Industrial Robot Trajectory Tracking Control Using Multi-Layer Neural Networks Trained by Iterative Learning Control

Abstract

Fast and precise robot motion is needed in many industrial applications. Most industrial robot motion controllers allow externally commanded motion profiles, but the trajectory tracking performance is affected by the robot dynamics and joint servo controllers, to which users have no direct access and about which they have little information. The performance is further compromised by time delays in transmitting the external command as a setpoint to the inner control loop. This paper presents an approach for combining neural networks and iterative learning controls to improve the trajectory tracking performance for a multi-axis articulated industrial robot. For a given desired trajectory, the external command is iteratively refined using a high-fidelity dynamical simulator to compensate for the robot inner-loop dynamics. These desired trajectories and the corresponding refined input trajectories are then used to train multi-layer neural networks to emulate the dynamical inverse of the nonlinear inner-loop dynamics. We show that with a sufficiently rich training set, the trained neural networks generalize well to trajectories beyond the training set as tested in the simulator. In applying the trained neural networks to a physical robot, the tracking performance still improves but not as much as in the simulator. We show that transfer learning effectively bridges the gap between simulation and the physical robot. Finally, we test the trained neural networks on other robot models in simulation and demonstrate the possibility of a general purpose network. Development and evaluation of this methodology are based on the ABB IRB6640-180 industrial robot and ABB RobotStudio software packages.

Flexible-Joint Manipulator Trajectory Tracking with Learned Two-Stage Model employing One-Step Future Prediction

“…The system is tested on two quadrotors with different dynamics and outperforms systems composed of ILC and proportional-derivative (PD) or proportional-integralderivative (PID) controllers. Another use of ILC involves production of the ground truth input trajectories to train NNs which approximate the inverse dynamics of the system by Chen et al [21]. The authors apply this approach to an industrial manipulator, training the model in simulation.…”

“…In order to learn the inverse dynamics approximation, we generate 45 minutes of functional trajectories θ d for the left arm of the Baxter robot. Random sinusoidal trajectories are often used to form the base of the training set [21] [22]. However, recording of such dataset takes the valuable robot hours away from the user.…”

Prediction and Compensation of Contour Error of CNC Systems Based on LSTM Neural-Network

“…The desired trajectory was used as the input of the fitted inverse model and the reference trajectory was obtained as the model output. In [19], an approach combining ANNs and ILC was proposed to improve tracking performance of multi-axis industrial robots. For a given desired trajectory, a high-fidelity dynamic simulator was used to iteratively refine the external instructions to compensate for the robot's inner-loop dynamics.…”

Two‐loop reinforcement learning algorithm for finite‐horizon optimal control of continuous‐time affine nonlinear systems

“…Here comes to the iterative learning control (ILC) technique that has been used in many applications that perform repetitive tasks within finite time interval 35,36 . The works 37,38 apply ILC to tuning the constant weight for the neural networks. Differently, our weight function is time‐varying and continuous.…”