基于改进SLAM算法的六足机器人运动轨迹规划

Motion Trajectory Planning of Hexapod Robot Based on Improved SLAM Algorithm

针对现有控制算法存在的轨迹控制偏差大和避碰能力差等问题,对六足机器人控制过程进行了研究,并提出一种基于改进SLAM算法的控制方案;对D-H模型进行了研究,分析了六足机器人的空间运动过程,采用高清摄像采集作业现场的图像同时实现高清摄像头的坐标转换,利用IMU单元提升SLAM算法模型的稳定性并进行误差与参数的标定,基于局部二值化模型提取现场图像的特征;在图像特征集训练中采用了CNN网络模型提升SLAM算法模型的数据训练能力,并根据与现场环境交互后最大折扣奖励值,提升机器人步态稳定性和局部区域的避碰效果;实验结果显示:改进SLAM算法实现了机器人全局范围内的轨迹路径寻优,路径耗时仅为35.4 s,在10次避碰测试中与障碍发生碰撞的次数为1次,优于其他避碰控制算法。

Existing control algorithms have the shortages of large trajectory control deviation and poor collision avoidance ability,this paper researches the control process of hexapod robot,and proposes a control scheme based on simultaneous localization and mapping(SLAM) algorithm.The Denavit-Hartenberg(D-H) model is studied,and the spatial motion process of the hexapod robot is analyzed.a high definition(HD) camera is used to acquire the images of the job site and realize the coordinate conversion of the HD camera at the same time.an inertial measurement unit(IMU) unit is used to improve the stability of the SLAM algorithm model and carry out the calibration of errors and parameters,the local binarization model is based on extracting the features from the field images.In the image feature set training,the convolutional neural network(CNN) network model is used to improve the data training ability of the SLAM algorithm model,and according to the maximum discount reward value after interacting with the field environment,the gait stability of the robot and collision avoidance effect of the local area are improved.Experimental results show that the improved SLAM algorithm can achieve a trajectory path optimization in global scope,with a path time of only 35.4 s,and only one of ten collision avoidance tests occurs the improved SLAM algorithm is superior to other collision control algorithms.