Annotated Bibliographyx - The-A-List

madbrainedmudlickΤεχνίτη Νοημοσύνη και Ρομποτική

20 Οκτ 2013 (πριν από 3 χρόνια και 11 μήνες)

63 εμφανίσεις

Hunckler

1

Eddie Hunckler

Advanced Research

Annotated Bibliography

9
-
26
-
2012


Morimoto, Jun, Gordon Cheng, Christopher G. Atkenson, and Garth Zeglin. "A Simple
Reinforced Learning Algorithm For Biped Walking."
IEEE
.

(
2004
)
: 3030
-
035.
Print.


In this study, a model
-
based learning algorithm is investigated for appropriately
placing
the

swinging leg. Several different methods for this learning model are used.

The first is
a model based on estimation. The second is a model
-
based method with a rew
ard system
for reinforcing learning ability. Through this simulation successful walking patterns were
learned or discovered through these trial processes.


Garcia, Mariano, Anindya Chatterjee, Andy Ruina, and Michael Coleman. "The Simplest
Walking Model: S
tability, Complexity, and Scaling."
Journal of Biomechanical
Engineering
.

Volume
120 (1998): 281
-
88. Print.


This article is a study of the dynamics of human walking in a simplified form to gain
foundational understanding for future more complex models. A

computer model of a
simple set of mass
-
less legs with frictionless hips was used to develop the walking
patterns and relationships. The hip mass in the simulation was much greater than the foot
Hunckler

2

mass. Several relationships were studied including a relati
onship between the slope of
the ground and the stance angle at a fixed point. They also researched the relationships
between the leg angles and the
time.


Hirai, Kazuo, Masato Hirose, Yuji Haikawa, and Toru Takenaka. "The

Development of
Honda Humanoid Robot."
IEEE
.

(1998): 1321
-
326. Print.


This article encompassed a large amount of Honda’s humanoid research. The first
research that they mention is modeling the walking gait of their robots after human
walking gaits. The ro
bot was built to model a human, but that was more difficult than
originally expected. The mechanical design was simplified to include the minimum
requirements for walking. The humanoid is equipped with an extensive amount of
sensors including gyroscopes
and accelerometers that provide live feedback on the
robot’s position allowing for
self
-
corrections
.


Huang, Qiang, Kazuhito Yokoi, Shuuji Kajita, Kenji Kaneko, Hirohiko Arai, Noriho
Koyachi, and Kazou Tanie. "Planning Walking Patterns for a Biped Robot."
IEEE
.

(2001): 280
-
89. Print.


In this article the authors describe a process for planning walking pattern of robots. They
use a method called zero moment point (ZMP). “The ZMP is defined as the point on the
ground about which the sum of all the moments of

active force equals zero.” The
researchers mapped a derived trajectory for the ZMP and the derived the hip positions to
Hunckler

3

achieve that trajectory. A biped robot was built to test the
efficiencies of the gait
developed through this method. This robot util
ized dynamic balance in its walking
sequences.


Kun, Andrew L., and W. Thomas Miller, III. "Adaptive Dynamic Balance of a Biped
Robot Using Neural Networks."
Robotics Laboratory, ECE Dept., University of
New Hampshire
.

1
-
5. Print.


Bipedal robots that rely on static balance for their walking abilities are slow, inefficient,
and relay on large feet to remain stable.
The goal of this research was to design and build
an adaptive dynamic balance scheme based on neural networks.
The con
trol was a
combination of preplanned
control as well as adaptive controls. The neural network
“training” allowed the robot to adapt and learn a better walking sequence. The biped
was able to reach 100 steps per minute and stop and start on command.


Manoonpong, Poramate, Tao Geng, Thomas Kulvicius, Bernd Porr, and Florentin
Wörgötter. “Adaptive, Fast Walking in Biped Robot under Neuronal Control and
Learning.”
PLOS Computational Biology
. Volume 3.
(
2007
)
: 1305
-
1320. Print.


The authors investigate a t
heory based on neuronal control of walking as well as

balance
associated with walking
. The research tries to mimic walking in a human by utilizing
multiple levels of control form basic actuator control to complex balance analysis.
Mechanisms for walking
must be at a level of reflexes in humans. This robot was
Hunckler

4

modeled off of a human design mimicking the proportional dimensions and weight
distributions. As a test to see the similarities of this robot to a human, a Froude number
was calculated to see how w
ell they corresponded.


Mayagoitia, Ruth E., Anand V. Nene, and Peter H. Veltink. "Accelerometer and Rate
Gyroscope Measurement of Kinematics: An Inexpensive Alternative to Optical
Motion Analysis Systems."
Journal of Biomechanics
.

(2002): 537
-
42. Print.


This article details the option of using accelerometers and rate gyroscopes to measure
kinematics of the sagittal plane. It compares the accuracy of these sensors to advanced
optical motion analysis systems. It states that this would be a suitable replac
ement for the
optical system except at high speed. This is due to strong jerks upon contacting the
ground with the foot. This test was performed on 10 males between the ages of 23 and
27 years.


Miller, W. Thomas, III. "Real
-
Time Neural Network Control o
f a Biped Walking Robot."
IEEE

(1994): 41
-
48. Print.


“The project goal here is to develop biped control strategies based on a hierarchy of
simple gait oscillators, PID controllers, and n
e
ural network learning, that do not require
detailed kinematic or dyn
amic models.” This project focuses on 10 degrees of freedom
biped robot. This robot uses dynamic balance

which is when the projected center of mass
is outside the area inscribed by the feet. In this method, the robot is falling, and the foot
Hunckler

5

needs to mo
ve into place to break the fall. This project used a neural network to learn
and develop the walking patterns for the biped.


Raibert, Marc, Kevin Blankespoor, Gabriel Nelson, and Rob Playter. "BigDog, the
Rough
-
Terrain Quadruped Robot."
The International

Federation of Automatic
Control
.

(2008): 10822
-
0825. Print.


The goal behind developing the BigDog was to create a walking robot that was self
-
contained, but addressed problems such as being self
-
contained and traversing rough
terrain.
BigDog is a quadru
ped robot based off of original MIT Leg Laboratory ideas of
simple dynamic walking robotic solutions. BigDog has over 50 sensors. The onboard
computer performs the computations for the high and low level control computations.
This system is capable of cl
imbing up and down slopes. When climbing up, it leans
forward, and when walking down, it leans backwards. When a slope is greater than 45
degrees the robot adjusts it gait as well to accommodate for the slope.