Research Of Service Platform For Cloud Robotics Based On Container Technology

The rapid development of robotics technology has brought considerable progress to the industrial,service,national defense industries and created great social value.As the next important field of robot application,the development and research of intelligent robots have risen to the national strategy level,which has aroused extensive concern all over the world.In technology,there are still some problems to overcome urgently.For example,the single function of robots can't adapt to multi-purpose scenarios,and there is a high cost of hardware with strong computing power and so on.Taking advantages of cloud computing in resources and data sharing,this thesis combines robotics with cloud computing,studying on the service platform for cloud robotics based on container technology,to build a flexible robot system framework in the form of cloud-based functional services,and explore the way to realize cloud robotics,which has strong scientific value and practical significance.To meet the demand for transferring robot computing tasks to the cloud,combining the technical advantages of container technology,this thesis designs a service platform architecture for cloud robotics.In this architecture,the operating platform is supported by the cooperation of communication middleware,the system management module,and system resource module normally and safely.The functional robot services are encapsulated in flexible Docker containers,which changes the resources using way for complex computing tasks.To ensure that the robots functional services can have a better performance of computing in high or low loads,and improve the resource utilization efficiency of service platform,there is a flexible container capacity scheduling algorithm in the resource management module.Considering the actual operation situation of different robotic services,this algorithm calculates the average load according to the CPU usage,mem ory usage and network usage of containers,and performs capacity elastic scheduling operation by preset upper and lower thresholds.During the period of resource flexibility expansion,users will not perceive the dynamic scheduling of functional service capacity,which ensures the stability of the platform.Considering that robot visual SLAM.is a-computationally intensive task,this thesis combines the construction idea of micro-service,split the task into cloud micro-services that can be implemented in the service platform,forming modular functional components such as tracking,local mapping,closed-loop detection,point cloud building so as to realize cloud unloading of complex data operations.The full use of the platform performance makes the application of sophisticated and advanced algorithms in low-cost robots possible.On this basis,through the functional service test experiments deployed in the cloud,it is verified that the service platform can effectively unload the tasks of the robot to the cloud,and reduce the hardware performance requirements of the robot.At the same time,modular functional service forms can expand the functions and use scenarios of robots,break the limits of traditional robots dealing with a single scene,and help to enhance the intelligence of robots.