基于极限学习机与模糊积分融合的机器人地面分类

发布时间：2018-04-12 00:17

本文选题：移动机器人 + 地面分类　；参考：《哈尔滨工程大学学报》2017年04期

【摘要】：为了提高移动机器人地面分类的准确率,采用奇异值分解和功率谱密度估计两种方法对振动信号进行特征提取。针对极限学习机的隐层节点冗余问题,给出改进的算法,并采用改进的极限学习机对地面分类。针对模糊积分参数耗时和积分函数不确定的问题,给出改进的方法,并基于2种特征采用改进的模糊积分对2个改进的极限学习机进行融合。在四轮移动机器人左前轮轮臂上安装三向加速度计和z向传声器,使之在沙、碎石、草、土、沥青地面上分别以5种速度行驶,采集车轮与地面相互作用的加速度和声压信号。根据改进的极限学习机和模糊积分融合算法,分别对每种速度下的5种地面进行分类,分类平均准确率为95.22%。实验验证了算法的有效性。
[Abstract]:In order to improve the accuracy of mobile robot ground classification, singular value decomposition (SVD) and power spectral density estimation (PSD) are used to extract the feature of vibration signal.Aiming at the hidden layer node redundancy of LLM, an improved algorithm is presented, and the improved LLM is used to classify the ground.In order to solve the problem of time consuming of fuzzy integral parameters and uncertainty of integral function, an improved method is presented, and two improved extreme learning machines are fused by improved fuzzy integral based on two characteristics.A three-direction accelerometer and a Z-direction microphone are installed on the left front wheel arm of a four-wheel mobile robot to drive on sand, gravel, grass, soil and asphalt at five speeds, respectively. The acceleration and sound pressure signals of the interaction between the wheel and the ground are collected.According to the improved extreme learning machine and fuzzy integration fusion algorithm, five kinds of ground are classified at each speed, and the average accuracy of classification is 95.22.The effectiveness of the algorithm is verified by experiments.
【作者单位】：哈尔滨工程大学机电工程学院;
【基金】：国家自然科学基金项目(60775060) 高等学校博士学科点专项科研基金项目(20122304110014)
【分类号】：TP242

【相似文献】