%0 Journal Article
%T 基于神经网络识别的垃圾分类平衡机器人<br>Garbage Sorting Balancing Bobot Based on Neural Network Recognition
%A 高昂
%A 贺文嘉
%A 董文硕
%J Artificial Intelligence and Robotics Research
%P 351-360
%@ 2326-3423
%D 2023
%I Hans Publishing
%R 10.12677/AIRR.2023.124038
%X 本机器人具有广泛应用前景，解决日益严重的环境污染和资源浪费问题，能在复杂环境中解决人工操作难以实现的自动化，精确化，智能化问题。该平衡机器人具有自主定位、自主移动、自主避障、通过垃圾、垃圾抓取与搬运、路径规划等功能。利用外界算力平台，通过神经网络构建垃圾模型，计算特征点，打包传入嵌入式AI计算设备(TX2)，通过TX2解算相应的模型，更精确的视觉识别物品模型。同时设计相对应程序使得机器人每个任务都安全可行。并且机器人可在狭窄环境下实现换向。加入视觉识别相机实现垃圾识别、以及放置到指定位置。其次，该机器人加入机械夹爪，可对大体积垃圾进行抓取放置。该机器人的底盘使用STM32作为主控芯片，可以迅速反映，快速获取电机，舵机等反馈数据。且应用PID算法进行控制，根据目标轨迹与当前误差，动态地调整控制参数，机器人四周都装有距离传感器可自动避障，且下方装配了定位轮实时监测机器人位置，若发生问题可以及时做出反应。<br />
The robot has a wide range of application prospects, to solve the increasingly serious environmental pollution and resource waste problems, can solve in the complex environment manual operation difficult to achieve automation, precision, intelligent problems. The balance robot has the functions of autonomous positioning, autonomous movement, autonomous obstacle avoidance, garbage, garbage grab and handling, path planning and so on. Using the external computing power platform, the garbage model is constructed through the neural network, the feature points are calculated, and the garbage model is packaged into the embedded AI computing device (TX2). The corresponding model is solved through TX2, and the object model is more accurate visual recognition. At the same time, the corresponding program is designed to make each task of the robot safe and feasible. And the robot can change direction in a narrow environment. A visual recognition camera is added to recognize garbage and place it in a designated position. Secondly, the robot adds a mechanial claw, which can grasp and place large volumes of garbage. The chassis of the robot uses STM32 as the main control chip, which can quickly reflect and obtain feedback data such as motor and steering gear. Moreover, PID algorithm is applied for control, and thr control parameters are dynamically adjusted according to the target trajectory and current error, Distance sensors are installed around the robot to automatically avoid obstacles, and the positioning wheel is equipped below to monitor the robott’s position in real time, so that it can respond in time if problems occur.
%K 神经网络，平衡机器人，PID，TX2，STM32，视觉识别<br>Neural Network
%K Balancing Robot
%K PID
%K TX2
%K STM32
%K Visual Identificaton
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=76020