Design of virtual reality fusion UAV cluster simulation system
[Objective]The validation of unmanned cluster algorithms and their functions and collaborative effects for various scenarios has become one of the main bottlenecks hindering the transition of unmanned cluster technology from theoretical research to engineering applications.First,the development of unmanned cluster algorithms is rapid,but most of them are at the theoretical and simulation levels,and there is a significant gap between them and practical applications.Second,unmanned cluster collaboration algorithms are often complex and are difficult to verify directly in physical systems.They are supported by many conditions unrelated to the algorithm itself,such as system stability,unmanned platform control,navigation and positioning,and limited computing resources.Finally,from the algorithm level to engineering applications,processes such as algorithm design,algorithm simulation,semiphysical simulation,physical system design,and physical system testing are usually involved.Simulation and semiphysical simulation are essential.However,current digital environments for unmanned cluster verification are mostly pure simulation environments,with effective semiphysical simulation environments as an important transitional step being rarely addressed.Therefore,constructing a semiphysical verification system that can verify unmanned cluster collaborative algorithms,as well as elements and scene design for practical applications,is of great theoretical and practical significance in verifying the application effects of unmanned clusters in many special scenarios.[Methods]This study constructed a computer-based digital simulation system and a semiphysical simulation cluster node based on an embedded onboard system:the virtual real fusion unmanned aerial vehicle(UAV)cluster simulation system.This system is based on algorithm simulation and semiphysical operation and manages tasks to be verified as a whole.It can achieve verification through embedded board interconnection for semiphysical verification and can be extended to add virtual cluster nodes for large-scale distributed virtual real fusion verification.Different modes of cluster nodes can achieve data exchange and task collaboration,with a core function of supporting scalable system architecture,distributed data exchange mechanism,and modular implementation of main functions.In the overall system architecture,the core algorithm logic is decomposed into two parts:centralized and distributed operators.The task logic and deployment methods of the two are further divided.For distributed nodes,multithreaded or wireless communication methods are used to achieve distributed data interaction according to their deployment methods.Meanwhile,the core algorithms of distributed nodes are modularized and encapsulated so that the algorithms have good reusability on different nodes.[Results]Taking the double-observation policy learning algorithm(DOPLA)as an example to verify the simulation effect of the system,this algorithm can make decisions without multiple iterations and optimize allocation through real-time decision-making.Experiments were conducted in task scenarios with different numbers of targets.To better demonstrate the distributed collaborative target coverage of large-scale drones under the DOPLA algorithm,100 targets were randomly selected for coverage.In this system,20 drones were used to perform collaborative target coverage tasks.[Conclusions]This system can serve as a simulation verification platform for the development of collaborative algorithms in drone clusters and a semiphysical verification environment for task elements and scenario design.The demonstration and verification of multimodal and virtual real fusion demonstrate the excellent scalability of the designed verification system in terms of architecture.This provides a strong platform support for comprehensive experiments,teaching,scientific research,training,and innovative activities of UAV cluster systems.
万方数据
维普
