多传感器信息融合在四足机器人避障中的应用  

Application of multi-sensor information fusion in obstacle avoidance of quadruped robot

作　　者：张天[1] 杨晨曦[2] 朱颖[1] 宋明辉[3] 张地[4] 

机构地区：[1]北京理工大学光电学院,北京100081 [2]北京理工大学信息与电子学院,北京100081 [3]北京理工大学机械与车辆学院,北京100081 [4]郑州大学国际学院,河南郑州450001

出　　处：《传感器与微系统》2015年第5期150-153,共4页Transducer and Microsystem Technologies

基　　金：总装预研计划资助项目(104060103)

摘　　要：为提高仿生四足机器人在复杂、动态环境下对障碍物位置信息的感知能力,针对机器人在结构化路面上以Walk步态行走的情况,对双目视觉传感器和超声测距传感器获取的障碍物距离信息进行融合研究。首先,对两种传感器获取的障碍物距离信息进行卡尔曼滤波,降低环境中杂波的影响,然后,根据STF融合算法,利用滤波后得到的两组状态向量的估计值和协方差矩阵进行融合处理。仿真结果表明:滤波后的距离信息的估计值曲线很好地跟踪了真实值曲线,说明卡尔曼滤波算法发挥了良好的滤波作用;与融合前相比,融合后的距离信息估计值的方差明显减小,说明融合后的障碍物位置信息更加准确,满足了仿生四足机器人在复杂、动态环境下作业和行进的精度要求。In order to improve ability of bionic quadruped robot to percept location information of obstacles in complex and dynamic environment,information fusion between binocular sensor and ultrasonic ranging sensor is researched under condition that the robot moves in walking gait on structured road surface. Firstly,distance information of obstacles from these two sensors is processed by Kalman filtering algorithm,which is helpful to alleviate interference caused by noise. Then,use two groups of state vector estimation value and covariance matrix obtained by filtering,fusion processing can be carried out by STF fusion algorithm. Simulation results show that curve of estimation value of distance information after filtering track curve of real value well,which shows that Kalman filtering algorithm plays an effective role in inhibiting noise; after fusion,the variance of the estimation value is distinctly smaller,which illuminates that the information of the obstacles becomes more accurate. Finally,the need for precision is met when the robot works in complex and dynamic environment.

关 键 词：双目视觉传感器 超声测距传感器 卡尔曼滤波算法 匀速模型 STF算法 机器人避障 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]