无人自行车LPV模型在线识别方法研究

Online identification of the LPV model of an unmanned bicycle

集总参数是无人自行车变参数线性(variable parameter linear,LPV)力学模型的核心要素之一,针对一种无人自行车关键集总参数难以直接准确获得的问题,提出一种在线估算其中时变运动参数和物理参数的方法。对于模型中时变车速的估算,主要通过车轮转速传感器以及车架加速度传感器的量测值,结合卡尔曼滤波技术来实现。对于模型中的物理结构参数,主要根据自行车的LPV模型构造逆向识别算法,以IMU,编码器,电流传感器等传感器测量的系统状态为输入,待识别的物理结构参数为输出,通过最小二乘法求解得到。物理样机的实验结果表明:利用模型识别结果计算得到的力矩与传感器实际测量的驱动电机输出力矩基本吻合,验证所提方法的可行性与正确性。

Lumped parameters are one of the key elements of variable parameter linear(LPV)mechanical model of unmanned bicycle.Aiming at the problem that it is hard to directly obtain the key parameters of the LPV model of an unmanned bicycle,an online identification method to estimate the time-varying motion parameters and physical parameters is proposed.The measurement value from the wheel speed sensor and acceleration sensor are lumped with Karlman filter technique to calculate the time-varying velocity of the LPV model.The identification model to get the physical structure parameters is constructed based on LPV model of the bicycle by referring to the system state from the IMU,encoder,current sensor and other sensors.The model is ultimately solved by mean of the least square method.The results of the physical prototype experiments demonstrate that the calculated torque obtained from the model identification parameters is basically consistent with the actual sensor measured torque,which verifies the feasibility and correctness of the proposed identification method.