Unmanned aerial vehicle cluster low-cost control system and method based on SDN and NFV

Abstract

The invention discloses a low-cost unmanned aerial vehicle cluster control system and method based on an SDN and an NFV. A task arrangement engine is used for analyzing task requirements into service function chains; the NFV resource management platform is used for receiving each service function chain and dynamically allocating the VNF resources in the NFV resource pool based on a dynamic scheduling model to obtain an optimal node combination of the optimal configuration of the VNF resources; the SDN integrated controller is used for receiving the optimal node combination and optimizing a communication path of the optimal node combination based on a flow scheduling model and a load balancing model to obtain an optimized network topology and an optimized communication link; and the distributed communication module is used for receiving the optimized network topology and the communication link, carrying out dynamic adaptation on the communication link and the network topology of each unmanned aerial vehicle based on a lightweight protocol, realizing a logic concentration and physical dispersion architecture through cooperation of an SDN and an NFV, and elastically expanding the cluster scale.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle cluster control and network communication, and particularly relates to an unmanned aerial vehicle cluster low-cost control system and method based on SDN and NFV. BACKGROUND

[0002] The unmanned aerial vehicle cluster has wide application in disaster relief, agricultural monitoring and other fields, but the traditional unmanned aerial vehicle cluster control method has the following problems:

[0003] (1) Low hardware resource utilization: each unmanned aerial vehicle needs to be independently configured with special hardware (such as a computing unit, a sensor, and a communication module), resulting in resource redundancy and cost increase.

[0004] (2) Poor flexibility and scalability: hardware and functions are tightly coupled, making it difficult to dynamically adapt to changes in task requirements or quickly expand the cluster size.

[0005] (3) Low control efficiency: relying on a central node or manual intervention, communication delay is high, and dynamic task response capability is insufficient.

[0006] (4) Strong dependence on manpower: manual configuration of network topology and task allocation is required, and the operation and maintenance cost is high. SUMMARY

[0007] The main purpose of the present application is to provide an unmanned aerial vehicle cluster low-cost control system and method based on SDN and NFV, which aims to virtualize all unmanned aerial vehicle hardware resources into a global resource pool, combine the centralized control capability of SDN, realize dynamic resource scheduling, flexible task deployment and efficient collaboration, significantly reduce hardware cost and manual dependence, and improve resource utilization and control efficiency.

[0008] To achieve the above purpose, the present application provides an unmanned aerial vehicle cluster low-cost control system based on SDN and NFV, comprising: a task orchestration engine for parsing task requirements into service function chains and sending to an NFV resource management platform; the NFV resource management platform is used for receiving each service function chain and dynamically allocating VNF resources in the NFV resource pool based on a dynamic scheduling model to obtain an optimal node combination of VNF resource optimal configuration and send to an SDN centralized controller, wherein the NFV resource pool is obtained based on virtualization of the computing, storage and sensor of all unmanned aerial vehicles in the unmanned aerial vehicle cluster; the SDN centralized controller is used for receiving the optimal node combination and optimizing the communication path of the optimal node combination based on a traffic scheduling model and a load balancing model to obtain an optimized network topology and communication link and send to a distributed communication module; the distributed communication module is used for receiving the optimized network topology and communication link and dynamically adapting the communication link and network topology of each unmanned aerial vehicle based on a lightweight protocol.

[0009] Optionally, the dynamic scheduling model comprises a resource utilization optimization sub-model and a service function chain orchestration sub-model; the resource utilization optimization sub-model is configured to receive each service function chain and perform dynamic allocation of VNF resources in the NFV resource pool based on a greedy algorithm or a linear programming algorithm; and the service function chain orchestration sub-model is configured to receive each service function chain and output a service function chain with minimum delay.

[0010] Optionally, the system further comprises a virtualization environment configured to abstract the local VNF resources into a standardized interface and run a resource registration protocol on the standardized interface to report the local VNF resources to the NFV management platform.

[0011] Optionally, the process of constructing the optimal node combination problem comprises calculating a total resource consumption of the VNF chain, calculating a total task completion time of the VNF, and constructing the optimal node combination problem according to a sum of the total resource consumption and the total task completion time.

[0012] Optionally, the SDN controller is further configured to construct a network state graph; the traffic scheduling model is configured to process the network state graph and solve a shortest communication link in the network state graph based on a Dijkstra algorithm with a weight sum of the communication links as an objective function; and the load balancing degree model is configured to process the network state graph to obtain a flow table rule with a load balancing degree less than a preset value.

[0013] Optionally, the system further comprises a node real-time monitoring module configured to calculate a node health degree based on a preset first expression and calculate a network load based on a second expression.

[0014] Optionally, the system further comprises an adaptive module configured to trigger the NFV resource management platform to perform automatic expansion when a remaining VNF resource in the NFV resource pool is less than a preset threshold value, and migrate a VNF resource to another drone node when the node health degree is less than a preset value, wherein a migration time is less than or equal to a ratio of a VNF resource state change time to a bandwidth of the drone node.

[0015] Optionally, the system further comprises a communication mode selection module configured to select a communication mode of each drone according to an environmental condition.

[0016] Optionally, an expression of the mixed integer programming model of the task orchestration is as follows:

[0017]

[0018] x ij ∈{0,1}

[0019] wherein x ij represents whether the task j is assigned to the node i, c ijdenotes the resource cost, λ denotes the delay weight coefficient, T total is the task completion time, R i denotes the resource vector of the node, D j denotes the task demand resource vector.

[0020] Optionally, the expression of the global network state is:

[0021]

[0022] wherein α, β, γ are adjustment coefficients, and the adjustment coefficients are adjusted according to the scene.

[0023] In order to achieve the above purpose, the application further provides a low-cost control method for a UAV cluster based on SDN and NFV, which is applied to the low-cost control system for the UAV cluster based on SDN and NFV provided in any one of the preceding embodiments, and the method comprises the following steps:

[0024] parsing the task demand into each service function chain;

[0025] dynamically allocating VNF resources in an NFV resource pool based on a dynamic scheduling model to obtain an optimal node combination of VNF resource optimal configuration, wherein the NFV resource pool is obtained based on the virtualization of the calculation, storage and sensor of all UAVs in the UAV cluster;

[0026] optimizing the communication path of the optimal node combination based on a traffic scheduling model and a load balancing model to obtain an optimized network topology and communication link;

[0027] dynamically adapting the communication link and the network topology of each UAV based on a lightweight protocol.

[0028] The present application proposes a low-cost control system and method for drone swarms based on SDN and NFV. A task orchestration engine parses task requirements into service function chains and sends these chains to an NFV resource management platform. The NFV resource management platform receives these service function chains and dynamically allocates VNF resources in the NFV resource pool based on a dynamic scheduling model. The optimal node combination for optimal VNF resource configuration is obtained and sent to an SDN centralized controller. The NFV resource pool is derived from the virtualization of computing, storage, and sensors across all drones in the drone swarm. Each drone abstracts its local hardware resources (CPU, GPU, sensors, etc.) into virtual resources. The resource pool supports dynamic expansion and contraction, with new drones automatically added to the resource pool and resources from faulty nodes automatically migrated to other nodes. Based on task requirements, the NFV management platform combines VNF chains (such as "image processing VNF + navigation VNF + communication VNF") from the resource pool to achieve on-demand function loading; the SDN centralized controller receives the optimal node combination, and optimizes the communication path of the optimal node combination based on the traffic scheduling model and load balancing model, obtains the optimized network topology and communication link and sends it to the distributed communication module, dynamically adjusts the data flow path to avoid link congestion. The flow table is sent down through the OpenFlow protocol to support multi-hop self-organizing network communication and reduce dependence on the central base station. Combined with the resource pool status, low-latency paths and high-computing power resources are allocated to high-priority tasks; the distributed communication module receives the optimized network topology and communication link, and dynamically adapts the communication link and network topology of each drone based on a lightweight protocol. This application realizes the "logical centralization + physical dispersion" architecture through the collaboration of SDN and NFV, and the cluster scale can be elastically expanded. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a system architecture diagram of an embodiment of a low-cost control system for drone swarms based on SDN and NFV.

[0030] Figure 2 A low-cost control flow chart provided for an embodiment of a low-cost control system for drone swarms based on SDN and NFV in this application;

[0031] Figure 3 This is an implementation flowchart provided for an embodiment of a low-cost control system for drone clusters based on SDN and NFV in this application.

[0032] The realization of the objectives, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0033] It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application.

[0034] The main solution of the embodiment of this application is: to virtualize all drone hardware resources into a global resource pool through NFV technology, and combine it with the centralized control capabilities of SDN to achieve dynamic resource scheduling, flexible task deployment and efficient collaboration, significantly reduce hardware costs and manual dependence, and improve resource utilization and control efficiency.

[0035] While SDN improves network management efficiency by separating the control and data planes, and NFV decouples hardware resources and enables on-demand allocation through virtualization, their collaborative application in drone swarms is still immature, particularly due to a technological gap in combining low cost with high flexibility.

[0036] This application provides a solution that: (1) improves resource utilization: pooling hardware resources increases utilization by over 40%, reduces resource fragmentation to less than 15%, and reduces the cost of single-machine configuration. (2) optimizes control efficiency: SDN centralized control reduces manual intervention and reduces task response latency by 30%. (3) enhances scalability: adding new drones is plug-and-play, and system expansion time is shortened to minutes. (4) reduces operation and maintenance costs: automated resource scheduling and fault recovery reduce manpower requirements by 80%.

[0037] Reference Figure 1 The first embodiment of the present application provides a low-cost control system for a drone cluster based on SDN and NFV. The low-cost control system 100 for the drone cluster may include a task orchestration engine 101, an NFV resource management platform 102, an SDN centralized controller 103, and a distributed communication module 104.

[0038] The task orchestration engine 101 is used to parse the task requirements into service function chains and send them to the NFV resource management platform 102;

[0039] The task scheduling engine 101 calls the virtualized function unit (VNF) in the resource pool according to the task requirements and generates the optimal task execution plan. The mixed integer programming model of task scheduling is as follows:

[0040]

[0041]

[0042] x ij ∈{0,1}

[0043] Among them, x ij Indicates whether task j is assigned to node i, c ij is the resource cost, λ is the delay weight coefficient, T total The time when the task is completed.

[0044] The NFV resource management platform 102 is configured to receive each service function chain and dynamically allocate VNF resources in the NFV resource pool based on a dynamic scheduling model to obtain an optimal node combination of VNF resource optimal configuration and send the optimal node combination to the SDN centralized controller 103, wherein the NFV resource pool is obtained based on virtualization of computing, storage and sensor of all unmanned aerial vehicles in the unmanned aerial vehicle cluster;

[0045] Specifically, the NFV resource management platform 102 virtualizes the computing, storage, sensor and other hardware resources of all unmanned aerial vehicles in the unmanned aerial vehicle cluster into a unified resource pool to realize on-demand dynamic allocation. The dynamic scheduling model mainly includes a resource utilization optimization sub-model and a service function chain (SFC) arrangement sub-model.

[0046] That is, the dynamic scheduling model includes:

[0047] a resource utilization optimization sub-model and a service function chain arrangement sub-model;

[0048] The resource utilization optimization sub-model is configured to receive each service function chain and dynamically allocate VNF resources in the NFV resource pool based on a greedy algorithm or a linear programming algorithm.

[0049] For example, it is assumed that the unmanned aerial vehicle cluster has N unmanned aerial vehicles, i.e. N nodes, and the resource vector of each node is R i =(r i1 ,r i2 ,…r ik ), where r ik represents the kth resource (such as CPU, memory, sensor) of the ith node. The total resource of the unmanned aerial vehicle cluster is:

[0050]

[0051] The task demand resource vector is D j =(d j1 ,d j2 ,…d jk ), and the goal is to minimize resource fragmentation, i.e.

[0052]

[0053] By using the greedy algorithm or the linear programming to dynamically allocate resources, D j ≤R total .

[0054] The service function chain arrangement sub-model is configured to receive each service function chain and output a service function chain with the minimum delay.

[0055] For example, a task is composed of an SFC chain of m virtual network functions (VNF), and the delay of the SFC chain is defined as:

[0056]

[0057] wherein, is the processing time of the kth VNF, is the data transmission delay between VNFs. The combination of nodes with minimum delay in the resource pool is selected by dynamic programming.

[0058] In an embodiment of the present application, the system further comprises a virtualization environment for abstracting the local VNF resources into a standardized interface and running a resource registration protocol on the standardized interface to report the local VNF resources to the NFV management platform.

[0059] Exemplarily, each drone is pre-installed with a lightweight virtualization environment (such as Docker) to abstract the local resources into a standardized interface:

[0060] Node_Profile = (CPU_cores, RAM_GB, Sensor_List, Location) (1)

[0061] The resource registration protocol of the standardized interface is based on MQTT or CoAP to report the VNF resource information to the NFV management platform.

[0062] The SDN centralized controller 103 is configured to receive the optimal node combination and optimize the communication path of the optimal node combination based on a traffic scheduling model and a load balancing model to obtain an optimized network topology and communication link and send to the distributed communication module 104;

[0063] The construction process of the optimal node combination problem comprises:

[0064] calculating the total resource consumption of the VNF chain;

[0065] calculating the total task completion time of the VNF;

[0066] constructing the optimal node combination problem according to the sum of the total resource consumption and the total task completion time.

[0067] Specifically, the NFV platform calculates the optimal node combination based on the following formula:

[0068]

[0069] The constraint conditions are as follows:

[0070] ΣNode_Resources ≥ ΣVNF_Demand( resource type)

[0071] In an embodiment of the present application, the SDN controller is further configured to construct a network state graph.

[0072] The traffic scheduling model is configured to process the network state graph, and solve the shortest communication link in the network state graph based on a Dijkstra algorithm with a weight sum of the communication link as an objective function.

[0073] The load balancing degree model is configured to process the network state graph to obtain a flow table rule with a load balancing degree less than a preset value.

[0074] In an embodiment of the present application, the SDN controller is further configured to construct a network state graph.

[0075]

[0076] wherein α and β are adjustment coefficients, BW(e) is a link bandwidth, and Delay(e) is a transmission delay.

[0077] The Dijkstra algorithm is used to solve the shortest path, and an objective function of the traffic scheduling model is:

[0078]

[0079] wherein P is a path set.

[0080] To prevent link congestion, a link load balancing degree of the load balancing degree model is defined as:

[0081]

[0082] The dynamic adjustment of the flow table rule ensures that all links satisfy L(e)≤θ (θ is a threshold value, usually 0.8).

[0083] The distributed communication module 104 is configured to receive the optimized network topology and communication link, and dynamically adapt the communication link and network topology of each unmanned aerial vehicle based on a lightweight protocol.

[0084] The distributed communication module 104 is configured to realize low-delay communication based on a lightweight protocol (such as MQTT), and support dynamic network topology adaptation.

[0085] In an embodiment of the present application, the system further comprises:

[0086] The node real-time monitoring module 105 is configured to calculate a node health degree based on a preset first expression, and calculate a network load based on a second expression.

[0087] In an embodiment of the present application, the system further comprises:

[0088] The adaptive module 106 is used to trigger the NFV resource management platform to automatically expand capacity when the remaining VNF resources in the NFV resource pool are less than a preset threshold;

[0089] When the node health is less than the preset value, the VNF resources are migrated to other drone nodes, where the migration time is less than or equal to the ratio of the VNF resource state change time to the drone node bandwidth.

[0090] In one embodiment of the present application, the system further includes:

[0091] The communication mode selection module 107 is used to select the communication mode of each UAV according to the environmental conditions.

[0092] refer to Figure 2 This paper provides a low-cost control method for drone swarms based on SDN and NFV, applicable to any drone swarm scenario requiring low cost, high flexibility, and multi-task collaboration. Regardless of the application scenario, this method is executed in the following four phases, each supporting modular configuration:

[0093] Phase 1: Hardware deployment and resource virtualization

[0094] (1) Node selection:

[0095] Core equipment: low-cost edge computing equipment as the main control unit.

[0096] Sensors / actuators: Optional according to scenario requirements (such as cameras, temperature and humidity sensors, robotic arms, etc.), with costs controlled at ≤500 yuan / node.

[0097] Communication module: LoRa+Wi-Fi dual-mode as standard, 4G optional (can be omitted in cost-sensitive scenarios).

[0098] Phase 2: Task Modeling and Resource Matching

[0099] (1) Task decomposition:

[0100] Input task requirements (e.g., "real-time video analysis + data return") are parsed by the task engine into an SFC chain:

[0101] Task_Chain = [VNF1(type, resource requirements), ..., VNF n ]

[0102] For example: agricultural inspection [image acquisition, edge AI inference, result compression, transmission]; logistics distribution [path planning, obstacle avoidance detection, package status monitoring].

[0103] (2) Dynamic resource allocation:

[0104] The NFV platform calculates the optimal node combination based on the following formula:

[0105]

[0106] The constraints are as follows:

[0107] ∑Node_Resources ≥ ∑VNF_Demand( Resource Type)

[0108] Allocation strategy: for compute-intensive tasks, prefer to allocate high CPU / GPU nodes; for delay-sensitive tasks, select nearby nodes and allocate high-bandwidth links.

[0109] Stage 3: Network optimization and task execution

[0110] (1) SDN control strategy:

[0111] The global network state is modeled as a graph G(V, E), where the nodes V are drones, and the edge E weights are:

[0112]

[0113] (coefficients α, β, γ are adjusted according to the scene, for example, disaster rescue focuses on energy consumption, and logistics distribution focuses on delay)

[0114] Path calculation: use improved Dijkstra algorithm.

[0115] (2) Communication mode selection:

[0116] Dynamic switching logic:

[0117]

[0118] where D th , Th BW are threshold values that can be configured according to the actual environment, for example, D th = 1km in urban environment, D th = 5km in the wild.

[0119] Stage 4: Monitoring and elastic expansion and contraction:

[0120] In an embodiment of the present application, the system further comprises a node real-time monitoring module for calculating node health based on a preset first expression, and calculating network load based on a second expression.

[0121] Specifically, the node health is represented by the first expression:

[0122]

[0123] The network load is represented by the second expression:

[0124]

[0125] In an embodiment of the present application, the system further comprises an adaptive module, configured to trigger the NFV resource management platform 102 to perform automatic expansion when the remaining VNF resources in the NFV resource pool are less than a preset threshold value; and migrate the VNF resources to other UAV nodes when the node health degree is less than a preset value, wherein the migration time is less than or equal to the ratio of the VNF resource state change time to the bandwidth of the UAV node.

[0126] Specifically, horizontal expansion: when Resource_Free<Threshold, triggering automatic expansion (adding nodes or waking up dormant nodes). Fault recovery: if H i <0.2, migrate its VNF to other nodes, and the migration time T migrate ≤VNF_State_Size / BW.

[0127] Reference Figure 3 On the basis of the above embodiments, the present application further provides a low-cost control method for a UAV cluster based on SDN and NFV, applied to the low-cost control system for a UAV cluster based on SDN and NFV provided in any one of the preceding embodiments, and the method comprises the following steps:

[0128] S10. Parsing a task requirement into service function chains;

[0129] S20. Dynamically allocating VNF resources in an NFV resource pool based on a dynamic scheduling model to obtain an optimal node combination with optimal VNF resource configuration, wherein the NFV resource pool is obtained based on the virtualization of the computing, storage and sensor of all UAVs in the UAV cluster;

[0130] S30. Optimizing the communication path of the optimal node combination based on a traffic scheduling model and a load balancing model to obtain an optimized network topology and communication link;

[0131] S40. Dynamically adapting the communication link and network topology of each UAV based on a lightweight protocol.

[0132] The above is only a preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation using the content of the specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A low-cost control system for drone swarms based on SDN and NFV, characterized by: The drones in the drone swarm include: The task orchestration engine is used to parse task requirements into service function chains and send them to the NFV resource management platform; The NFV resource management platform receives service function chains and dynamically allocates VNF resources in the NFV resource pool based on a dynamic scheduling model. It obtains the optimal node combination for the optimal VNF resource configuration and sends it to the SDN centralized controller. The NFV resource pool is based on the virtualization of computing, storage, and sensors of all drones in the drone cluster. The SDN centralized controller receives the optimal node combination and optimizes the communication path of the optimal node combination based on the traffic scheduling model and load balancing model. It obtains the optimized network topology and communication link and sends it to the distributed communication module. The distributed communication module is used to receive the optimized network topology and communication links, and dynamically adapt the communication links and network topology of each drone based on a lightweight protocol.

2. The low-cost control system for drone swarms based on SDN and NFV as claimed in claim 1, characterized in that: The dynamic scheduling model includes: Resource utilization optimization sub-model and service function chain orchestration sub-model; The resource utilization optimization sub-model is used to receive each service function chain and dynamically allocate VNF resources in the NFV resource pool based on a greedy algorithm or a linear programming algorithm; The service function chain orchestration sub-model is used to receive various service function chains and output the service function chain with the minimum delay.

3. The low-cost control system for drone swarms based on SDN and NFV as claimed in claim 1, characterized in that: Also includes: The virtualized environment is used to abstract local VNF resources into standardized interfaces and run the resource registration protocol on the standardized interfaces to report local VNF resources to the NFV management platform.

4. The low-cost control system for drone swarms based on SDN and NFV as claimed in claim 2, characterized in that: The construction process of the optimal node combination problem includes: Calculate the total resource consumption of the VNF chain; Calculate the total VNF task completion time; The optimal node combination problem is constructed based on the sum of resource consumption and task completion time.

5. The low-cost control system for drone swarms based on SDN and NFV as claimed in claim 1, characterized in that: The SDN controller is also used to construct a network status diagram; The traffic scheduling model is used to process the network state graph, and takes the minimized weight sum of the communication links as the objective function, and solves the shortest communication link in the network state graph based on the Dijkstra algorithm; The load balancing model is used to process the network state diagram to obtain a flow table rule in which the load balancing degree is less than a preset value.

6. The low-cost control system for drone swarms based on SDN and NFV as claimed in claim 1, characterized in that: The system further comprises: The node real-time monitoring module is used to calculate the node health based on a preset first expression, and to calculate the network load based on a second expression.

7. The low-cost control system for drone swarms based on SDN and NFV as claimed in claim 1, characterized in that: The system further comprises: The adaptive module is used to trigger the NFV resource management platform to automatically expand capacity when the remaining VNF resources in the NFV resource pool are less than the preset threshold; When the node health is less than the preset value, the VNF resources are migrated to other drone nodes, where the migration time is less than or equal to the ratio of the VNF resource state change time to the drone node bandwidth.

8. The low-cost control system for drone swarms based on SDN and NFV as claimed in claim 1, characterized in that: The system further comprises: The communication mode selection module is used to select the communication mode of each drone according to the environmental conditions.

9. The low-cost control system for drone swarms based on SDN and NFV as claimed in claim 1, characterized in that: The expression of the mixed integer programming model of the task scheduling is: x ij ∈{0,1} Among them, x ij Indicates whether task j is assigned to node i, c ij represents resource cost, λ represents delay weight coefficient, T total is the task completion time, R i Represents the resource vector of the node, D j Represents the resource vector required by the task.

10. A low-cost control method for drone clusters based on SDN and NFV, characterized in that: The method applied to the low-cost control system for drone swarms based on SDN and NFV as described in any one of claims 1 to 9 includes: Analyze task requirements into service function chains; Dynamically allocate VNF resources in the NFV resource pool based on a dynamic scheduling model to obtain the optimal node combination for optimal VNF resource configuration. The NFV resource pool is based on the virtualization of computing, storage, and sensors of all drones in the drone cluster. Optimize the communication path of the optimal node combination based on the traffic scheduling model and load balancing model to obtain the optimized network topology and communication link; The communication links and network topology of each UAV are dynamically adapted based on a lightweight protocol.

Images