A new hybrid robot-human control system is under development at the University of Nevada, Reno (UNR), with the goal of assisting the operator, in order to improve machine efficiency and ore control, and reduce operator fatigue and machine wear.
A wide variety of surface mining and construction tasks require rapid removal and handling of massive quantities of material. The goal here is to improve efficiency, accuracy and ease of use for loading shovels and other excavation equipment used in the mining and construction industries.
Effective manual coordination of current, multi-link excavating machines requires a large amount of training and experience. Even experienced shovel operators are susceptible to fatigue or become temporarily distracted, resulting in suboptimal performance, ore dilution, and excessive machine wear.
The mining industry of the future will increasingly employ machine automation and robotics to enhance the efficiency of material handling operations. However, new concepts are needed for assisting the human operator in these operations, until application of fully-automated machines becomes practical. Behind the concepts for the human-robot interlace, is the need to translate the human operator's dimensional space into a realm that computers can understand. Hence, the birth of the Cartesian Control Forward Kinematics Virtual Machine makes this real-time universal coordinate translation possible.
Current mainstream technology for control of excavators, involves hydraulic joysticks connected to proportioning valves that actuate the main hydraulic rams to position and power the excavator's bucket. Control of the bucket's position during all phases of excavation is what the operator needs to focus on, in order to remain productive. This type of control invokes the operator's attention to regulate excavation rate, dynamic positioning and relative excavation force simultaneously. By replacing one or more of these attentive actions with a robotic control, all aspects of the excavator's performance will improve.
In order to communicate with a computer interface, all relevant operations need to be concurrently interactive with the operator. The intended interaction between robot and human occurs in the human's joystick manipulations. Therefore, digital joysticks are fundamentally required to interact with a computer. By retrofitting a hydraulically-controlled excavator with digital controls and electo-hydraulic proportioning valves, digital control signals from electronic joysticks can be input dynamically through a variable transformation processor.
The interface box provides signal routing and conditioning between control system components, and allows the operator to switch between manual and robotically-assisted (RA) modes. In RA mode, a suitable Software Kinematics Transformation(SKT) is selected by the operator for the current task. The joystick control signals supplied by the operator are then re-distributed according through this SKT to produce optimally-coordinated control signals for the electronic-hydraulic valves.
The SKT transforms the joystick input provided by the operator to produce coordinated machine motion according to a task-specific coordination pattern. Position sensors provide machine parameters needed for evaluation of feedback for correction of system nonlinearities & external disturbances. External positioning sensors and workspace model data can also be incorporated into the control algorithm and/or displayed on a screen for the operator's evaluation. Use of feedback control significantly improves accuracy and repeatability of trajectory, by correcting for small modelling errors and hydro-mechanical system nonlinearities.
At the heart of the new Robot-Human Control System is a software-based machine kinematics transformation. Software Kinematics Transformation allows the operator to select a virtual machine kinematics that is most appropriate for the current excavation task. The revolute kinematics defined by the machine geometry (hardware kinematics) naturally produces a nonlinear motion pattern, which is not necessarily optimal for a given task. On a standard, manually controlled machine each control joystick used by the operator is typically assigned to a single link. Consequently, the operator must carefully coordinate the individual machine link velocities in a complex nonlinear fashion in order to produce the desired trajectory. The SKT transforms the joystick input provided by the operator to produce coordinated machine motion according to a task-specific coordination pattern. Several computer simulations were developed in order to test SKT algorithms.
This video clip shows the prototype Bobcat®435 excavator with Cartesian virtual machine being tested with joystick signals generated by the interface computer.
The prototype Bobcat® 435 excavator with Cartesian Software Kinematic Transformation example programmed specifically for trenching was used in a set of realistic trenching exercises to benchmark performance using robotic-assistance. This kinematics transformation reduces the operator's control to a linear interface, and dramatically simplifies machine control for any desired trajectory. This mode is especially suited for tasks which require a linear trajectory (e.g. trenching, slope maintenance). More specifically, the bucket pin trajectory is robotically-assisted in the following figure and video.
An innovative "bucket-steering" kinematics, which is specifically designed for front shovel loading operations in open pit mining, is currently under development.
This operator control allows movement of the bucket position in the major x-axis. That means forward and back from the operators position in the Bobcat's cab.
The left and right joystick motion controls position of the bucket in the major y-axis. That means up and down from the operator's objective in the cab.
Forward and back (transverse) joystick action modifies bucket angle in relation to the major x-axis.
The prototype Bobcat® 435 excavator with Cartesian Software Kinematic Transformation robotically-assisted virtual machine was reconfigured to the loader arrangement and a series of exercises compared performance using robotically-assisted and manual modes. Both computer simulation and prototype test results indicate that robot-human control via task-specific Software Kinematics Transformation, improves machine control accuracy and ease of use. This video depicts the virtual machine running a bucket angle stabilization mode.
Dr. George Danko
Office: (775) 7844284