Beruflich Dokumente
Kultur Dokumente
Automation in Construction
journal homepage: www.elsevier.com/locate/autcon
A fast path planning method for single and dual crane erections
Yu-Cheng Chang, Wei-Han Hung, Shih-Chung Kang
Department of Civil Engineering, National Taiwan University, Taiwan
a r t i c l e
i n f o
Article history:
Accepted 27 November 2011
Available online 2 January 2012
Keywords:
Crane
Erection
Dual crane
Path planning
Robotic
a b s t r a c t
This research aims to develop a method for planning the erection path automatically and efciently. The proposed method is comprised of two steps. The rst step is to convert the scene of the crane erection into a conguration space, in which the crane's load capacity and the obstacles in the environment have been included.
The second step is to nd a collision-free path in the conguration space by using the probabilistic road map
(PRM) method. Three tests were conducted to validate the action, crane placproposed method in this research. The results show that the proposed method is efcient, and can generate effective erection paths
for operating in near real-time scenarios. The method is appropriate for both single and dual crane erection,
and can help engineers plan more easily, and verify erection-planning decisions such as crane seleement, and
logistcs.
2011 Elsevier B.V. All rights reserved.
1. Introduction
In addition to the usual single crane erection, dual-crane cooperative erection has become more common in modern construction projects in recent years. Particularly, in industrial construction, it is often
necessary to transport large facilities; this requires cranes with a capacity of 5001000 t. Although available large cranes can move
weights of up to 1300 t, their operation may be restricted by the limited space on site, while the construction costs will increase if a larger
and more expensive crane is rented. A useful, often used alternative is
to utilize two less expensive cranes to perform a cooperative crane
erection [10,17]. However, in the cooperative dual crane erection process, the two cranes need to work together to maintain the equilibrium of heavy loads. The complexity of cooperative dual crane erection
is far higher than that of single crane operation, and this can lead to
high risk situations during construction [26]. If a feasible and safer
erection path can be pre-planned for cooperative dual crane erection,
then such high risk situations can be reduced to a minimum; further
justifying the importance of erection path planning.
Erection path planning is a complex topic, and there are three major
difculties involved. Firstly, the load of the crane should be within its lifting capacity during the erection process; the upper limit of the capacity
varies with the angle of the boom, which makes planning more difcult.
Secondly, collisions among the crane, the lifting object, and any obstacle
should be avoided. This makes the planning difcult when there are numerous obstacles on site, and the volume of the lifting object is large.
Thirdly, the cable of the crane must be kept vertical plumbed during a cooperative dual crane erection in order to avoid increased tension from
Corresponding author.
E-mail address: sckang@caece.net (S.-C. Kang).
0926-5805/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.autcon.2011.11.006
469
end point is not nil. If the computational time is not limited, PRM
can certainly be used to nd a feasible path.
The PRM method is based on probability, where the basic process
is to repeatedly guess the collision-free points and try to link them
into a collision-free path. The computation time for PRM depends
on the number of path nodes. The fewer nodes there are, the shorter
is the computation time. However, the probability of nd a feasible
path is also lower, and vice-versa, having more nodes leads to longer
computation time and a higher probability of nding a feasible path.
The PRM method is suitable for erection path planning. Erection
path planning is different from the maze problem of robotics, which
involves guring out where the corner is and how to navigate
through narrow paths. In construction practice, we usually maximize
the working space for cranes. This means PRM can compute (or
guess) a collision-free path in a very short amount of time, and only
a few nodes are sufcient to plan an erection path. For a more complicated erection operation, more nodes can be used to nd a feasible
path. Furthermore, regarding efciency, a crane operator may be
more interested in obtaining a feasible path quickly rather an optimal
path slowly. This is consistent with the main concept of the PRM
method. Therefore, this research makes use of the PRM method for
erection path planning, making it possible to nd a feasible erection
path in near real-time automatically.
3. Erection path planning for single crane erection
In this section we will introduce how a Conguration Space (C-Space)
is built for the single crane erection procedure, and nding a collision free
path from the C-space as the erection path for single crane operations.
3.1. Assumptions
The method for erection path planning presented in this research
is valid only under the following assumptions:
1. For the calculation of path planning, all the obstacles are assumed
to be static during the erection operation, except for the crane and
the lifting object.
2. There is no change in the location where the crane is set up during
the erection operation. In almost all erection operations, the crane
is set up at a xed location and the crane cannot move during the
erection operation.
3. The sway of the lifting object during the erection process is
accounted for using a larger boundary for the lifting object
model. We can measure the possible sway range and set an outer
boundary. Then we use the boundary to computer the collisionfree path. This avoids virtually all possible collisions due to the
sway. Crane operators are asked to minimize the cable sway to reduce risks. They usually stop moving the lifting object when it is
swinging until the object becomes static. However, the problem
of whether the dimensions of the lifting object model are enough
is beyond the scope of this research.
3.2. Procedure of erection path planning
For the erection path planning method presented here, the ow is
divided into three parts as shown in Fig. 1. Before planning an erection path, the user must provide information about the selected
crane including its lifting capacity and initial location. A C-space is
then constructed for the selected crane, and path planning is then
conducted in the C-space. Finally, operation planning is performed
for the hoisting, and a collision free and feasible erection path is
planned. If a feasible erection path cannot be found, users then can alternate the crane with other cranes that have better capacity or adjust
the initial location of the selected crane until they can nd a feasible
erection path.
470
nodes in C-space, and then connect all possible nodes to form different paths. Finally, we select the optimal path from all possible paths
connecting the starting point to the end point, which forms the erection path for the crane.
To calculate the optimal path from found paths, we use the variation in angle of the crane base-swing, the variation in angle of the
crane boom-luff, and variation in length of the hoist height as the
basis for evaluation, and for developing a Pcost function to calculate
the costs for the erection path In , as shown in Eq. (1). The reason
we do not use the length of the path for evaluation is that the position
of the lifting object is the end-point result of the three degrees of freedom (base-swing angle, boom-luff angle, and the hoist height). The
shortest path (e.g. a straight path in the 3-dimension space) is usually
more difcult to follow for the crane operators because they have to
simultaneously control these three degrees of freedom with changing
velocities. For a human being, it is easier, safer, and more stable to
control only one or two degrees of freedom with the same velocity.
Therefore, we use the variation of these three degrees of freedom
for the cost measurement. Fig. 4 shows an example of cost estimation
by the variation of base-swing angle and boom-luff angle.
In order to have the same evaluation criterion for angle (baseswing angle and boom-luff angle) and length (hoist height), the developed cost function uses the time of operation to reach the vibration as shown in Eq. (2), where t represents the total time of
operation needed for the total variation of base-swing angle , t represents the total time of operation needed for the total variation of
boom-luff angle , and th represents the total time of operation needed
for the variation of hoist height Hob _ max. The reason we use Hob _ max instead of in=11|hi| will be discussed in the next section. Therefore,
Eq. (2) can be transformed into Eq. (3), where is the angular speed
of the base-swing, is the angular speed of the boom-luff, and Vh is
the speed of hoisting. Then we can use Eq. (3) as the cost function to
471
Table 1
Algorithm of cObstacleCheck.
Algorithm cObstacleCheck(, ): determine the conguration(, ) is C-obstacle or not
: base-swing angle.
: boom-luff angle.
IF obstacle collided with boom(, ) or base() THEN
the conguration(, ) is C-obstacle
ELSE
getHoistHeightRange(, )
IF hmin hmax THEN
the conguration(, ) is C-obstacle
ELSE the conguration(, ) is not C-obstacle
Table 2
Algorithm of getHoistHeightRange.
Algorithm getHoistHeightRange(, ): nd hoist height range hmin and hmax
: base-swing angle.
: boom-luff angle.
h: hoist height
h: hoist height increase in each interaction
hmin: minimal hoist height for (, )
hmax: maximal hoist height for (, )
LET h = 0
REPEAT:
IF hmin not found THEN
IF object(, ) does not collided with boom(, ) THEN
hmin = h
ELSE
h = h + h
ELSE
IF object(, ) collided with obstacle, ground, or base() THEN
hmax = h
ELSE
h = h + h
UNTIL hmax found
RETURN hmin and hmax
the erection path may tend to cross the obstacle from the top. However the retract/release action of the cable costs more time and make
the erection more inefcient. Therefore the cost function can be dened as Eq. (1), where the W represents 1, the W represents
1, and Wh represents a Vh 1.
n1
P cost In t t t h
P cost In
n1
i1 ji j
1
V h
n
Hob _ max
Vh
W
W
Wh
Fig. 3. Layout of a crane's 2D C-space.
P cost In
In
measure the path. The larger the calculated result of the Pcost function,
the longer it will take to execute the path, and the harder it will be to
operate the crane. The path with the smallest calculated result of the
Pcost function is the optimal path.
However in practice, sometimes we may multiply the time cost of
each degree-of-freedom by a weight coefcient (W, W, and Wh) before summing together as shown in Eq. (4). These coefcients can signicantly inuence the erection path chosen. For example if we lower
the weight of the crane base-swing angle, the method tends to nd a
solution that needs fewer operations on the other parts. The erection
path found then is easier and faster to execute for crane operators.
However, it may increase the working area of the crane and therefore
raise the risk when lifting, making it unsuitable for a narrow working
space. Conversely, if we lower the weight of hoist height variation,
n1
P cost In W i1 ji j W i1 ji j W h H obm ax
n1
i1 ji j
H obm ax
n1
i1 ji j V h
1
H obm ax
n1
i1 ji j
4
Erection path represented by connecting n node congurations where the 1-st node is the starting conguration and
the n-th node is the end conguration;
Variation in angle of the crane base-swing angle during
the path section connecting the i-th node to the (i + 1)-th
node in degrees;
Variation in angle of the crane boom-luff angle during the
path section connecting the i-th node to the (i + 1)-th node
in degrees;
Number of nodes for the path;
The highest obstacle among congurations along the path; in
meters (m). Fig. 5 shows an example of a path with four
nodes, and where Hob _ max is the maximum value of the Hob
in the Fig. 5. Since the 2D C-space method in this investigation
does not include a parameter for the hoist height, we use the
maximum height of obstacle that the lifting object needs to
move past as a rough estimate for the change in hoist height.
Angular speed of base-swing, in (degrees/s).
Angular speed of boom-luff, in (degrees/s).
Speed of hoisting, in (meter/s).
The weight of the crane base-swing angle change in the cost
estimation for the path.
The weight of the crane boom-luff angle change in the cost
estimation for the path.
The weight of the hoist height change in the cost estimation
for the path.
The factor of the crane base-swing change in the cost estimation for the path.
The factor of the crane boom-luff change in the cost estimation for the path.
The factor of hoist height change in the cost estimation for
the path.
472
In the existing research on erection path planning [26], the operation of crane is expressed by a 3D C-space under the condition that
the crane does not move during the erection operation. The coordinates
of C-space are the crane base-swing angle , the boom-luff , and the
hoist height h. All crane operations are expressed in the C-space by
[, , h].
In this research, we have found that it is not necessary to consider
the hoist height as one of the coordinates in the C-space. This is because in the erection operation, we change the hoist height to lift
the load higher in order to avoid collisions between the load and
the ground or obstacles. In fact, we only need to know the extent of
hoist height reduction during the erection operation to enable the
load to clear the obstacles. After the load is lifted during the erection
operation, the load must be raised to a height above the obstacles in
the planned path to ensure that the load does not collide with either
the ground or the obstacles. Therefore, in this investigation, we have
found the minimum hoist height hmin, the maximum hoist height
hmax, and the height of obstacle Hob corresponding to each conguration in the C-space, we have also simplied the 3D crane C-space to a
2D conguration space [, ]. This method signicantly reduces the
computational time and complexity of path planning, and thus we
are able to achieve a faster erection path planning procedure.
4. Erection path planning for dual cooperative crane
This study extends the aforementioned method for single crane
erection path planning to a method for a dual, cooperative crane.
The path planning ow chart is still the same as in Fig. 1, except
that in order to model the dual crane system, two 2D C-spaces are
used. A method was developed to connect the two C-spaces constructed for each crane, and nd all possible path nodes so that erection path planning can be conducted for dual cooperative cranes.
4.1. Building the C-space for dual cranes
In order to describe the dual crane system, we built the individual
C-space for each of the two cranes, and address them as two cranes,
crane A and crane B. The C-space for crane A is CA and that for crane
B is CB.
We rst dene the coordinate ranges for CA and CB. The weight of
the object is W. For the object, since the dual crane cooperative operation is evenly divided between the two cranes, the weight of W/2 is
separately substituted into the weight tables of crane A and crane B to
nd crane booms largest luff angle max(W) and its smallest luff angle
min(W). Conguration CA represents a possible connection point between crane A and the object, and Conguration CB represents a possible connection point between crane B and the object.
We then use the method given in Section 3.3 to separately determine the regions for CA-obstacle and CB-obstacle. However, unlike the
case of single crane operation, the inspection of object collision is performed by the use of the object and a partial crane model at the end of
each connection. Since the purpose of the dual crane C-space is to nd
the connectable congurations for the crane and the object, collision
inspection between the object and obstacles is carried out at the
stage of path planning. After completing the C-obstacle region determination, we record the cable extension range and heights of obstacles for each conguration.
4.2. Path planning for dual cranes
connection point between the reference crane and the load. Next, we
search from the other crane's corresponding connection point, and
use the connection congurations between the two cranes and object
as a set of path nodes. This process continues until an adequate number of nodes have been obtained and then the nodes are connected to
form the path. Finally, we select the optimal solution from all possible
paths that connect the starting point to the goal as the erection path
for dual cooperative crane operation. The ow chart for dual cooperative crane path planning is shown in Fig. 7.
To explain the method of the dual crane path planning proposed in
this research further, we present a simple case of dual cooperative crane
path planning as an example, as shown in Fig. 8. First, we enter the roadmap into a path node taken from the connecting point (CA _ s,CB _ s) between the object and the two cranes at the starting position. Then, we
enter the roadmap into a path node taken from the connecting point
(CA _ g,CB _ g) between the object and the two cranes at the goal position.
We then advance into the process of path node sampling.
During the path node sampling process, cranes A and B take turns
to serve as the reference node for sampling. When crane A is the reference node, a sample is arbitrarily taken from CA (indicated as CA _ n)
that is not within the region of C-obstacle. CA _ n then act as the connection point between the load and crane A. The next step is to nd
a possible connection point CB _ n, corresponding to CA _ n, which connects crane B and the object. Consider CA _ n as the center of a circle
and use the height of obstacle Hob corresponding to CA _ n as the height
473
to be lifted. Then, on the XY plane of the workspace with XYZ coordinate system, rotate the object so that it is parallel to the line C A C B ,
which connects the centers of rotation of the booms of the two cranes
as shown in Fig. 8. If at this time, the object in the XYZ coordinate system does not collide with either the obstacle or the cranes and CB _ n
and CB are not in the region of C-obstacle, then we use (CA _ n, CB _ n)
as a node and insert it into the roadmap.
However, if the object collides with the obstacle, then subsequently we take CA _ n as the center of the circle and rotate the object in the
XY plane both clockwise and counterclockwise until there is no further collision. If the rotation angle of the load is 1 b 2 or 2 b 2 and
CB _ n is not C-obstacle, then we use (CA _ n, CB _ n) as a node and insert
it to the roadmap as shown in Fig. 9.
The reason for alternatively selecting cranes A and B as the reference node is to provide a variety of paths. In theory, the more diverse
the sampling points, the higher the possibility of nding collision-free
paths. Since the PRM method is based on probability where we randomly sample the nodes in the conguration space and try to arrive
at a feasible solution instead of checking all possible solutions and
choose the optimum one. Therefore, alternative selection of the reference point can increase the randomness without adding to computational cost. Since this study uses the connection point of the load and
the end of the crane as the center of the circle, the boom is turned in
order to avoid the obstacle so that when a different crane is used as
the reference-sampling node, various nodes will be possible.
When connecting the collision-free path, we connect the nodes
along with the connection point between the object and crane using
straight lines to form the path in the working space with the XYZ coordinate system, as shown in Fig. 10. To judge whether the path between the nodes is a collision-free path, the following three-step
process is needed. First, we nd the height Hobject to lift the object
over the obstacle without collision for the section of erection path between the two nodes. Secondly, we inspect the connection points between the object and crane in the trajectory of path to see whether
they will go past the C-obstacle in the spaces between CA and CB. If
CB_n
CA_n
CB_n
Fig. 9. Node sampling in cooperative dual crane path planning, when an object collides
with an obstacle.
474
they do pass the C-obstacle, then this shows that collision will occur
at the loading point of the path so that this section between nodes
will not be a part of the path. Finally, we inspect the amount of
cable extension needed to raise the height to Hobject and see if the
length exceeds what is permitted. If it exceeds the length limit, it
means that it is not possible to lift the object over the obstacle by
using the path, and therefore the section between these nodes cannot
be a part of the path. If the section between two nodes is judged to be
a collision-free path, then record the height Hobject, which will be used
in the calculation for optimal path and in planning for hoist height.
After completing the collision-free path connection, those paths
that successfully go from the starting point to the end point through
the nodes are designated as allowable erection paths. If no such
path can be found, then further sampling of nodes should be performed and they should be added to the roadmap. The process continues until a path can be found that connects the starting point to
the end point.
For calculation for the optimal path, we make use of the Pcost function proposed in Section 3.4 to calculate the costs for the erection
path. The costs for the paths of crane A and crane B are then separately estimated by using the Pcost function in which results are then combined. The larger the resulting value is, the more time it will take to
execute the path and the harder it will be to operate the cranes.
Fig. 11. Taking required height that ensure the object avoids an obstacle as the erection
path.
We introduce the computer software Erection Planner in this section and its architecture and development environment for erection
as the load capacity table of the crane, the weight of the object, and
coordinates of starting and end points for the object.
5.2.2. C-space builder
The purpose of this part is to build the C-space. The C-space is built
based on the method for building the C-space presented in Sections 3.3
and 4.1, and is built according to the type of erection operation. To implement the collision detection needed in the C-space building procedure,
we use PhysX to build simple physical models for the crane, the object,
and the obstacles. In the PhysX physical action simulation, the baseswing angle, the boom-luff angle, and the position of object are changed
if any collisions occur.
5.2.3. Path planner
The purpose of this part is to implement the method of path planning proposed in Sections 3.4 and 4.2, and essentially carry out erection planning to obtain path information including the base-swing
angle and the boom-luff angle.
5.2.4. Hoisting planner
This part plans the hoisting operation for the erection path by the
method of hoisting planning described in Sections 3.5 and 4.3, and
sends out the completed erection path information.
5.2.5. Scene visualization
In this part, a virtual erection scene is built using XNA, and the trajectory of the object is shown in the erection scene. Fig. 5.4 shows the
resulting erection path using the current method. The visualization
part can be integrated with other 3D simulation tools for more advanced and graphically detailed visualization e.g. Maya, 3ds Max and
Blender. Users can retrieve the project information from the Erection
Project Information Input and combine it with the planned output of
the Hoisting Planner (which contains the crane congurations in the
path) and then adapt it to their visualization or simulation tool.
6. Experiment result and discussion
In order to verify the method of erection path planning proposed
in this research, we have conducted a series of erection scene experiments to test if the proposed method can be applied to both single
crane erection operation, and dual cooperative erection operation.
475
476
Fig. 15. Erection path of Experiment 1 (The goal position of the object is visualized using translucent yellow color).
position of the object and the red part shows the end position). Two
different paths were found: one went around obstacle B from the
left side, while the other from the right side. Afterwards, we hoped
to nd paths that did not go past any obstacles but passed through
the narrow space between the cranes and obstacles. The results are
shown in Figs. 20 and 21.
6.3. Experiment 3: Efciency test
In Experiment 3, we tested using the same two erection scenes
from Ref. [2]. We imitated the method of building the 3D C-space
and also made use of Genetic Algorithm based method of erection
planning, both reported in Ref. [2], and compared it with the method
presented in this research. In addition, we also compared results
obtained by using the proposed PRM and the planning conducted either in both 3D and 2D C-spaces.
6.3.1. Erection Scene 1
Erection Scene 1 used the scene 1 from the dual cooperative crane
erection of [2] as shown in Fig. 22. The object was a beam (10 m in
length, and with height and width equal to 1 m), and the two cranes
were connected to either end of the beam. The two cranes were required to lift the object together from the ground and place it on
the platform on top of an obstacle 6 m high. The body of the crane
was 10 m long, 6 m wide, and the length of the boom was 45 m,
with base-swing angular speed of 4/s, boom-luff angular speed of
1/s, and hoist height change speed of 1 m/s. Fig. 23(a) shows the
477
can adjust the weight values to generate feasible and suitable erection paths for different specications and site environments.
The developed method can potentially benet many aspects of
current construction practice. The possible benets of this research
are summarized in the following paragraphs:
Crane selection: Rather than having a qualitative crane selection, the
methods and tools developed in this research can be used to provide
quantitative information about the differences from using the various candidate cranes. Similarly, for large or high-rise construction
projects, this research developed tools that provide data to enable
proper evaluation of the compromises and benets of using of multiple cranes at the site as opposed to using a single crane.
Crane placement: The methods developed in this research enable
computers to simulate the erection processes automatically by
utilizing the planned erection path. They can also be extended to
obtain erection times produced by placing the crane at different
locations and searching for an optimum location within the site
that will minimize erection times.
Logistics for scheduling material deliveries to the site: This study can
be used to improve various aspects of the logistics at construction
sites. In particular, the generation of a precise and detailed erection plan prior to construction can minimize onsite storage requirements by delivering only the materials that will be erected
soon to the site. For example, structural elements can be delivered
only one day prior to their erection, or for a site where no onsite
temporary storage is available, they can be delivered just a few
hours prior to erection. Furthermore, elements to be lifted by the
crane can be delivered to the site and even placed in the delivery
truck in accordance with the plan for lifting them using the crane.
Planning and visualization of erection virtually prior to actual erection
activities: This research provides tools for planning and visualizing
Fig. 18. Experiment 2 Substituting path 1 when the path did not go through the middle obstacle.
478
Fig. 19. Experiment 2 Substituting path 2 when the path did not go through the middle obstacle.
Fig. 20. Experiment 2 Substituting path 1 when the path did not go through any obstacle.
Fig. 21. Experiment 2 Substituting path 2 when the path did not go through any obstacle.
479
Fig. 23. Scene 1 of Experiment 3 Erection path planned: (a) Genetic Algorithm in 3D C-space; (b) by PRM in 3D C-space; and (c) by PRM in 2D C-space.
Fig. 24. Scene 2 of Experiment 3 Erection path planned: (a) Genetic Algorithm in 3D C-space; (b) by PRM in 3D C-space; and (c) by PRM in 2D C-space.
Problem 1
GA
(3D C-space)
PRM
(3D C-space)
PRM
(2D C-space)
36.33
71.47
12.29
31.51
11.26
69.81
13.15
32.77
0.51
67.64
10.32
26.69
Table 4
Scene 2 of Experiment 3. Efciency comparison for different methods of path planning.
Planning method
Problem 2
GA
(3D C-space)
PRM
(3D C-space)
PRM
(2D C-space)
59.42
144.96
24.81
106.11
11.32
143.55
23.22
109.78
0.52
142.93
19.14
93.06
480
results of rst two tests show that the proposed method can help nd
a collision-free erection path that satises safety requirements, and
can assist engineers in solving the path-planning problem. From the
third experiment, we found that the proposed method of erection
path planning is time efcient when used to nd a feasible erection
path compared to existing methods. The obtained path by the proposed method is also easier to operate. The method is also exible
in being able to nd suitable erection paths under different conditions, as required on the erection site.
References
[1] E.U. Acar, Morse decompositions for coverage tasks, International Journal of Robotics Research 21 (4) (2002) 331345.
[2] M.S. Ali, N.R. Babu, K. Varghese, Collision free path planning of cooperative crane
manipulators using genetic algorithm, Journal of Computing in Civil Engineering
19 (2) (2005) 182193.
[3] Autodesk, 3ds maxURL, http://www.autodesk.com.tw/2010.
[4] R.A. Brooks, T. Lozano-Perez, A subdivision algorithm in conguration space for
ndpath with rotation, AI Memo 684, AI Laboratory, MIT, Cambridge, MA, 1982.
[5] H. Choset, K.M. Lynch, S. Hutchinson, G. Kantor, W. Burgard, L.E. Kavraki, S. Thrun,
Principles of Robot Motion: Theory, Algorithms, and Implementations, The MIT
Press, 2005.
[6] R. Daily, D.M. Bevly, Harmonic potential eld path planning for high speed vehicles, American Control Conference, 2008.
[7] D.E. Goldberg, J.H. Holland, Genetic algorithms and machine learning, Machine
Learning 3 (2) (1988) 9599.
[8] L.J. Guibas, J.C. Latombe, S.M. Lavalle, D. Lin, R. Motwani, A visibility-based pursuitevasion problem, International Journal of Computational Geometry and Applications 9 (4) (1999) 471493.
[9] J. Guldner, Sliding mode control for gradient tracking and robot navigationusing
articial potential elds, IEEE Transactions on Robotics and Automation 11 (2)
(1995) 247254.
[10] W.C. Hornaday, C.T. Haas, Computer-aided planning for heavy lifts, Journal of
Construction Engineering Management 119 (3) (1993) 498515.
[11] S.C. Kang, E. Miranda, Planning and visualization for automated robotic crane
erection processes in construction, Automation in Construction 15 (4) (2006)
398414.
[12] L.E. Kavraki, P. Svestka, J.C. Latombe, M.H. Overmars, Probabilistic roadmaps for
path planning in high-dimensional conguration spaces, IEEE Transactions on
Robotics and Automation 12 (4) (1996) 556580.
[13] O. Khatib, Real-time obstacle avoidance for manipulators and mobile robots, International Journal of Robotics Research 5 (1) (1996) 9098.
[14] A.M. Ladd, L.E. Kavraki, Measure theoretic analysis of probabilistic path planning,
IEEE Transactions on Robotics and Automation 20 (2) (2004) 2292428.
[15] R.A. Lewis, A.K. Bejczy, Planning considerations for a roving robot with arm, Proceedings of the 3rd International Joint Conference on Articial Intelligence, 1973.
[16] Y. Li, K.C. Ng, D.J. Murray-Smith, G.J. Gray, K.C. Sharman, Genetic algorithm automated approach to design of sliding mode control systems, International Journal
of Control 63 (4) (1996) 721739.
[17] K.L. Lin, C.T. Haas, Multiple heavy lifts optimization, Journal of Construction Engineering Management 122 (4) (1996) 354362.
[18] Y. Liu, S. Arimoto, Path planning using a tangent graph for mobile robots among
polygonal and curved obstacles, International Journal of Robotics Research 11
(4) (1992) 376382.
[19] T. Lozano-Perez, M.A. Wesley, An algorithm for planning collision-free paths
among polyhedral obstacles, Communications of the ACM 22 (10) (1979)
560570.
[20] Microsoft, XnaURL, http://msdn.microsoft.com/en-us/xna/default.aspx2008.
[21] M. Mitchell, An Introduction to Genetic Algorithms, Massachusetts Institute of
Technology, 1996.
[22] A. Mohri, M. Yamamoto, G. Hirano, Cooperative path planning for two manipulators, IEEE International Conference on Robotics and Automation, , 1996,
pp. 28532858.
[23] NVIDIA, Nvidia physx sdk 2.8.1 documentation, , 2008.
[24] B. Oommen, S. Iyengar, N. Rao, R. Kashyap, Robot navigation in unknown terrains
using learned visibility graphs, IEEE Transactions on Robotics and Automation 3
(6) (1996) 672681.
[25] H.R. Reddy, K. Varghese, Automated path planning for mobile crane lifts, Computer-Aided Civil and Infrastructure Engineering 17 (6) (2002) 439448.
[26] H.I. Shapiro, J.P. Shapiro, L.K. Shapiro, Cranes and Derricks, Mc Graw Hill, 1991.
[27] P. Song, V. Kumar, A Potential Field Based Approach to Multi-Robot Manipulation,
IEEE International Conference on Robotics and Automation, 2002.
[28] X. Wang, Y.Y. Zhang, D. Wu, S.D. Gao, Collision-free path planning for mobile
cranes based on ant colony algorithm, Journal of Key Engineering Materials
467469 (2011) 11081115.
[29] J. Zhang, W.L. Lo, H. Chung, Pseudo-coevolutionary genetic algorithms for power
electronic circuits optimization, Pseudo-Coevolutionary Genetic Algorithms for
Power Electronic Circuits Optimization 36 (4) (2006) 590598.
[30] C. Zhang, H. AlBahnassi, A. Hammad, Improving Construction Safety through
Real-Time Motion Planning of Cranes, The International Conference on Computing in Civil and Building Engineering (ISCCBE)-2010, 2010.