Research On External Force Estimation And Direct Teaching Of Robots Without Force Sensors

With the proposal of " Human-machine integration ",a major challenge facing emerging robotics applications is the unstructured environment.In an unstructured environment,robots are required to have the ability to sense external forces.Currently,force sensors are mainly used to realize the perception of external force information,and the cost is relatively high.In the process of human-machine collaboration,direct teaching technology can adapt to frequent changes of production tasks,quickly reprogram the robot,and can better meet the needs of modern industrial high productivity.This paper proposes an external force estimation method based on Kalman filter,which relies only on the measurement signals(current,angular displacement)of the motor.Based on the idea of the disturbance observer,the a priori information of external joint torque is introduced into the generalized momentum dynamic model,a linear state-space equation is established,and combined with standard Kalman filter to estimate the external joint torque.In order to successfully apply the proposed method,a nonlinear friction model was established for the nonlinearity and uncertainty of friction at low speeds,and the measurement noise and the uncertainty of friction related to speed were modeled as the disturbance torque obeying the Gaussian distribution.In addition,in order to improve the identification effect of the dynamic model parameters,the limitations of the traditional least squares method are analyzed,and the weighted least squares are used to estimate the parameters,and the improvement effect brought by the method is proved in the experiment.In order to apply the external force estimation method to the direct teaching of robots,this paper combines the motion state information of the robot with a speed-dependent dynamic threshold to preprocess the estimated external torque error.A startup planning method is proposed to reduce the startup torque threshold caused bystatic friction.The processed external torque is combined with admittance control to achieve direct teaching without force sensors.Finally,on the designed experimental platform,the effectiveness of the external force estimation method is verified,and based on the estimated external joint torque,a direct teaching experiment of the robot joint space is carried out.The experimental results show that the estimated external torque and the proposed The preprocessing strategy can realize the direct teaching function.