In order to realize the intelligent manufacturing of production line in 3C,medicine,food and other industries,collaborative robots with collision detection emerge as the times require.Currently,the common solution is to install multiple sensors on the robot or its surrounding which increases the complexity of the structure and the manufacturing cost.In this dissertation,the lightweight and modular collaborative robot developed by our lab is taken as the research object and in-depth research is carried out to realize the collision detection function based on algorithm instead of sensors.According to the the robot's structure model,connecting rod coordinate system is built.The Jacobi matrix representing the generalized transmission ratio from joint space to cartesian space is obtained and the dynamic equation based on Lagrange dynamics formula is established.The simulation is carried out by using Adams and Matlab to verify the correctness of the dynamic equation.According to the robot's energy equation and momentum equation,the collision operator based on energy and the external torque observer based on momentum including the first and second order are established respectively.The collision detection algorithm is simulated by using Adams and Matlab and offline simulation is carried out by using the off-line preservation function of the motor driver to verify the correctness.A reasonable method is provided for the setting of collision threshold at each joint and the zero-moment mode is put forward to avoid harm to people.Modeling and experiment of identification algorithm for joint mass and centroid position of robot is proposed,and the related parameters are solved by least square method.The correctness of the identification algorithm is verified by comparing the calculated and measured values of joint drive torque.The Stribeck friction model is used to describe and identify the joint friction torque during the harmonic drive which removes the friction in the first order external torque observer and makes the observation value only affect by the impact force.The external torque observer is written in C# language to realize the on-line detection,and the interface is designed to realize the real-time monitoring of the observation value.Finally,the experiment is carried out on the collaborative robot and results show that the observation value follows the change of the external torque in real time and stays under the threshold in normal operation of robot.However,it immediately exceeds the threshold when a collision occurs and the robot switches to the zero-moment mode which verifies the correctness of the collision detection algorithm. |