GPU parallel computing driven large-scale unmanned aerial vehicle cluster simulation method

Abstract

The invention discloses a GPU parallel computing driven large-scale unmanned aerial vehicle cluster simulation method, which comprises the following steps: S1, building an unmanned aerial vehicle numerical simulation model, generating a code and optimizing the code; s2, optimizing a memory access mode of cluster parallel simulation; s3, optimizing a simulation task division strategy according to the characteristics of unmanned aerial vehicle cluster state updating; s4, according to the S1, S2 and S3, cluster parallel simulation is achieved on the code level; s5, realizing acceleration of a cluster formation algorithm by using parallel computing strategies of S2 and S3; and S6, integrating the cluster parallel simulation in S4 and the formation control algorithm in S5 by using a UDP protocol. According to the multi-core parallel computing architecture based on the GPU, efficient simulation of the large-scale unmanned aerial vehicle cluster is achieved, and the technical problems that a traditional CPU simulation method is poor in real-time performance, difficult to achieve and the like in simulation of more than thousands of large-scale unmanned aerial vehicle clusters are solved.

Description

Technical Field

[0001] This invention belongs to the field of high-performance computing and UAV swarm simulation, specifically involving a GPU parallel computing-driven method for simulating large-scale UAV swarms. Background Technology

[0002] Unmanned aerial vehicle (UAV) swarms can accomplish complex tasks such as target reconnaissance and combined strikes through multi-UAV collaborative operations, offering advantages such as large operational radius, wide reconnaissance range, and low life-cycle cost. In recent years, UAV swarms have been increasingly widely used in human production and daily life fields such as flight performances, express delivery and logistics, precision agriculture, and urban transportation. UAV swarm simulation technology is an important tool for studying UAV swarm systems. For large-scale UAV simulation, the main problem is that as the simulation scale continues to increase, the economic and time costs required for traditional simulations based on central processing units (CPUs) also increase significantly, and real-time performance deteriorates severely. Therefore, improving the simulation efficiency of large-scale UAV swarms has become a crucial issue that urgently needs to be addressed.

[0003] Currently, drone swarm simulation mainly relies on traditional CPU simulation methods, which are primarily divided into two categories: single-machine simulation and distributed parallel simulation. Single-machine simulation refers to implementing the entire simulation process on a single computing node, offering advantages such as simple development and low implementation cost. However, its scalability is significantly limited by the computing resources of a single node.

[0004] The introduction of distributed parallel simulation technology provides a relatively effective solution for the development of large-scale cluster simulation. This technology improves the real-time performance and computational efficiency of simulation by breaking down complex simulation tasks and distributing them to multiple computing units in a computer cluster for parallel execution. While distributed parallel simulation can effectively alleviate the computational resource limitations of single-machine simulation, supporting larger-scale UAV swarm simulations typically requires increasing the number of CPU nodes to enhance computing power. As the node scale increases, the complexity of system communication grows exponentially, not only increasing the difficulty of data synchronization and collaborative control but also leading to a significant increase in system construction and maintenance costs. Under the dual constraints of cost control and real-time requirements, existing simulation scales are generally limited to tens to hundreds of UAVs.

[0005] Graphics Processing Units (GPUs), with their powerful parallel computing capabilities, exhibit significant advantages over traditional CPUs when handling large-scale, high-density floating-point operations. With the continuous maturation and standardization of GPU programming architectures, their widespread application in fields such as artificial intelligence, aerospace, biomedicine, and military operations has fully validated the powerful potential of GPU parallel computing, providing a new technical approach for large-scale swarm simulation. Addressing the need for efficient simulation of large-scale UAV swarms, this patent provides a GPU parallel computing-driven large-scale UAV swarm simulation method. By fully utilizing the parallel advantages of GPUs, it overcomes existing simulation scale limitations while balancing cost-effectiveness and real-time simulation, enabling efficient parallel simulation of over a thousand UAVs. Summary of the Invention

[0006] This invention provides a GPU-parallel computing-driven method for simulating large-scale UAV swarms, addressing the challenge that traditional CPU-based serial computing methods cannot meet the demands of high efficiency, high real-time performance, and low cost in large-scale simulations.

[0007] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows: a GPU parallel computing-driven simulation method for large-scale UAV swarms, comprising the following steps: S1: Build a numerical simulation model of the drone; S2: Optimize the memory access mode for cluster parallel simulation; S3: Optimize the simulation task partitioning strategy based on the characteristics of UAV swarm status updates; S4: Based on S1, S2 and S3, implement cluster parallel simulation at the code level; S5: Accelerates the cluster formation algorithm by using the parallel computing strategies of S2 and S3; S6: Integrates S4 cluster parallel simulation and S5 formation control algorithm using UDP protocol.

[0008] In summary, the present invention has the following beneficial effects: Firstly, this invention optimizes the simulation task partitioning strategy based on the characteristics of large-scale UAV swarm simulation. Considering the parallel simulation characteristics of the large-scale UAV swarm involved in this invention, its state update process is mainly determined by dynamic equations and kinematic constraints. Since the state evolution of each UAV within a simulation time step is spatially independent and data interaction is minimal, it highly aligns with the data parallel paradigm. Based on this, this invention models the UAV swarm state update, optimizes load balancing and execution efficiency, and designs thread granularity and spatial mapping models, ultimately optimizing the simulation task partitioning strategy.

[0009] Secondly, a GPU parallel simulation framework is built for large-scale UAV swarm simulation. Addressing the problem of large-scale UAV swarm simulation, this invention systematically constructs a parallel simulation mechanism tailored to the execution characteristics of GPUs, starting from the methodological design level. This method uses CUDA as the foundation for parallel computing, combines the independent characteristics of the simulation objects, employs a fine-grained data parallelism strategy, and evenly distributes simulation tasks to parallel threads based on static load balancing principles and thread mapping models, achieving efficient utilization of computing resources. This invention, while ensuring simulation accuracy and system stability, breaks through the scale limitations of traditional CPU simulation methods, enabling real-time simulation of large-scale UAV swarms exceeding a thousand aircraft on conventional consumer-grade graphics cards.

[0010] Third, GPU parallel computing is used to accelerate the reconstruction of large-scale UAV swarm formation algorithms. The centralized leader-follower formation algorithm, due to its independent computational characteristics, is highly similar to large-scale UAV swarm simulation, and the same GPU parallel simulation framework can be directly used to accelerate the reconstruction of the formation control process. This significantly improves system real-time performance while supporting the control of formations of thousands of UAVs. Attached Figure Description

[0011] Figure 1 This is a flowchart of the method of the present invention. Figure 2 This is a structural diagram of the fixed-wing unmanned aerial vehicle model of Embodiment 1 of the present invention.

[0012] Figure 3 This is a data storage structure diagram for cluster parallel simulation in Embodiment 1 of the present invention.

[0013] Figure 4 This is a flowchart of the cluster parallel simulation task in Embodiment 1 of the present invention.

[0014] Figure 5 This is the pseudocode flow of the main loop control logic abstraction of the overall simulation framework in Embodiment 1 of the present invention.

[0015] Figure 6 This is a flowchart illustrating the specific calculation process of the formation control algorithm in Embodiment 1 of the present invention. Detailed Implementation

[0016] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are for illustrative purposes only and are not intended to limit the scope of the present invention.

[0017] Example 1 Reference Figure 1-6 A GPU-parallel computing-driven method for simulating large-scale drone swarms includes the following steps: S1. Construction of numerical simulation model for unmanned aerial vehicles; Using a swarm of fixed-wing unmanned aerial vehicles (UAVs) as the simulation object, a complete six-degree-of-freedom dynamics and kinematics model is constructed. The fixed-wing UAV simulation model is implemented using a commonly used dedicated modeling and simulation environment, and the model is modularly encapsulated. The constructed model has waypoint following capabilities, enabling autonomous flight based on a preset path and real-time output of UAV position and attitude information. The model structure is as follows: Figure 2 As shown. After the model is validated, simulation code is generated. The code is then optimized to adapt to the GPU parallel computing framework, thereby improving the computational performance of large-scale UAV swarm simulation.

[0018] S2, memory access optimization for cluster parallel simulation, such as Figure 3 ; Based on the access patterns and sharing requirements of different types of data, they are mapped to the appropriate GPU memory space.

[0019] S3. Optimize the simulation task partitioning strategy based on the characteristics of cluster simulation; Based on CUDA for parallel computing and considering the independence of the simulation objects, a fine-grained data parallelism strategy is adopted. According to the static load balancing principle and thread mapping model, the simulation tasks are evenly distributed to parallel threads to achieve efficient utilization of computing resources.

[0020] S4. After steps S1, S2 and S3, implement cluster parallel simulation at the code level; Based on the S3 simulation task partitioning strategy, the specific execution flow of a large-scale UAV swarm simulation task is as follows: Figure 4 As shown. To achieve parallel computing and data flow scheduling throughout the simulation process, the constructed simulation system adopts a master-slave control mode. The master unit is primarily responsible for advancing and scheduling the global simulation clock, ensuring time synchronization throughout the simulation process; while the device unit focuses on performing large-scale UAV state update calculations, utilizing CUDA kernel functions to achieve parallel solution of the UAV's physical state. The main loop control logic of the overall simulation framework can be abstracted as follows: Figure 5 The pseudocode flowchart shown systematically demonstrates the specific control mechanism for the host and devices to collaboratively advance the simulation.

[0021] S5. Accelerate the cluster formation algorithm using the parallel computing strategies of S2 and S3; The centralized leader-follower formation control algorithm is adopted, and its specific calculation flowchart is as follows: Figure 6 As shown. Since this algorithm has independent computational characteristics similar to cluster parallel simulation, a GPU simulation architecture similar to cluster parallel simulation is used to accelerate and reconstruct the formation control process, thereby ensuring the real-time performance of the system.

[0022] S6. Integrate S4 cluster parallel simulation and S5 formation control algorithm using UDP protocol.

[0023] The S4 cluster parallel simulation sends the current position information of all UAVs to the formation control algorithm in real time. The S6 algorithm calculates the expected position information of each following UAV based on the currently set formation and feeds it back to the S4 parallel simulation system. The S4 parallel simulation system updates the expected state of the UAV cluster accordingly and finally outputs the real-time position and attitude information of the cluster.

[0024] Example 2 A GPU-parallel computing-driven simulation method for large-scale UAV swarms, similar to Example 1, includes the following steps: S1. Construction of numerical simulation model for unmanned aerial vehicles; Using a quadrotor drone swarm as the simulation object, a complete six-degree-of-freedom dynamics and kinematics model is constructed. The quadrotor drone simulation model is implemented using a commonly used dedicated modeling and simulation environment, and the model is modularly encapsulated. After model validation, simulation code is automatically generated, and GPU parallel computing framework adaptation is achieved through code optimization, thereby improving the computational performance of large-scale drone swarm simulation.

[0025] S2, Memory access optimization for cluster parallel simulation; Based on the access patterns and sharing requirements of different types of data, they are mapped to the appropriate GPU memory space.

[0026] S3. Optimize the simulation task partitioning strategy based on the characteristics of cluster simulation; Based on CUDA for parallel computing and considering the independence of the simulation objects, a fine-grained data parallelism strategy is adopted. According to the static load balancing principle and thread mapping model, the simulation tasks are evenly distributed to parallel threads to achieve efficient utilization of computing resources.

[0027] S4. After steps S1, S2 and S3, implement cluster parallel simulation at the code level; To achieve parallel computing and data flow scheduling throughout the simulation process, the constructed simulation system adopts a master-slave control mode. The host computer is primarily responsible for advancing and scheduling the global simulation clock, ensuring time synchronization throughout the simulation process; while the device side focuses on performing large-scale UAV state update calculations, utilizing CUDA kernel functions to achieve parallel solutions for the UAV's physical states.

[0028] S5. Accelerate the cluster formation algorithm using the parallel computing strategies of S2 and S3; A centralized leader-follower formation control algorithm is adopted. Since this algorithm has independent computing characteristics similar to cluster parallel simulation, a GPU simulation architecture similar to cluster parallel simulation is used to accelerate and reconstruct the formation control process, thereby ensuring the real-time performance of the system.

[0029] S6. Integrate S4 cluster parallel simulation and S5 formation control algorithm using UDP protocol.

[0030] The S4 cluster parallel simulation sends the current position information of all UAVs to the formation control algorithm in real time. The S6 algorithm calculates the expected position information of each following UAV based on the currently set formation and feeds it back to the S4 parallel simulation system. The S4 parallel simulation system updates the expected state of the UAV cluster accordingly and finally outputs the real-time position and attitude information of the cluster.

[0031] in conclusion This patent proposes a GPU parallel computing-driven simulation method for large-scale UAV swarms. The effectiveness of the proposed method was verified on two different types of UAVs: fixed-wing and quadcopter.

[0032] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A heterogeneous image intelligent localization method for unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: S1: Build a numerical simulation model of the drone; S2: Optimize the memory access mode for cluster parallel simulation; S3: Optimize the simulation task partitioning strategy based on the characteristics of UAV swarm status updates; S4: Based on S1, S2 and S3, implement cluster parallel simulation at the code level; S5: Accelerates the cluster formation algorithm by using the parallel computing strategies of S2 and S3; S6: Integrates S4 cluster parallel simulation and S5 formation control algorithm using UDP protocol.

2. The GPU parallel computing-driven large-scale UAV swarm simulation method according to claim 1, characterized in that, In step S1, a numerical simulation model of the UAV is built, and the code is optimized for later parallel computing.

3. The GPU parallel computing-driven large-scale UAV swarm simulation method according to claim 1, characterized in that, In step S2, a suitable memory mapping strategy is designed based on the data characteristics and access patterns involved in the cluster parallel simulation process.

4. The GPU parallel computing-driven large-scale UAV swarm simulation method according to claim 1, characterized in that, In step S3, the simulation task partitioning strategy is optimized based on the characteristics of UAV cluster state updates, including parallel modeling of UAV cluster state updates, load balancing and execution efficiency optimization, and thread granularity and space mapping model design.

5. The GPU parallel computing-driven large-scale UAV swarm simulation method according to claim 1, characterized in that, In step S5, the cluster formation algorithm is accelerated according to the parallel simulation strategy of steps S2 and S3.

Images