摘要
六足机器人采用不同的步态参数针对不同的非结构环境可提高运行效率、增强平稳性。为实现该目标,预先完成三个步骤:第一,机载摄像头用来捕获环境信息;第二,基于环境样本库,使用卷积神经网络训练环境识别模型,因其在视觉问题上的泛化能力,卷积神经网络可以自动提取图像特征并降低建模工作量;第三,调整六足机器人的步态参数以获得某一给定环境下的最优步态。在此基础上,六足机器人根据环境识别结果选择相应的最佳步态,并实现环境自适应。最后进行对比实验,结果表明,应用环境自适应方法的六足机器人可以在复杂环境中耗能更低,并获得更高的速度和更优良的平稳性。
Hexapod robots use different appropriate gait parameters for different unstructured environments can improve walking speed and enhance stability. In order to realize this goal, three steps of works are accomplished in advance. First, a camera on - board is used to capture the environment images. Second, the environment recognition model is trained based on the environmental sample libraries using convolutional neural network, because convolutional neural network can extract im- age features automatically and reduce the workload of modeling. Third, the gait parameters of the hexapod robot are tuned to obtain the optimal gait for a given environment. On this basis, the hexapod robot can select the corresponding optimal gait ac- cording to the environment recognition result, and realize the environment self - adaptation. Finally, the comparison experi- ments are conducted, the results show that the hexapod robot with the environment self - adaptation method can obtain higher speed and better stability with less energy consumption in a mixed environment.
作者
傅汇乔
留沧海
唐开强
江浩
FU Huiqiao;LIU Canghai;TANG Kaiqiang;JIANG Hao
出处
《现代机械》
2018年第2期1-6,共6页
Modern Machinery
基金
四川省科技计划重点项目(18MZGC0203)
西南科技大学生创新基金项目(cx17-093)
关键词
六足机器人
卷积神经网络
环境自适应
步态
hexapod robot, convolutional neural network, environment self - adaptation, gait
作者简介
傅汇乔（1996-），男，本科生，研究方向：智能机器人。;留沧海（1966-），男，教授，博士，研究方向：机器人技术等。
