Page 64 - ASME DSCC 2015 Program

Technical Program

nonlinear Model Predictive Controller for an unmanned Ground ConTRIBuTED SESSIon
Vehicle on Variable Terrain 1-13-1 TP7 Industrial Applications
Invited session paper. DSCC2015-9982 Elijah Pierce A 4:00pm–6:00pm
Andrew Eick, David Bevly, Auburn University, Auburn, AL, United States
Session Chair: Joshua Vaughan, University of Louisiana at Lafayette
Rough, off-road terrain contains multiple hazards for an unmanned ground
Session Co-Chair: Jiong Tang, University of Connecticut
vehicle (UGV). In this paper, hazards are classified into three groups: obsta-
cles, rough traversable terrain, and rough untraversable terrain. These three
Remote lead Through Teaching by Human Demonstration Device
types of hazards create a rollover risk for a UGV. A nonlinear model predic-
Contributed short paper. DSCC2015-9808
tive controller (NMPC) that is capable of navigating a UGV through these
Hsien-Chung lin, Te Tang, Masayoshi Tomizuka, Wenjie Chen, University
hazards is presented. The control algorithm features a nonlinear tire model
of California Berkeley, Berkeley, CA, United States
which more accurately captures the dynamics of the UGV when compared
to a linearized tire model, and has a fast enough run time for real time imple- Industrial robots are playing increasingly important roles in production lines.
mentation. On an actual vehicle, the UGV is assumed to be equipped with The traditional pendant programming method, however, is unintuitive and
a perception based sensor, such as a Light Detection And Ranging (LiDAR) time-consuming. Its complicated operation also sets a high requirement
unit, to provide information of the terrain roughness, grade, and elevation. on users. To simplify the robot programming process, many new methods
This information is used by the NMPC to safely control the vehicle to a target have been proposed, such as lead through teaching, teleoperation, and
location. However, for the purposes of this paper, control inputs and terrain human direct demonstration. Each of these methods, however, suffer from
are simulated in CarSim, and the feasibility of real time implementation is their drawbacks. To overcome the drawbacks, a novel robot programming
investigated. method, remote lead through teaching (RLTT), is introduced in this paper.
In RLTT, the operator uses a device to train the robot remotely, allowing the
Vehicle Rollover Prevention using the Zero-Moment Point in an lQR
demonstrators to use the mature lead through teaching techniques in a safe
output Regulator
environment. In order to implement RLTT, the human demonstration device
Invited session paper. DSCC2015-9624
(HDD) is also designed to transfer the demonstration information from the
Paul Stankiewicz, nicolas ochoa lleras, Robert leary, Sean n. Brennan, human to the robot.
Pennsylvania State University, University Park, PA, United States
Crane Workspace Mapping using Image Segmentation and QR Codes
This research investigates vehicle control techniques for rollover prevention
Contributed regular paper. DSCC2015-9823
in a collision avoidance scenario. The zero-moment point (ZMP) is used to
M. Sazzad Rahman, Joshua Vaughan, University of Louisiana at Lafayette,
evaluate the vehicle’s current and near-future rollover propensity with the
Lafayette, LA, United States
purpose of predicting and correcting impending wheel lift. Specifically, a
linear-quadratic (LQ) output regulator is utilized to safely navigate the vehicle This paper presents a novel approach of mapping a crane workspace using
through a collision avoidance maneuver, while employing a weighting a combination of QR code-based and image-segmentation-based mapping
scheme that explicitly accounts for rollover prevention through the ZMP. algorithms. Known objects in the workspace are labeled with a QR code,
Results show that ZMP regulation is able to reduce the peak rollover threat and a database contains the information of the objects. A camera mounted
to the vehicle. Additionally, it is shown that regulation of ZMP in the near on the crane trolley takes pictures as the crane moves through the work-
future (previewed ZMP) does not necessarily result in a safe maneuver at space. The images are then used to produce an image-segmentation-based
the present time. map of the workspace. To produce the QR code-based map, the QR codes
in the images taken with the camera are decoded, and the information
of the corresponding objects are read from the database file. The object
position and orientation are calculated from the position and orientation
of the QR codes, and the map is drawn. Results showed that the mapping
algorithms are more reliable together than they are individually.

64