Resource Scheduling For Unmanned Aerial Vehicle Control System Based On Cyber-Physical Systems

Cyber-Physical Systems(CPS)is a complex system of information processing,network transmission and physical control,in which,with the integrity of information technology,communication technology and control technology,information processing and physical processes are deeply intertwined on spatial and temporal scales.CPS is regarded as the core technology to promote the new-round industrial revolution.CPS not only relays and shares the information,but also fulfills corresponding control tasks through dialogue and cooperation between the system components.Unmanned Aerial Vehicle(UAV)is one of the key areas of CPS application.Applying the principles and approaches of CPS to UAV promotes the deeper integration of a variety of heterogeneous resources(e.g.ground control system,the UAV body,sensors,smart devices).It also achieves accurate,efficient,real-time and intelligent integrated control of the UAV,and enhances s resource utilization,dynamic reconfigurability and accessibility of UAV control system thanks to accurate sensing technology,reliable communication technology and powerful computational intelligence.Due to the uncertainty of tasks,network delay and packet loss occurring in the UAV control system based on Cyber-Physical Systems(UAVCPS),it is difficult for the UAV to perform well.First of all,the failure of the task because of the breakdown or inaccessibility of UAV component(s)shows that an alternative architecture should be developed to ensure the fulfillment of tasks.Secondly,the uncertainty caused by the constraints of computational resources and network resources requires sounder control and resource utilization for proper resource scheduling and task fulfilment.Thirdly,since resources are limited,special attention should be paid to local system overload in extreme cases.In addition,with abundant parallel computing resources and network resources,CPS can divide UAV's intelligent tasks into multiple subtasks.One critical issue is how to best organize these subtasks and manage the task in a coordinated way.Therefore,flexible and efficient resource scheduling technology has become one of the key technologies in CPS research.This paper studies the scheduling of computational resources and network resources of CPS and covers the following five aspects:(1)This paper studies the structure of CPS,and proposes a CPS architecture based on bus.Service-oriented architecture(SOA)backs effective resource scheduling by making the components more reliable,shielding the heterogeneity of resources,encapsulating resources,and coordinating interaction interface.According to the idea of SOA,CPS can be divided into service layer,middleware layer,network layer and physical layer.A system resourcescheduling framework is designed with aspect to task priority scheduling,cycle optimization and task allocation mechanism,which provides the framework for the further study of CPS.(2)Computational resource scheduling for UAVCPS is studied.To cope with irregular task arrival as well as uncertain task execution time and communication time,ISMHEFT(an improved HEFT algorithm for stochastic multi-DAG schedule)algorithm is proposed.The algorithm describes the task through directed acyclic graph(DAG),and takes into account task entry sequence and importance in handling irregular arrival of multiple DAGs so that it can calculate the priority of all subtasks within a unified DAG structure.Unlike HEFT scheduling algorithm,ISMHEFT algorithm represents time with the sum of standard variance and average value so as to figure out more accurate rankup speed and execution time,receive sounder task priority and task allocation,and promote efficiency.(3)A Two-level Fuzzy Feedback Improved-EDF Scheduling(TLFFIE)is designed to accommodate uncertain complex dynamic environment with unpredictable noise.This scheduling first schedules computational resources between bus and nodes by adapting to dynamically changing task priority,and monitoring the priority of each node in consideration of response time and packet loss.Response time and packet loss are used to regulate duty cycle in the hope of reallocating the resources of UAVCPS.The Matlab simulation and prototype system results show that this scheduling can reduce control system packet loss,which in turn better utilizes and improves the system's control performance.(4)To mitigate real-time reduction as well as task failure caused by overload and out-of-order delivery in CPS system,an effective task allocation mechanism is studied.A request division load balancing model is built and we propose a virtual node transition strategy based on PI feedback.This strategy exploits incremental PI control,dynamically monitors the bus node's assignment weight according to real-time feedback of node performance,and adjusts virtual nodes locally via a virtual node transition algorithm to maintain the stability of Hash Space.This approach can realize dynamic load balancing,reduce resource consumption of the server,lessen communication overhead caused by remote data access,and enhance the real-time performance of control systems.(5)An UAV plant protection system is built.Experiments on task priority,cycle optimization,resources utilization,task finish time and packet loss show that the design is reliable and that the scheduling is effective.In conclusion,this paper focuses on the architecture design and resource scheduling of UAVCPS,analyses task priority scheduling,cycle optimization and tasks allocation,and proposes corresponding strategies and scheduling algorithms.The research results bear boththeoretical and practical significance to improve the efficiency of real-time performance and promote the resource management technology of CPS.