Design And Control System Of Small Humanoid Soccer Robot

Humanoid robot is a cross-disciplinary product of mechanical engineering,electronic engineering,computer science and automatic control technology.It is an important research direction in robotics.This article designes a robot physical prototype according to the rules of the2019 Robo Cup small humanoid robot competition,and completes its control system,so that the robot has certain action and perception capabilities to complete the soccer game.In this paper,the mechanical structure and hardware of the robot are firstly designed and selected.The robot contains 18 degrees of freedom and uses a bus-type digital steering gear as an actuator.It senses external information through a camera and an inertial measurement unit.The main control devices are divided into two layers,the upper layer for the main computation,and the lower layer for the hardware control.The circuit is divided into power supply,communication and interaction and other parts.The inverse kinematics solution of the robot uses an analytical method to establish the foundation for robot walking and motion control.This paper implements a micro robot control system based on multi-process and TCP protocol.It is divided into different modules and has a low coupling between modules,and different modules can communicate by broadcasting message or remote procedure call.The control system is divided into main controller,hardware controller,debugging system,auxiliary system,motion control system and vision system according to their functions.The main controller includes kinematics,upper and lower communication systems,decision system based on finite state machine,etc.The hardware controller is used for the control of servos and other equipment,and the reading of sensor data.The debugging system is based on Qt GUI framework,and has the functions of image debugging,parameter calibration,robot remote control and monitoring.In the auxiliary system,the cross compilation technology based on CMake and the automation script based on Python are proposed.The motion control system is used to plan the walking and special actions of robot.In this paper,a robot simulation model based on Webots is built for the testing of motion control.In the planning of walking,the center of mass trajectory is planned using the inverted pendulum model and the zero moment theory,and the swing leg trajectory is planned using polynomials.For the special actions of the robot,the paper proposes a method based on key frame pose interpolation in the action,and designs a tool.In the vision system,the preprocessing stage completes the collection of different image data,the conversion of image color space and storage format,the image size scaling based on bilinear interpolation,and the image distortion correction,and uses CUDA to accelerate image preprocessing.In the target extraction stage,the field boundary extraction is realized based on statistics and RANSAC fitting method,and goal and football recognition are completed based on the convolutional neural network.