基于扩展卡尔曼滤波的两轮机器人姿态估计

Attitude estimation based on extended Kalman filter for a two-wheeled robot

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波的方法进行补偿,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Levenberg-Marquardt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.采用扩展卡尔曼滤波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.将滤波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.

Aiming at the error from inertial sensors of a two - wheeled self - balanced robot, a compensating algorithm based on the extended kalman filter （EKF） was proposed. According to the inertial sensor error characteristic obatined from experiments, the error mathematical models were established by Levenberg- Marquardt nonlinear least - square iterative fit method. The error of gyro and accelerometer was calibrated by computer simulation. Using an extended kalman filter to fuse the data from the gyro and accelerometer and compensate for the sensor error, an optimal estimation for attitude was achieved. Results of simulation and field experiment demonstrated that the attitude error was suppressed validly, then an accurate and low - cost estimation of attitude was achieved, which proved that the attitude estimation method was effective and feasible.