基于扩展卡尔曼滤波两轮机器人姿态估计.pdfVIP

39 12 V ol39 N o12 2 0 0 7 12 JOURNA L OF HAR BIN INST ITUTE OF TECHNOLOGY D ec. 2007 王晓宇, 闫继宏, 秦 勇, 赵 杰 , 15000 1, Em a il: wxyh it197745@ yahoo. com. cn) : 针对两轮自平衡机器人惯性传感器存在误差的问题, 提出基于扩展卡尔曼滤 的方法进行补偿, 从 而实现机器人姿态的最优估计. 利用实验获得的惯性传感器误差特性, 采用 L evenbe rg - M arqua rdt非线性最 小二乘迭代法拟合数据, 从而建立机器人导航用惯性传感器陀螺仪和加速度计误差的数学模型, 并对误差进 行标定. 采用扩展卡尔曼滤 将传感器的数据进行融合并对误差进行补偿, 得到机器人姿态的最优估计. 将 滤 后的模型应用到两轮自平衡机器人系统, 实验结果表明改进后的系统误差得到了有效的抑制, 从而验证 了采用低成本的惯性传感器进行机器人的姿态估计是有效可行的. : 扩展卡尔曼滤 ; 误差建模; 姿态估计; 数据融合; 随机漂移误差 : TP242 : A : 0 367 - 6234 2007) 12- 1920 - 05 A ttitude estim ation based on extended K alm an filter for a tw owheeled robot WANG X iaoyu, YAN J ihong, Q IN Y ong, ZHAO Jie R obotics Institute o f H IT, H arb in Institu te o f T echno logy, H a rbin 15000 1, Ch ina, E - m a il: w xyh it197745@ yahoo. com. cn) Abs tract: A mi ing at the error from inertia l sensors of a wt o- w hee led se lf- balanced robot, a com pensating algorithm based on the extended ka lm an f ilter EKF ) w as proposed. A ccording to the inertial sensor error character istic obatined from expermi ents, the error m athem atica lm ode lsw ere established by L evenberg- M ar quardt nonlinear least- square iterative fit m ethod. T he erro r of gyro and acce lerom eter w as calibra ted by com puter smi u lation. U sing an ex tended kalm an f ilter to fuse the da ta from the gyro and accelerom eter and com pensate for the senso r erro r, an opt mi al estmi ation for attitude was achieved. Resu lts o f smi ulat ion and fie ld expermi ent dem onstrated that the att itude erro rw as suppressed va lid ly, then an accura te and low - cost estmi a t ion of attitude w as achieved, wh ich proved tha t the attitude e