深度神经网络下机器人滑移量检测与路径规划  

Robot Slip Detection and Path Planning Based on Deep Neural Network

作　　者：孙滨 SUN Bin(College of Information Engineering,Zhengzhou University of Industrial Technology,He’nan Zhengzhou 451150,China;College of Computer and Information Engineering,He’nan Normal University,He’nan Xinxiang 453000,China)

机构地区：[1]郑州工业应用技术学院信息工程学院,河南郑州451150 [2]河南师范大学计算机与信息工程学院,河南新乡453000

出　　处：《机械设计与制造》2023年第10期134-138,共5页Machinery Design & Manufacture

基　　金：2022年度河南省重点研发与推广专项(科技攻关):社交网络中虚假节点分析与识别技术研究(222102210159);2020年度河南省重点研发与推广专项(科技攻关):基于物联网与大数据的多系统融合的煤矿安全生产监测监控研究(202102210361);2019年河南省教育厅高校青年骨干教师培养资助项目:集成神经网络下的骨龄评价和身高预测研究(2019GGJS279)。

摘　　要：为保证机器人行驶路径最短时避免发生滑移,提出基于深度神经网络的机器人移量检测与路径规划方法。通过分析机器人运动学情况,结合机器人滑移量会随轨迹、地面平整度及障碍物影响范围发生变化,构建运行轨迹方程组,求得会影响滑移量的因变量值;建立位姿计算坐标系,设置机器人滑轮初始位姿坐标,计算发生滑移后末端位姿,更新后代入滑移公式求得滑移量;以障碍物避障和路径最短为规划标准,采用深度神经网络算法激活避障及路径最短的约束函数,输出规划结果。通过对比实验结果证明所提方法的有效性,在所提方法控制下滑移量检测结果与实测值表达一致,规划路径在保障高效避障的同时确保是最短行驶距离,优于对比方法,可靠性强。In order to avoid slippage when the robot travel path is the shortest,a robot movement detection and path planning method based on deep neural network is proposed.By analyzing the kinematics of the robot.Combined with the fact that the slid-ing amount of the robot will change with the trajectory,ground flatness and the influence range of obstacles,construct the run-ning trajectory equation set,and obtain the dependent variable value that will affect the slip;establish a coordinate system for pose calculation,and set the initial position of the robot pulley Attitude coordinates,calculate the terminal pose after slip occurs,update the descendants and enter the slip formula to obtain the slip amount;take obstacle avoidance and the shortest path as the planning criteria,use the deep neural network algorithm to activate the constraint function of obstacle avoidance and the shortest path,output the planning result.The effectiveness of the proposed method is proved by the comparative experimental results.The detection results of the slip amount controlled by the proposed method are consistent with the measured values.The planned path ensures the shortest driving distance while ensuring efficient obstacle avoidance,which is superior to the comparative method and has strong reliability.

关 键 词：深度神经网络 运行轨迹 机器人滑移量 初始位姿 约束函数 路径规划 

分 类 号：TH16[机械工程—机械制造及自动化] TP388[自动化与计算机技术—计算机系统结构]