The Design Of Remote Real-Time Control System Of Humanoid Robot Based On Cloud Platform

The humanoid robot is the most ideal robot that is able to interact with people.It can get new knowledge constantly by interacting with the environment and complete a variety of tasks in a way that its designers can’t imagine. It can adapt to dynamic and unstructured environment.Because of the finite capacity and the ability of processing of Airborne Equipment, the ability to perform complex tasks of the robot is limited. With the rapid development of wireless network technology, the remote control of robot is possible. This article is based on the cloud platform and designs a massive processing capability of multi robot remote control system, through the airborne camera we can capture real-time images and build the background processing system in the cloud platform, the background processing system make judgments to images according to set rules, and return control commands to the robot to complete the corresponding action.The specific contents are as follows:1. Based on cloud platform, this article designs the overall architecture of the remote control system of humanoid robot, including information acquisition module, information processing module, communication module, processing module (cloud platform), cluster control module. With the information input, the work flow of systems:the humanoid robot Darwin-OP collects environmental information (information acquisition module)by airborne camera, then through the information pretreatment module, the compressed image is transmitted to cloud platform(umaster-worker architecture), information gets to the master node of cloud platform, master node makes decision according to the load condition of each worker node, the task will be passed to the worker node is reasonable, after that, the result gives the humanoid robot control module feedback by master node, which makes the corresponding control action.2. Specific design and implementation of each module. This article selects specific implementation method of each module through the investigation. The information acquisition module is the airborne camera of Darwin-OP humanoid robot. Darwin-OP is a humanoid robot that uses video4Linux2V4L2to do image acquisition with USB camera. Image preprocessing module adopted information compression algorithm that is JPEG image compression algorithm. The communication of humanoid robot and cloud platform communication module is based on TCP/IP network communication mode; the internal node communication of cloud platform uses ZeroMQ Communication Middleware; The cloud platform is IaaS layer service based on cloud computing, the virtual machine of cloud platform is used to build master-worker infrastructure, the number of virtual machines changes according to the needs. The control module is located in humanoid robot onboard PC, using Darwin-OP action-editor to write their own platform of humanoid robot control instruction.3. Prove the effective design of experiment. The experiment was divided into five parts:image compression, image recognition experiment, simulation load test experiment, The robot motion editing experiment, the whole experimental system. Through the experiment, results are in line with expectations, showing that the design of system is effective.