Research On The Active Disturbance Rejection Control Strategy For The Flexible-joints Of Industrial Robots

As an important role in China’s manufacturing industry,the position control precision and rapid response performance of industrial robot joints directly affect the production quality and efficiency.In the existing robot joints,harmonic gears with high reduction ratio,high load weight ratio and high precision are usually used to meet the overall lightweight design requirements.However,harmonic gears will make it difficult to design the robot joint controller in terms of command tracking and disturbance attenuation.1)The harmonic gear behaves like a torsional spring,which makes the control degrees of freedom of the flexible joint mismatch the output degrees of freedom.The traditional rigid joint control method will cause the underdamped vibration for a long period of time in the positioning process,and additional static error of the position will occur when undergone load disturbance.2)The rotational inertia of the permanent magnet synchronous servo motor is significantly reduced by using the harmonic gear,so it will be confronted with more complex mechanical and electrical transient process in the current loop control.It is necessary to improve the disturbance-rejection performance of the current loop to provide better torque output for the flexible joint.In view of the above problems,this paper mainly studies the topology design of flexible joint position controllers and current controllers.First of all,since the flexible joint is a high-order system containing coupling units and cascade units,the basic problems of applying Active Disturbance Rejection Control(ADRC)are discussed.When determining the topology of controllers,decoupling controllers are set separately for the coupling units,and related cascade units are combined together according to the requirements of control bandwidth and disturbance observation.In this way,the design procedure is clarified for the simplified system.When the sampled signal is influenced by noise,the system stability can be ensured by sacrificing the observer bandwidth.In order to avoid noise amplification in cascade systems,tracking differentiator is used to filter the command signals.Secondly,by analyzing the influence of smaller motor inertia and digital control delay on the current loop,it is concluded that the main disturbance of the current loop is the slope disturbance.The influence of control delay is suppressed by angle compensation in coordinate transformation directly.For the influence of smaller motor inertia,the traditional decoupling control strategies such as current feedback decoupling,complex vector decoupling and disturbance attenuation control strategies such as second-order ADRC are analyzed firstly.Then a current control strategy is proposed based on composite observer and model compensation.The strategy has strong parameter robustness and good suppression ability of slope-type disturbance,which is verified in the simulation and experiment via the hardware-in-the-loop platform.Finally,for the underdamped vibration and load torque disturbance in the position control,the traditional methods with limited state feedback are analyzed and compared firstly.Then a position control strategy based on two-stage state feedback and ADRC is proposed.In order to save hardware cost,the method only uses the feedback values from the motor position sensor and link position sensor while the joint torque is calculated with the nominal value of flexibility.The motor angular velocity and link angular velocity are observed by the extended state observer,thus full-state-feedback is realized and equivalent damping of the system is improved.At the same time,on the basis of avoiding continuous noise amplification by using tracking differentiator,the load torque disturbance is observed by using extended state observer and the then compensated.This method has strong parameter robustness and is less affected by sampling noise,while improves tracking performance and disturbance rejection performance with less sensor requirements.Simulation results show its effectiveness.