You are on page 1of 5

The Research of Motion Planning for Humanoid Robots

Lu Xueqin, Qiu Rongfu
Department of Information and Control Engineering
Shanghai University of Electric Power
Shanghai, China
e-mail: lvxueqin@shiep.edu.cn
Abstract-This paper presents research advances in motion
planning for humanoid robots, which has been a focus of
robotics research in recent years. The twelve-link model of
humanoid robot is analyzed. Then the intelligent control
techniques on path planning and gait planning have been
introduced. The three algorithms of gait planning are
fuzzy logic algorithm, neural network algorithm and
genetic algorithm. The analysis and estimation aim to
provide a typical overview on motion planning for
humanoid robots.
Keywords-humanoid robot; path planning; gaits
planning; fuzzy logic algorithm; neural network algorithm;
genetic algorithm.
I. INTRODUCTION
The research in humanoid robots is now on its way
of diverging into various categories. The research on
such areas as artificial intelligence, hardware
development, realization of biped locomotion, and
interaction with the environment are gaining a rapid
phase of development with the help of the rapid growth
of technology [1].
By today, it has come to a situation, where the robot
should be able to perform a wide variety of functions
that helps people in their daily life. Humanoid robots
are designed to imitate human movement, behavior, and
activities. These robots can sense, actuate, plan, control,
and execute activities. Among the successful humanoid
robots include ASIMO from Honda, QRIO from Sony,,
and Advanced Media [2]. In China some famous
University like National University of Defense
Technology, Harbin Institute of Technology has
developed Forerunner and HIT   , but there are some
gaps on key technologies between China and Japan.
Such robots require both path planning and gait
planning.
While in the field of humanoid robot, there are some
limitations in the biped walking velocity, stability and
adapting to environment. At some externs, the function
of humanoid robot is depended by motion planning. For
humanoid robots poses a particular challenge.
Developing practical motion planning algorithms for
humanoid robots is a daunting task given that such
robots typically have 30 or more degrees of freedom
(DOF) [3]. The problem is further complicated by the
fact that humanoid robots must be controlled very
carefully in order to maintain overall static and dynamic
stability. These constraints severely restrict the set of
allowable configurations and prohibit the direct
application of existing motion planning techniques.
In the paper, we introduced the classic model of the
humanoid robot first, which is the foundation of
designing humanoid robot. Then, we discuss the motion
planning which focus on path planning and gait
planning. The three methods will be showed to solve the
problem of the gait planning. We also discuss the
advantages and disadvantages of the algorithms. At last,
we may get an expectation of development of the
motion planning for humanoid robot.
II. THE MODEL OF HUMANOID ROBOT
The accuracy of the Kinetic model directly decides
the actual movement characteristics of the humanoid
robot. The model designed more accurately, the results
of calculation will be more closed to the actual motion
characteristics. Otherwise, the calculation results may
be very different from the actual movement. Humanoid
robot is a complex structure and nonlinear system,
which has many degrees of freedom and redundant
multi-variable, so it is very hard and cockamamie to
establish a very precise model; Even if accurate models
can be established, the work of calculation will be very
2009 International Conference on Artificial Intelligence and Computational Intelligence
978-0-7695-3816-7/09 $26.00 © 2009 IEEE
DOI 10.1109/AICI.2009.335
322
heavy and it is not conducive to analyze problems and
unable to meet the requests of real-time control. In order
to establish a comparison of the kinetic equation which
can truly reflect the actual humanoid robot motion
characteristics, we need to be simplified the robot
platform. At present, the international usually use the
seven-link dynamic model (in figure 1).

Figure 1. Seven-links model
According to this kinetic model, we can be able to
design a better walking gait. However, it increased the
coordination of moving function, which increase the
degree of freedom in the mechanism, the direct
application of two-legged robot model of the seven-link
will not be able to describe more complex movements.
Thus, we should consider the effect of upper limbs and
head to the robot movement, basing on the seven-link
dynamic model of the humanoid robot. Then we may
establish a more appropriate model. Following the idea
of seven-link model, the humanoid robot between any
two adjacent joints of the connecting rod are seen as a
rigid body, and assuming that the quality of each rigid
body uniformly distributed and the center of mass at its
geometric center. Ignoring the effects of the hands to the
system, we define that its head, arm and forearm as the
center of mass which is located at the geometric center
of the rigid body, so the robot platform can simplify a
model of twelve-link which is connecting joints each
other. Some or all of the connecting 12 rods of the
coordination of movement complete the robotic
movement. Humanoid robot model of the connecting 12
robs are shown in figure 2 [4].

Figure 2. Twelve-link model
In order to further simplify the model for gait
analysis and design, we usually use the method that
there are no interaction between sagittal movement and
lateral movement, the sagittal movement and lateral
movement are shown in the figure 3 [5].

Figure 3. Models of two planes
III. THE METHODS OF MOTION PLANNING FOR
HUMANOID ROBOT
A. Path planning
Path planning can create an ordered sequence of
objective points that the robot must attain. Figure 4
present the path planning problems in the research [6].
Kffner and Chestnutt [7] introduced a classic gait
planning strategy called Game-Theoretic, the core
ideology is heuristic search, the principle of the strategy
is shorn if fig.5. The more about the principle we can
refer the journal [7]. But the off-line method often took
more time or produced longer plans. And now some
researchers have advanced the reformulation of HSP
323
(Heuristic Search Planner), which is based on the same
ideas and heuristic as HSP, but searches backward form
the goal rather than forward form the initial state. As a
result it can produce better plans, often in less time [8].

Figure 4. The basic path planning problem

Figure 5. Principle of heuristic search
Another path planning method is basing on the
sensor. It is a real-time planning method, mainly used in
dynamic noholonomic environment. It can solve the
problem of real-time and practicality. For example,
vision sensing is the most effective form for humanoid
robots understanding their environment, recognition of
the objects around them and navigating from one point
to another.
B. Gait planning
Because of the high DOFs and complex mechanical
structure, the precise dynamics model of the humanoid
is difficult to obtain. Therefore, intelligent techniques
are naturally a good choice in planning and control of
humanoid robots. The gait planning of humanoid robot
is achieved by the synergies between the off-line gait
planning and the on-line postural adjustment.
1) Fuzzy logic method
In designing a fuzzy algorithm, one usually needs to
determine parameters such as the defuzzzification
method, the membership functions, and the fuzzy rules.
There exist some standard defuzzification methods such
as the maximum criterion method, the mean of
maximum method, and the center of area method [9].
The fuzzy navigation algorithm is summarized as
follows:
Step 1: Given the initial and the desired final
configurations.
Step 2: Use the information measured by the sensors
to determine the parameter.
Step3: Use the fuzzy rules and the center-of-area
defuzzification method to determine the priorities.
Step 4: If has the highest priority, then the robot is
driven to move in the direction.
Step5: Repeat the procedure in Step 2 through Step 5
until the goal configuration is reached.
The above method will increase the accuracy of
self-positioning and environment recognition of the
robot. Another advantage is the development of fuzzy
navigation algorithm, which is simple and easy to
implement.
However, due to the changeable posture of humanoid
robot and the uncertainty environment, the fuzzy control
methods are extremely complex. So the current methods
can’t get an ideal experimental result.
2) Neural network method
Since neural network have an inherent capability of
learning and approximating nonlinear functions, it is
attractive to apply them in robot control systems for
modeling and compensation purposes. Intelligent
control systems can benefit from the advances in
artificial neural networks as a tool for on-line learning
optimization, and optimal policy making. The basic
concept of the neromodulatory neural network model is
schematically illustrated in figure 6 [10].
324

Figure 6. The concept of NN
It is observed that the proposed method has high
adaptability against environmental perturbations.
However, generally, the evolved neural controllers
could be fragile in inexperienced environments,
especially in real worlds.
3) Genetic algorithm method
It is considered that GA can be efficiently applied of
trajectory generation of the biped natural motion on the
basis of energy optimization as well as for walking
control of biped robots and for generation of
behavior-based control of these systems.
The hierarchical trajectory generation method
consists of two layers: one is the evolutionary
programming (EP) layer which optimizes the
interpolated configuration of biped locomotion robots.
The trajectory of biped is generated using ZMP stability
conditions. The GA optimization method is presented in
figure 7 [11].

Figure 7. GA process
The research of advanced GA is on the possible
optimization criteria that are path length, energy
consumption, smoothness and time needed for
movement [12]. The GA has shown very good results in
environments of different kind, but not in all possible
environments.
All in all, the extended GA works very promisingly.
Utilising the active search overcomes the drawbacks of
the earlier version of the path planning software. Thus,
the extended GA is suitable for universal usage now.
IV. DISCUSS
Each of the proposed paradigms has its own merits
and drawbacks. To overcome the drawback, certain
integration and synthesis of hybrid techniques
(symbiotic intelligence) are needed for efficient
application in humanoid robotics. For example, the
method of humanoid walking gait using GA-based
neural network is able to generate near-optimal walking
gait at any speed in feasible range [13]. GPS (Global
Positioning System) should be considered in the path
planning and vision is the most effective sensing form.
In addition to steering the robot, it can also play a key
role in achieving robot’s intelligence and its interaction
with outside world [14].
V. CONCLUSION
This paper has proposed the path planning and
typical gait planning algorithm, and conducted an
in-depth analysis and evaluation. Inadequate for a
variety of algorithms, we propose the use of hybrid
algorithms, GPS and advanced sensors to improve the
stability and capacity of the humanoid robot.
In spite of the intensive development and
experimental verification of various humanoid robots, it
is important to further improve their capabilities using
advanced hardware and arithmetic to make humanoid
robots more autonomous, intelligent and adaptable to
the environment and humans.
ACKNOWLEDGMENTS
This work was supported by Leading Academic
Discipline Project of Shanghai Municipal Education
Commission, Project Number: J51301.

325
REFERENCE
[1] Ill-Woo Park, Jung-Yup Kin, Jungho Lee, et al,
“Mechanical Design of Humanoid Robot Platform
KHR-3”, Proceedings of 2005 5th IEEE-RAS
International Conference on Humanoid Robots, 2005, pp.
321-326.
[2] Hideki Kondo,Yu Ogura, Hiroyuki Aikawa, et al, “Biped
Humanoid Robot Capable of Being Used as Human
Motion Simulator” , Experimental Robotics, 2008, pp.
523-531.
[3] Lakhmi C. Jain, Anas Quteishat and Chee Peng Lim,
“Intelligent Machines: An Introduction Studies in
Computational Intelligence” , Studies in Computational
Intelligence 2007, Springer Berlin / Heidelberg, Vol. 70,
pp. 1–9.
[4] Yang-Qing, “The research of the real-time motion
planning for humanoid robots”, Thesis of Master Degree,
National University of Defense Technology, 2005.
[5] T.Furuta, T.Tawara, Y,Okumura, “Design and construction
of a series of compact humanoid robots and development
of biped walk control strategies” , Robotics and
Autonomous Systems, 2001, Vol 37, pp. 81-100.
[6] S.M. LaValle, “Robot Motion Planning:A Game-Theoretic
Foundation” , Algorithmica, 2000, issue 26, pp. 430-465.
[7] Kufner J J, Nishiwaki K, lagallli S, et a1, “Motion
planning for humanoid robots”, Transactions in
Advanced Robotics, 2005, issue 15, pp.365-374.
[8] S.Biundo, M.Fox, “Planning as heuristic search: new
results”, Lecture Notes In Computer Science,
Springer-Verlag London, UK, 1999, Vol. 1809,
pp.360-372.
[9] Tsong-LI LEE, CHIA-JU WU, “Fuzzy Motion Planning of
Mobile Robots in Unknown Environments”, Journal of
Intelligent and Robotic Systems, Kluwer Academic
Publishers Hingham, MA, USA, 2003, Vol 37, Issue 2,
pp. 177 -191.
[10] Toshiyuki, Kondo, “Evolutionary design and behavior
analysis of neuromodulatory neural networks for mobile
robots control”, Applied soft computing, 2007, issue 7,
pp. 189-202.
[11] DUŠKO KATI´C and MIOMIR VUKOBRATOVI.
“Survey of Intelligent Control Techniques for Humanoid
Robots”, Journal of Intelligent and Robotic Systems,
Kluwer Academic Publishers Hingham, MA, USA, 2003,
Vol 37, issue 2, pp.117-141.
[12] M.Gemeinder,M.Gerke, “GA-based path planning for
mobile robot systems employing an active search
algorithm”, Applied Soft Computing, 2003, issue 3, pp.
149-158.
[13] Zhe Tang, Changjiu Zhou and Zengqi Sun, “Humanoid
walking gait optimization using GA-Based neural
network”, LNCS, Vol 3611, 2005, pp. 252-261.
[14] M.Fikret Ercan and Changjiu Zhou, “Vision based
motion planning of humanoid robots”, Lecture Notes in
Control and Information Sciences, 2004, Vol 299, pp.
212-227.
[15] Xia Zeyang, Chen Ken,Xiong Jing and Fu chenglong,
“Advances in research on motion planning for humanoid
robots”, High Technology Letters, 2007, Vol 17, issue 7,
pp. 1092-1099. (in Chinese).



326