Design And Implementation Of ROS-based Cloud Robot System

Traditional robots cannot upgrade and iterate storage and computing power quickly due to long development cycles and high hardware costs.To address this problem,this thesis designs and implements a ROS-based cloud robot system.Our research covers the following techniques.1)Design and implementation of local robot.The robot is designed in three layers including the upper layer,the middle layer and the lower layer.The upper layer puts some sensors.The middle layer controls robot and carries ROS.The lower layer is responsible for driving the robot motion.The kinematic model equations of the robot are given in con-junction with the structure design of the robot.A local robot is implemented by selecting the appropriate hardware device in combination with the robot operating system ROS.2)Presenting communication of cloud robot system.We use the client-server archi-tecture as the communication architecture of cloud robot system.We use Web Socket as the transport protocol for cloud robot due to its less control overhead,stronger real-time performance,ability to maintain connection status,and better compression.On the mes-sage specification of cloud robot system,we encapsulate the message with JSON data type and can specify the name of the algorithm requested in the encapsulated data.3)Testing some algorithms using cloud robot.The cloud server can encapsulate various algorithms for local robots to call to extend their functionality.We implement a ROS-based object tracking algorithm by combining YOLOv5 and Deep SORT,which can directly receive video stream data from ROS for processing.4)We design the software and hardware system.We implement the local robot and a ROS-based cloud robot and implement the object tracking algorithm and deploy the system in the cloud server.