Automated Feed-Forward Learning for Pressure-Compensated Mobile Hydraulic Valves With Significant Dead-Zone

Abstract

Hydraulic manipulators on mobile machines, whose hydraulic actuators are usually controlled by mobile hydraulic valves, are being considered for robotic closed-loop control. A feed-forward-based strategy combining position and velocity feedback has been found to be an effective method for the motion control of pressure-compensated mobile hydraulic valves that have a significant dead zone. The feed-forward can be manually identified. However, manually identifying the feed-forward models for each valve-actuator pair is often very time-consuming and error-prone. For this practical reason, we propose an automated feed-forward learning method based on velocity and position feedback. We present experimental results for a heavy-duty hydraulic manipulator on a forest forwarder to demonstrate the effectiveness of the proposed method. These results motivate the automated identification of velocity feed-forward models for motion control of heavy-duty hydraulic manipulators controlled by pressure-compensated mobile hydraulic valves that have a significant input dead zone.

Nonlinear Model Learning for Compensation and Feedforward Control of Real-World Hydraulic Actuators Using Gaussian Processes

“…The mathematically-derived equations of flow that describe the dynamics of actuators [8], [9], or the piecewise functions that are characterized by physically meaningful parameters (such as the maximum input currents within deadband) [1], [2], provide us with models that contain very few parameters which can be tuned for a real actuator by nonlinear optimization [1]. One practical aspect of the piecewise models is that since the function is non-decreasing by design, the inverse is very straightforward to obtain [2] for control applications. However, the capacity of these models are further limited, for instance, a fixed parameter for actuator deadzone [2], [22] may not be the ideal solution for accurate control.…”

“…One practical aspect of the piecewise models is that since the function is non-decreasing by design, the inverse is very straightforward to obtain [2] for control applications. However, the capacity of these models are further limited, for instance, a fixed parameter for actuator deadzone [2], [22] may not be the ideal solution for accurate control. More importantly, the training objective for these models uses the same least-squares objectives as NNs, which does not explicitly account for the uncertainties and noise in measurements.…”

“…Despite its long history of developments and improvements, this classical approach to model-based control is not devoid of pitfalls and challenges in the case of hydraulic actuators in heavy-duty machines. Most system identification approaches in modelling the dynamics of hydraulic actuators assume either a physical model with few parameters [1], [2] that are unable to capture complex nonlinear effects and may require excessive This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ controller tuning, or neural networks (NNs) [3]- [7] that contain many parameters, yet require lots of data for training.…”

Autonomous robotic rock breaking using a real‐time 3D visual perception system

“…In this study, velocity feed-forward learning for each valveactuator pair was performed using the algorithm proposed in Nurmi and Mattila (2017). The algorithm identifies a feed-forward model of the valve-actuator pair by driving the actuator in a sinusoidal trajectory, while at the same time using adaptive control methods to map valve control and actuator velocity.…”

HEAP -- The autonomous walking excavator

“…The piston velocity is controlled using a feed-forward lookup table from desired piston velocities to valve currents and a PI-controller running at 100Hz to generate the reference current for the pilot valve. This is the most commonly used way of controlling heavy-duty hydraulic manipulators with valves that have a significant dead zone [29]. The look-up table is identified over the entire range of applicable piston velocities with 13 value pairs.…”