the future integrated and information-based communication scenarios, the network will have high dynamics, strong confrontation, and self-organization characteristics. At the same time, collaborative applications between multiple nodes will become the mainstream. In this scenario, in order to improve the anti-interference performance of the network and reduce interference between nodes and network energy consumption, directional communication technology is usually used. Due to the mobility of nodes, the neighbor discovery process needs to be completed quickly in the high-frequency directional networking process. However, the use of directional narrow beams for communication reduces interference while increasing the difficulty of neighbor discovery. Therefore, the network needs to be able to manage and control the beams in a refined manner. Therefore, under directional communication technology, how to intelligently manage airspace resources and align and access dynamic beams between multiple nodes is the first problem that needs to be solved.

[0003] At present, the research on time slot scheduling strategies in the process of neighbor discovery basically assumes that nodes in self-organizing networks use omnidirectional antennas to communicate. Time slot scheduling strategies can be divided into two categories. The first category is a deterministic time slot scheduling strategy, that is, according to a pre-established strategy, the working time slots of network nodes appear regularly and periodically. The advantage of this type of algorithm is that the worst-case discovery delay can be theoretically derived. The disadvantage is that it is not flexible enough and the average discovery delay is high. Typical deterministic scheduling strategies include Disco, U-Connect, SearchLight, etc. The second category is a random time slot scheduling strategy, that is, the working time slots of nodes appear randomly according to a certain probability. The advantage of this type of algorithm is that the average discovery delay is low and it is more flexible. The disadvantage is that it has a long tail effect and the discovery delay has no boundaries. Typical random scheduling strategies include Birthday, PSBA, etc.

[0004] Some scholars have proposed applying the Quorum system to beam scanning. The Quorum system can ensure that any two nodes can achieve beam alignment within a certain period of time, but this method has poor flexibility and cannot cope with changes in the network environment. In addition, some researchers have used reinforcement learning for beam scanning. They define the direction that the beam can point to as a finite set, and use this set as the action space of reinforcement learning. They use the results of neighbor discovery as rewards and perform beam configuration through a trained reinforcement learning model. This method does not take into account the problem of small amount of data and insufficient computing power of a single node in actual scenarios, and is not friendly to newly added nodes in the network because new nodes need to collect data again for learning. Summary of the invention

[0005] The technical problem solved by the present invention is: to overcome the shortcomings of the prior art, to provide a wireless self-organizing network resource scheduling simulation method based on location information, to realize the use of node location information to update beam pointing during the flight of multiple nodes in the network to improve neighbor discovery efficiency, and to maintain efficient link transmission during the flight of the nodes.

[0006] The technical solution of the present invention is: a wireless self-organizing network resource scheduling simulation method based on location information, wherein:

[0007] Step 1. Establish a node motion model, create an offline training scenario based on the node motion model, train the DQN model parameters, and store the trained model parameters in each gateway node;

[0008] Step 2: Link building is performed in the subnet according to the fixed allocation time slot strategy. After the initial link building is completed in the subnet, the DQN-based beam scanning algorithm is used to perform initial link building between subnets. That is, the offline trained DQN model parameters are used to guide the beam scanning direction of the current link building node in the next time slot to perform initial link building between subnets. Each node between subnets forms its own initial airspace resource adjacency state table.

[0009] Step 3. After the initial link between subnets is established, during the actual flight process, each node periodically broadcasts its own location information, and other nodes save it in their own airspace resource adjacency status table after receiving it. When the current sending node A sends information to the adjacent receiving node B, the sending node A queries its own airspace resource adjacency status table to obtain the location information of the receiving node B.

[0010] Step 4. Take the sending node A as the origin of the coordinate system and calculate the two rotation angles α and β corresponding to the receiving node B.

[0011] Step 5. According to α, β and the sector number table, the sector number of the receiving node B is obtained by looking up the table, and the sector number information of the receiving node B in the airspace resource adjacency status table is modified accordingly;

[0012] Step 6: Return to Step 3 for each node whose position has changed and continue to update the information in the airspace resource adjacency status table.

[0013] Preferably, the node motion model includes two parameters: node speed and node motion offset angle. After the parameters are set, during the link establishment phase between subnets, the gateway node divides its own airspace into N sectors.

[0014] Preferably, the DQN model input is the node motion model parameters, the number of sectors N, and the number of one-hop neighbor nodes and two-hop neighbor nodes in each sector; the trained DQN model parameters are the sector number and the corresponding action value function, i.e., the Q value.

[0015] Preferably, the process of establishing a link between subnets based on the DQN beam scanning algorithm is as follows:

[0016] Step 2-1, initialize the airspace resource adjacency status table;

[0017] Step 2-2, the current link-building gateway node runs the trained DQN model parameters offline according to the current airspace resource adjacency status table to obtain the next time slot scanning sector;

[0018] Step 2-3, the current link-building gateway node selects the next time slot receiving or sending state according to the time slot usage mode of the service channel in the initial link-building phase, and points the beam of the current node to the next time slot scanning sector obtained by the DQN model;

[0019] Step 2-4: If the current time slot is in the sending state, the current link-building gateway node sends a hello message to the sector to be scanned, and carries the local airspace resource adjacency state table; if the current time slot is in the receiving state, the neighbor node hello message is received and the local airspace resource adjacency state table is updated;

[0020] Step 2-5, go to Step 2-2, until all the gateway nodes that need to be linked are traversed, and the initial inter-subnet link is completed.

[0021] Preferably, the airspace resource adjacency status table includes neighbor node information of the node and sector information corresponding to the neighbor node.

[0022] Preferably, the current processing node is taken as the origin of the coordinate system, and two rotation angles α and β corresponding to the neighboring nodes are calculated, specifically:

[0023] The coordinates of the sending node A are (x0, y0, z0), and the coordinates of the receiving node B are (x1, y1, z1); where:

[0024]

[0025] If y1 ≥ y0, then

[0026] On the contrary,

[0027] Preferably, each node in the network carries multiple antennas distributed around the node. The number is determined based on actual conditions and should be able to cover at least 360-degree communication around the node.

[0028] When two nodes A and B communicate, the sending node A will calculate the pitch angle and azimuth angle of node B relative to node A based on its own position, its posture and the position of the receiving node B, and select the corresponding antenna for communication.

[0029] Preferably, when two nodes A and B communicate, node A calculates the distance between the two nodes based on its own position and the position of node B, divides the communication rate into different gears according to different communication distances, and determines the corresponding transmission link frame length for information transmission.

[0030] Preferably, when different communication distances are divided into different communication rate gears:

[0031] Short distance corresponds to high transmission rate, longer transmission frame length and larger transmission capacity; long distance corresponds to low transmission rate, shorter transmission frame length and smaller transmission capacity.

[0032] Preferably, the time slot usage of the service channel in the initial link establishment phase is as follows: each node sends or receives randomly in each time slot with a probability of 1 / 2.

[0033] Compared with the prior art, the present invention has the following advantages:

[0034] (1) The present invention proposes a DQN-based intelligent beam scanning method, which uses the model parameters after offline training, the airspace resource adjacency status table during the actual flight process, and the received neighbor node adjacency status table to update the beam pointing, guide the node's beam scanning direction in the next time slot, and speed up the neighbor discovery process;

[0035] (2) The present invention proposes a method for beam alignment using node location information, posture information, and antenna installation position, and dynamically changes the link transmission frame length according to the distance relationship between nodes, fully utilizing link resources and improving transmission capacity. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a schematic diagram of the hierarchical clustering network architecture in the present invention;

[0037] Figure 2 It is a schematic diagram of the process flow of the beam scanning algorithm based on DQN in the present invention;

[0038] Figure 3 Schematic diagram of beam position in the present invention;

[0039] Figure 4 Schematic diagram of the rotation angle in the present invention. DETAILED DESCRIPTION

[0040] Aiming at the needs of quickly completing network construction and inter-node links during flight in large-scale wireless networking scenarios, the present invention proposes an airspace resource scheduling method for quickly discovering neighbors and maintaining efficient link transmission.

[0041] like Figure 1 As shown, it is a schematic diagram of the hierarchical clustering network architecture of the present invention. The subnet to which each node belongs is determined according to network planning. Each subnet is configured with a unique subnet number. Orthogonal frequency hopping is supported between subnets. The network management node manages the node resources in the subnet. Data is exchanged between subnets through gateway nodes. Figure 2 The figure is a flowchart of the beam scanning algorithm based on DQN.

[0042] First, determine the type of this node, i.e., gateway, network management, and ordinary node. Different nodes are responsible for different functional modules, as follows: 1) Ordinary nodes, which have basic communication, forwarding and other functions, and are managed by network management nodes. 2) Network management nodes, based on ordinary nodes, have additional network management functions. Perform identity management and resource management on all nodes in the subnet. 3) Gateway nodes, based on ordinary nodes, have additional inter-subnet cluster communication functions, becoming subnet aggregation points and forwarding points for cross-subnet communication.

[0043] The link building in the subnet is carried out according to the fixed allocation time slot strategy. After the initial network building in the subnet is completed, the link building stage between subnets adopts the beam scanning algorithm based on DQN and is applied in actual scenarios based on offline training. The model parameters after offline training are stored in the node, and the local airspace resource adjacency status table and the received neighbor node adjacency status table are used as input during the actual flight to guide the node's beam pointing in the next time slot, speed up the neighbor discovery process, and finally form the airspace resource adjacency status table of the nodes between subnets. Its main process is as follows Figure 2 The specific steps are as follows:

[0044] 1. Establish a node motion model, which mainly includes two parameters: node speed and node motion offset angle. After setting the parameters, in the inter-subnet link establishment stage, the gateway node divides its own airspace into N sectors. According to the node motion model, an offline training scenario is created to train the DQN model parameters. The DQN algorithm input is the node motion model parameters, the number of sectors N, and the number of one-hop neighbors and two-hop neighbors in each sector (for example, the airspace resource adjacency state table of node 1 is [2,3], and the airspace resource adjacency state table of node 5 is [7,8]. Then, after node 1 and node 5 are connected, nodes 2, 3, and 5 are one-hop neighbors of node 1, and nodes 7 and 8 are two-hop neighbors of node 1). The trained model parameters are stored in each gateway node. The trained DQN model parameters are the sector number and the corresponding Q value (action value function).

[0045] 2. The gateway node runs the trained DQN model according to the current status (the airspace resource adjacency status table (mainly including neighbor nodes and the sectors where the neighbor nodes are located) and its own airspace resource adjacency status table sent by the neighbor nodes) to obtain the sector number to be scanned in the next time slot (that is, the sector number corresponding to the large value of the Q value).

[0046] 3. The node selects the receiving / transmitting state of the next time slot according to the time slot usage of the service channel in the initial network construction phase, and directs the beam of the current node to the scanning sector of the next time slot. The time slot usage of the service channel in the initial network construction phase is as follows: each node sends or receives randomly in each time slot, with a probability of 1 / 2.

[0047] 4. If this time slot is in the sending state, the gateway node sends a hello message in the direction of the sector to be scanned, and carries the local airspace resource adjacency status table (mainly including its own neighbor nodes and the sectors where the neighbor nodes are located).

[0048] 5. If the current time slot is in the receiving state and a hello message is received from a neighboring node, the local airspace resource adjacency state table and the neighbor adjacency state table are updated, the current state is updated, and step 2 is performed until the initial network establishment process is completed.

[0049] In three-dimensional space, the position of any point can be represented by coordinates (x, y, z). The line connecting the space point and the origin is denoted as u, the projection of u on the XOY plane of the body coordinate system of the current node (sending node A) is denoted as v, the rotation angle from the positive direction of the X axis to the projection v on the XOY plane is denoted as α, α∈[0,2π), and the rotation angle from the positive direction of the Z axis to the line u to u is denoted as β, β∈[0,π], as follows: Figure 4 The specific position of a point in space can be determined by the two rotation angles α and β, and its specific coordinates can be obtained by the distance between the space point and the origin.

[0050] The beam alignment algorithm based on the position information during the flight of the node calculates the deviation angle of the neighboring node based on the position information of the neighboring node, its own position and attitude information, and thus updates the sector information where the neighboring node is located. The specific algorithm flow is as follows:

[0051] 1. After the initial link between subnets is established, during the actual flight process, each node periodically broadcasts its own location information, and other nodes save it in their own neighbor location information table after receiving it. When the network completes the initial link establishment and operates normally, taking the current sending node A and the adjacent receiving node B as an example, when node A wants to send information to node B, node A knows its own location, and then queries its saved neighbor location information table to obtain the location information of node B;

[0052] 2. Get the coordinates (x0, y0, z0) of node A (sending node) and the coordinates (x1, y1, z1) of its neighbor node B (receiving node) at this moment;

[0053] 3. According to Figure 3 Schematic diagram of the rotation angle, calculate the two rotation angles α and β corresponding to the neighbor node B when the node A is the origin of the coordinate system: If y1 ≥ y0, then On the contrary,

[0054] 4. Based on α and β and the sector numbering rule, the table is searched to obtain the sector number num_1 where the neighbor node B is located at this moment. The sector numbering rule is as follows:

[0055] Since each beam uniquely corresponds to a set of azimuth and elevation angles (α, β), the beam sector number and its corresponding (α, β) are shown in the following table, where x is the beam lobe width,

[0056] Sector number β β 1 [0,x) [0,x) 2 [0,x) [x,2x) 3 [0,x) [2x,3x) … … … c [0,x) [(c-1)x,π] c+1 [x,2x) [0,x) c+2 [x,2x) [x,2x) … … … 2c [x,2x) [(c-1)x,π] … … … 3c [2x,3x) [(c-1)x,π] … … … r*c-1 [(r-1)x，2π),2π) [(c-2)x, (c-1)x)x) r*c [(r-1)x，2π),2π) [(c-1)x,π]

[0057] 5. Change the sector number of the neighbor node in the neighbor location information table to num_1;

[0058] 6. Each node in the network carries multiple antennas, which are distributed around the node. The number is determined based on the actual situation and needs to be able to provide 360-degree communication around the node. When two nodes A and B communicate, node A will calculate the pitch angle and azimuth angle of node B relative to node A based on its own position, its posture and the position of node B, and select the corresponding antenna for communication.

[0059] 7. Calculate the distance between node A and node B based on the position coordinates of node A and node B.

[0060]

[0061] 8. Divide the communication rate into different gears according to different communication distances, determine the corresponding transmission link frame length for information transmission, and transmit information in different gears to utilize link resources and increase transmission capacity.

[0062] Short distance corresponds to high-speed transmission rate, longer transmission frame length and larger transmission capacity; long distance corresponds to low-speed transmission rate, shorter transmission frame length and smaller transmission capacity; different rate gears are divided to adaptively change the transmission capacity according to the distance, which can make better use of resources compared to a fixed transmission rate.

[0063] The contents not described in detail in the specification of the present invention belong to the prior art known to the professional and technical personnel in this field.

Claims

1. A wireless self-organizing network resource scheduling simulation method based on location information, characterized in that include: Step 1. Establish a node motion model, create an offline training scenario based on the node motion model, train the DQN model parameters, and store the trained model parameters in each gateway node; Step 2: Establish links in the subnet according to the fixed allocation time slot strategy; After the initial link building is completed within the subnet, the DQN-based beam scanning algorithm is used to perform initial link building between subnets. That is, the offline trained DQN model parameters are used to guide the beam scanning direction of the current link building node in the next time slot to perform initial link building between subnets. Each node between subnets forms its own initial airspace resource adjacency state table. Step 3. After the initial link between subnets is established, during the actual flight process, each node periodically broadcasts its own location information, and other nodes save it in their own airspace resource adjacency status table after receiving it. When the current sending node A sends information to the adjacent receiving node B, the sending node A queries its own airspace resource adjacency status table to obtain the location information of the receiving node B. Step 4. Take the sending node A as the origin of the coordinate system and calculate the two rotation angles α and β corresponding to the receiving node B. Step 5. According to α, β and the sector number table, the sector number of the receiving node B is obtained by looking up the table, and the sector number information of the receiving node B in the airspace resource adjacency status table is modified accordingly; Step 6: Return to Step 3 for each node whose position has changed and continue to update the information in the airspace resource adjacency status table.

2. The method for simulating wireless self-organizing network resource scheduling based on location information according to claim 1, characterized in that: The node motion model includes two parameters: node speed and node motion offset angle. After setting the parameters, during the link establishment phase between subnets, the gateway node divides its own airspace into N sectors.

3. The method for simulating wireless self-organizing network resource scheduling based on location information according to claim 1, characterized in that: The input of the DQN model is the node motion model parameters, the number of sectors N, and the number of one-hop neighbor nodes and two-hop neighbor nodes in each sector; The parameters of the trained DQN model are the sector number and the corresponding action value function, i.e., the Q value.

4. The method for simulating wireless ad hoc network resource scheduling based on location information according to claim 1, characterized in that: The process of establishing links between subnets based on the DQN beam scanning algorithm is as follows: Step 2-1, initialize the airspace resource adjacency status table; Step 2-2, the current link-building gateway node runs the trained DQN model parameters offline according to the current airspace resource adjacency status table to obtain the next time slot scanning sector; Step 2-3, the current link-building gateway node selects the next time slot receiving or sending state according to the time slot usage mode of the service channel in the initial link-building phase, and points the beam of the current node to the next time slot scanning sector obtained by the DQN model; Step 2-4, if the current time slot is in the sending state, the current link-building gateway node sends a hello message to the sector to be scanned, and carries the local airspace resource adjacency state table; If the current time slot is in the receiving state, the neighbor node hello message is received and the local airspace resource adjacency state table is updated; Step 2-5, go to Step 2-2, until all the gateway nodes that need to be linked are traversed, and the initial inter-subnet link is completed.

5. The method for simulating wireless self-organizing network resource scheduling based on location information according to claim 1, characterized in that: The airspace resource adjacency status table includes the neighbor node information of the current node and the sector information of the corresponding neighbor node.

6. The method for simulating wireless ad hoc network resource scheduling based on location information according to claim 1, characterized in that: Taking the current processing node as the origin of the coordinate system, calculate the two rotation angles α and β corresponding to the neighboring nodes, specifically: The coordinates of the sending node A are (x0, y0, z0), and the coordinates of the receiving node B are (x1, y1, z1); where: If y1 ≥ y0, then On the contrary, 7. The method for simulating wireless ad hoc network resource scheduling based on location information according to claim 1, characterized in that: Each node in the network carries multiple antennas distributed around the node. The number is determined based on actual conditions and should be able to cover at least 360 degrees of communication around it. When two nodes A and B communicate, the sending node A will calculate the pitch angle and azimuth angle of node B relative to node A based on its own position, its posture and the position of the receiving node B, and select the corresponding antenna for communication.

8. The method for simulating wireless self-organizing network resource scheduling based on location information according to claim 1, characterized in that: When two nodes A and B communicate, node A calculates the distance between the two nodes based on its own position and the position of node B, divides the communication rate into different gears according to different communication distances, and determines the corresponding transmission link frame length for information transmission.

9. A method for simulating wireless self-organizing network resource scheduling based on location information according to claim 8, characterized in that: When different communication distances are divided into different communication rate levels: Short distance corresponds to high transmission rate, longer transmission frame length and larger transmission capacity; long distance corresponds to low transmission rate, shorter transmission frame length and smaller transmission capacity.

10. The method for simulating wireless self-organizing network resource scheduling based on location information according to claim 1, characterized in that: The time slot usage of the service channel in the initial link establishment phase is as follows: each node sends or receives randomly in each time slot with a probability of 1 / 2.

Images