Research On Meal Delivery Robot Design

Affected by the COVID-19,a large number of restaurants closed down and the labor service industry was devastating blow.In order to improve the dining environment and reduce the risk of epidemic spread,this article designed a meal delivery robot based on ROS(Robot Operating System)that integrates image recognition,robotic arm control,and AGV(Automated Guided Vehicles)navigation and obstacle avoidance.Compared with the traditional design scheme,the ROS-based design is more generalized,and the algorithm can be transplanted into the simulation software equipped with it for reliability testing,reducing repeated tests and preventing imperfect algorithms from damaging the robot hardware.The research on all algorithms involved in robotic arms and AGVs also closely follows the development of the robotics field.Therefore,the design and research of meal delivery robots has practical significance and practical value.The specific research contents and innovations of this article are as follows:First,design and complete the overall scheme of robot software and hardware according to actual needs.In terms of hardware,the overall connection diagram of each device is given,the reason for component selection is explained,and the functional parameters of each component are introduced in detail.In terms of software,the overall design plan of the software platform is given,and the four modules of recognition,capture,map building,and path planning are discussed in detail.Introduced ROS and the three simulation platforms used in this article,and explained the establishment of Solidworks,URDF(Unified Robot Description Format,),and Gazebo models used in the robot simulation platform.Secondly,three kinds of image preprocessing work were carried out: color gamut conversion,Gaussian blur,and opening and closing operations,which improved the image quality before the incoming recognition algorithm.According to the requirements of the design plan,the Canny edge detection algorithm has been improved in four aspects:adaptive threshold setting,histogram equalization,norm reset,and Gaussian kernel disassembly optimization complexity,which improves the recognition accuracy and reduces the algorithm operation hours.Thirdly,the camera calibration is performed to reduce the system error of the equipment,and the scale coefficient of the conversion between the world coordinate system and the image coordinate system is obtained.The contour recognition algorithm has also been improved and innovated in four aspects,including time and distance threshold setting,precise sub-pixel coordinates and non-maximum suppression.And then according to the image determine the deflection angle of the object,the position of the centroid,and the radius of the maximum circumscribed circle according to the identified edge.Use inverse kinematics to calculate the steering angle of the robotic arm and the opening angle of the mechanical claws,and convert them into PWM(Pulse Width Modulation)pulse width,unify the motion time of each servo,and use PID(Proportion,Integral,Derivative)to control the robotic arm to complete the grasp.Then,the front-end optimization of SLAM(Simultaneous Localization And Mapping)was carried out,including two parts: odometer calibration and lidar motion distortion removal.Use the processed sensor data for SLAM-Gmapping mapping to ensure the quality of the map.On the built raster map,A*(A Star)is used for optimal global path planning,and DWA(Dynamic Window Algorithm)is equipped to avoid local obstacles,and quickly re-plan the path to achieve accurate navigation.Finally,build the robot entity to complete the joint debugging of the meal delivery robot's software and hardware.Design three targeted experiments to verify the functions of its image recognition,robotic arm and AGV chassis.Analyzed and solved the problems in the experiment,the improved meal delivery robot performed well in terms of accuracy and stability.Prove that the design of robot software and hardware is feasible and satisfy service requirements.