足式机器人腿部倾角传感器信号处理研究

足式机器人在自主行走时,一般通过倾角传感器来测量腿部转动角度计算足端位置,然而目前足式机器人腿部倾角传感器测量时易受噪声干扰、温度等因素的影响,导致测量精度低,足端位置估计不准确.针对以上问题,提出新的倾角传感器信号处理方法,首先利用卡尔曼滤波方法对倾角传感器输出信号进行滤波预处理,然后把滤波信号和倾角传感器输出温度值作为建立的双输入单输出RBF神经网络模型的输入变量,采用蚁群聚类算法的并行寻优特征和自适应调整挥发系数方法来确定RBF神经网络基函数位置.实验结果表明,提出的算法能很好地滤除倾角传感器信号中的噪声,实现了倾角信号的温度补偿,测量误差能够控制在0.75％以内,具有实际运用价值.

Study on signal processing of leg tilt sensor of legged robot

When walking independently,a legged robot usually measures the rotation angle of the legs and calculates the position of the foot by means of the tilt sensor,however,the leg angle sensor is easily affected by the influence of various factors,such as noise,temperature,leading to low accuracy of measurement and the foot end position estima-tion. To solve the above problems,put forward a new angle sensor signal processing method,firstly by using Kalman filter for Angle sensor output signal filtering preprocessing,and then output the filtered signal and temperature value of the tilt sensor as a double input single output RBF neural network input variables of the model,using the ant clus-tering algorithm in parallel optimization characteristics and adaptive adjustment of volatile coefficient method to lo-cate the RBF neural network basis function. The experimental results show that the proposed algorithm can filter the noise of the tilt sensor signal well,and achieve the temperature compensation of the inclination signal. The measure-ment error can be controlled within 0.75％,which has practical application value.