Applied Mechanics and Materials Online: 2010-12-06

ISSN: 1662-7482, Vols. 44-47, pp 355-359

doi:10.4028/www.scientific.net/AMM.44-47.355
© 2011 Trans Tech Publications Ltd, Switzerland

Development of Autonomous Bucket Wheel Reclaimer with Laser

Vision

Jinbo Chen1,2,a, Zhengpeng Yuan1,b, Zhenbang Gong1,c

1
Department of Precision Mechanical Engineering, Shanghai University, Shanghai, China
2
State Key Laboratory of Robotics, Shenyang, China
a
[email protected], [email protected], [email protected]

Keywords: Bucket Wheel Reclaimer, Vision, Autonomous, Laser Scanner

Abstract: An autonomous system of the bucket wheel reclaimer is developed to improve capability
of the reclaimer in dark or bad whether. It obtains the 3D shape of the pile by the laser scanners
mounted the reclaimer, and extracts the landing position through computer vision techniques. All
the operations of the reclaimer can be controlled automatically in the remote control room. The
system can control simultaneously six reclaimers. And it has been successfully implanted to the
reclaimer in the stockyards of Majishan of Bao Steel Co. and show satisfactory result.

Introduction
Bucket wheel reclaimers, as a kind of bulk material handling Machines, are widely used to dig
raw materials from the stock pile and transfer the materials to other factories in modern industry [3].
In Figure 1, a picture of a bucket wheel reclaimer is shown. The traditional operation of the
reclaimer is performed by the operator manually, which involves traveling, slewing, and pitching
through an operation lever in the cab or the remote control room with the aid of some video cameras.
However this operation is not an easy job because the operator’s eyes or the videos cannot
recognize the real 3D shape of the stock pile precisely, especially in dark and bad weather. If the
bucket wheel collided with the pile or other machines, it will be a serious accident. So we
developed an autonomous bucket wheel reclaimer with the laser scanners mounted on both sides of
bucket wheel. The laser scanner, as the eyes of the reclaimer, can “see” the 3D shape of the pile
even in dark or bad weather, and the reclaimer can automatically dig the pile without the
interruption of the workers.

Figure 1: A Picture of a Bucket Wheel Reclaimer

All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of Trans
Tech Publications Ltd, www.scientific.net. (#551199666, São Luís, Brazil-22/01/21,21:12:55)
356 Frontiers of Manufacturing and Design Science

Imaging System of the Laser

To obtain the 3D shape of the pile, two laser scanners are installed separately at the two sides of
the reclaimer, as shown in Figure 2(b), where only the scanner on the right side is shown, and the
left one is behind the wheel. Two aspects are considered in this installation scheme. One is that the
pile is always explored by the laser scanners without any occlusion when the reclaimer is moving. The
other one is that the distance between the pile and the scanner is moderate because the material of
the pile is not good for the reflection of the laser. In addition, for each of the laser scanners, a shield
is added to protect the scanner from rain or dust in the outside work environment, as shown in
Figure 2(a). And a blower is used to clean the lens of the scanners. To ensure stable measurement,
the mounting brackets with good rigidity are needed.

a) b)
Figure 2: Installation of the Laser Scanners
The type of the laser scanners we choose is the SICK LMS-221. The configuration of the laser
scanner is reset for the environment. Its sensitivity is adjust to dark material according to the bad
reflection condition. The view of field and the angular resolution are set to 180° and 0.25° ,
respectively. We use mode 32m which is the maximum distance returning from the laser. The data
link is the RS422 at 500kbps.

Figure 3: The coordinates of a reclaimer

A reclaimer can be treated as a robot arm. Using Danevit-Hartennberg notation[1], coordinate
frames are assigned to the reclaimer as shown in Figure 3. Frame 4 and frame 6 are located the left
and right laser scanners, respectively, and Frame 6 is for the wheel landing position. According to
forward kinematics solutions, The general transformation matrices between Frame 0 and 4 can be
derived as shown in formula (1) below.
 Applied Mechanics and Materials Vols. 44-47 357

Cθ3Cθ 4 −Cθ3 Sθ 4 Sθ3 a4Cθ3Cθ 4 + d 4 Sθ3 

 − Sθ −Cθ 4 0 − L3 − L2 − a4 Sθ 4 
T4 = 
0 4
(1)
 Sθ3Cθ 4 − Sθ3 Sθ 4 −Cθ3 a4Cθ 4 Sθ3 − d 4Cθ3 + d1 
 
 0 0 0 1 

Where d4 is the distance from origin of Frame 3 coordinate to intersection of Z3 & X4 along Z3; a4
is the distance from intersection of Z3 & X4 to origin of Frame 4 coordinate along X4.
The laser scanner is a polar coordinate system. For the left scanner, the coordinate of the point in
Frame 4 is calculated from below formula.
X4 = ( x4 , y4 , z4 )T = (r ⋅ sin( β ), r ⋅ cos( β ),0)T

Where r is the distance returning from the laser scanner , and β is the angle of this beam。So

the coordinate X0 of the same point in Frame 0 is calculated from blew formula.

 x  x4 
 y y 
X0 =   = T40  4  = T40 X 4 (2)
z  z4 
   
1  1
For the right scanner, the operation is the same. So the 3D coordinates of the points of the pile
surface in frame 0 are obtained with the movement of the reclaimer, and a height map is generated
by mapping the range data to intensity value in the x-z image plane. In addition, Non-Local Means
filter is explored to remove the noise of the image data [2]. The 3D shape of a pile is shown in
Figure 4.

Figure 4: the 3D Shape of a Pile

Extraction of the Landing Points

In the operation of the reclaimer, the wheel bucket swings back and forth at the given landing
height. Then it feeds to the next position and repeat the same swing movement, as shown in Figure
5(a). So it is essential to extract the landing pints of each swing. First, the contour lines at the
landing height of the bucket are detected by the contour tracking techniques[4], as shown in Figure
5(b). Then, the curve of the candidates are extracted at the given feed under the constraint condition
that the length(r) is constant，as shown in Figure 5(c). All points on the lines are assumed to be
possible landing points and their feasibility are investigated. If a point brings about overload in the
358 Frontiers of Manufacturing and Design Science

slewing operation, it will be excluded from the set of candidates of landing points. All points on the
contour line are examined if they belong to candidates for landing points by investigating the
digging depth of the bucket wheel. In other word, each of the landing points must satisfy
D<DEPTHmax to avoid overload. DEPTHmax is the maximum digging depth under the nonoverload
condition.

a) b) c)
Figure 5: Extraction of the Landing Points

If the Landing points are confirmed, the movement parameter of the reclaimer can be calculated
using blew formula by inverse kinematics solutions.
− y − L2 − L3
θ 4 = arccos( )
a6
x
θ3 = arcsin( ) −α
a62 cos 2 θ 4 + d 62
d1 = z + d 6 cos θ3 − a6 cos θ 4 sin θ3
Where d6 is the distance from origin of Frame 3 coordinate to intersection of Z3 & X6 along Z3; a6
is the distance from intersection of Z3 & X6 to origin of Frame 6 coordinate along
a6 cos θ 4
X6; α = arctan( ) . P ( x, y, z ) is one of the landing points.
d6

Implement of the System

The whole architecture of the system is sketched in Figure 6(a), which shows the various
formats of information exchanges between the computers and the sensors. The computers in the
control room are connected to PLCs, cameras and Lasers in the reclaimers via Ethernet. The server
manages all operation data in the yard and status of all reclaimers. It gets information for stockpiles
from the image system and the database when they perform reclaiming. It also controls collision
avoidance of all reclaimers in the yard. The image system manages the yard height map which
calculated from the data of the lasers and the attitude of the reclaimers from the OPC server which
communicates with the PLCs of the reclaimers. The client dispatches the operation commands of
the PLCs over the OPC server, and monitor the status of the reclaimers from the information of the
PLCs, the cameras and the 3D shapes of the piles via the HMI which is shown in Figure 6(b). The
client can switch the work mode between the automatic mode and the manual mode. In the
automatic mode, it simultaneously controls six reclaimers. In the manual mode, the workers can
operate the reclaimer by the joystick in the control room. The PLC receives the command from the
 Applied Mechanics and Materials Vols. 44-47 359

client and performs the position control of the moving axis of the reclaimer which consists of
translation on the rail, pitching and slewing.

a) b)
Figure 6: the Stucture and HMI of the System

Conclusions
The proposed System used two laser scanner to create the 3D shape of the pile, and extracted the
landing points of the reclaimer by computer vision techniques. To avoid overload, each of the
candidates are examined. In the control room, the worker can manager six reclaimers
simultaneously in automatic mode. The system has been implanted to the reclaimer in the
stockyards of Majishan of Bao Steel Co.. The error of position between trajectory of the bucket and
the surface of the pile is within 15cm in the landing test, and this error can be accepted in the
operation of the reclaimer.

Acknowledgement
It is a project supported by State Key Laboratory of Robotics（No. RL2006001）, and “Shu
Guang” project which supported by Shanghai Municipal Education Commission and Shanghai
Education Development Foundation（No. 09SG37）, China.

References
[1] Craig, J. J., Introduction to Robotics, Mechanics and Control, 3rd Edition, Addison Wesley, 2005
[2] M. Mahmoudi and G. Sapiro, Fast Image and video denoising via nonlocal means of similar
neighbourhoods, IEEE signal processing letters, Vol. 12, No. 12, 2005.
[3] Robinson G.K., “How much would a blending stockpile reduce variation?” Chemometrics and
Intelligent Laboratory Systems, 74 (2004)121-133
[4] J. Denzler and H. Niemann. Active rays: Polar-transformed active contours for real-time contour
tracking. Real-Time Imaging, 5:203 – 213, 1999.