RoboSapien Based Autonomous Humanoid Robot

embarrassedlopsidedΤεχνίτη Νοημοσύνη και Ρομποτική

14 Νοε 2013 (πριν από 3 χρόνια και 6 μήνες)

59 εμφανίσεις

RoboSapien Based Autonomous Humanoid Robot


Nick Repka


University of Calgary, Department of Computer Science

2006


Interim Progress Report


Abstract:




Most human robot interaction labs today do not have access to cutting edge
humanoid robots due to
their current scarcity and high cost. Capable toy
-
based
humanoids do exist, such as the RoboSapien, but they lack the necessary computing
horsepower to be useful to researchers. The aim of this project is to follow and adapt the
methods of previous multi
-
a
gent systems researchers and amateur robot enthusiasts to
create a RoboSapien based autonomous robot that can be used for further human robot
interaction research. The attempt will be made to add the CPU power of a handheld
computer and to close the contro
l loop by adding external sensors. While it is true that
most of the ideas covered have been attempted before there is as yet no comprehensive
methodology for building cheap and effective research tools on this platform. Without a
detailed method for produ
ction of these tools most human robot interaction researchers,
lacking the ability to build humanoid robots themselves, simply do without them. If clear
instructions existed to build these machines more human robot interaction researchers
could have access

to humanoid robots. The final goal then is to provide this detailed
methodology; complete and fully documented, so that RoboSapien based research
platforms can be easily and inexpensively produced.
Introduction:



This report covers the
considerations a
nd activities carried out
for this project between February 9 and
March 15 2006. The three major
objectives covered in this time frame
were to be the reviews of required
hardware and software and the
acquisition of materials. The hardware
review covers two

major items, the PDA
computer and the camera. The software
review covers the software required to
make the hardware run and tools to
modify or create new programming.


In covering the material
acquisition phase we will discuss the
selected hardware and s
oftware
components in depth. Reasoning for
these selections will be explained and
known issues will be covered. Again the
materials will be broken down into
hardware and software categories for
this discussion.


Hardware Review:



The hardware for this pro
ject
consists of two major items: the PDA
and the camera. Another consideration
was the robot platform itself but this was
relatively minor for reasons that will be
explained later on.


PDA Choice



The PDA’s considered for usage
included line
-
ups by Hewl
ett Packard,
Toshiba and Dell. A soccer playing
RoboSapien
1

built by a Freiburg
University group used a Fujitsu
-
Seimens
Loox720 PDA. This PDA unfortunately
is no longer on the market and cannot be
acquired reliably in the used electronic
market. Our aim wa
s to select a hand
-
held with capabilities similar to this
PDA to enable us to reuse code provided
by this group.


Toshiba’s machine was initially
attractive, as it should fully support code
available from Freiburg University’s
Human Robot Interaction team
.
Combined with a camera also known to
work with the code the basic software
proof of concept scheduled for after this
report would be a plug and play affair.
Toshiba’s current model (e830)
2

was
probably the best overall fit for this
project.


Despite this

there were issues
with the Toshiba PDA. They are not
particularly easy to come by. Toshiba at
this time is manufacturing only one
model of PDA, the e830, and appears to
be repositioning itself outside of the
PDA market. Given that in the future
others at
the I
-
lab may wish to replicate
this work in this project it makes little
sense to base it on a machine that
appears to be disappearing. A functional
problem was it’s larger than average
mass, around 200g, that might have



1

Behnke, Muller, Schreider, “Playing Soccer
with RoboSapien”, University of Freiburg,
Institute for Computer Science, 2005.

2

Toshiba Corporation, e830,
http://www.toshiba.ca/web/product.grp?lg=en&s
ection=1&group=22&product=36
90&category=

interfered with the RoboSapien’s g
ait.
The most pressing reason it wasn’t
considered seriously when choosing
hardware was its high cost of around
$600.


Hewlett
-
Packard is probably the
largest remaining maker of PDAs and it
would be less than thorough to ignore
their offerings. Their IPAQ
3

handheld
computer lineup offered a number of
possible solutions. Among those
considered were the IPAQ “hx” models
numbers: 2190, 2490, and 2570. All
weighing in around 165g, actually less
than the Loox720 (170g) used by the
Freiburg group they could be e
asily
integrated into the robot.


The 2190 in the end was
eliminated due to having only 60% of
the Loox720’s processing power. The
other two, while capable performers,
lacked the necessary power to cost
ratios. H.P. tends to market itself as a
premium br
and and this distinction
comes with a price jump. In the end the
HP PDAs came with a sticker price $150
above competitive PDA models by other
manufacturers. In the end H.P. was ruled
out due more to being overpriced than
anything else.


Dell Computers bett
er known as
a desktop PC manufacturer also has a
solid and well supported PDA line
-
up.
The Axim
4

model line is already in its



3

Hewlett
-
Packard, Handhelds,
http://h10010.www1.hp.com/wwpc/ca/en/sm/W
F02a/12139188
-
12139190
-
12139190.html?a=/

4

Dell Computers, Dell Axim Handhelds,
http://www1.ca.dell.com/content/products/catego
ry.aspx/handhelds?c=ca&cs=CABSDT1&l=en&
s=bsd

4
th

or 5
th

generation and is comparable to
any other line of PDAs in power and
quality. The current model generation is
designated

the Axim x51. This line
-
up
has all the features expected of modern
PDAs, and matches the other line
-
ups in
all things save one; cost. The Axim x51
models are on average $100 to $150 less
expensive. Given our mandate to create
an inexpensive robot research

platform
the low cost of the Dell was persuasive.
The Axims also weigh in between 167
and 175g making them ideal for
implanting in the RoboSapien.


The one downside was that the
current Axim models are not available
with Microsoft Pocket PC 2002 or
2003
5
.

This creates an issue with camera
compatibility. A vital concern especially
considering this was the PDA acquired
for this project and the issue was
unknown to us until a week after getting
the unit.


Camera Choice



When considering expansion slot
camera
s for PDAs there are a couple
formats to consider. SDIO and CF. CF
or Compact
-
flash is considered the better
choice. CF offers more choice in
cameras and often better support from
manufacturers. A number of cameras
were looked at and were evaluated on
comp
atibility, resolution and cost.





5
Microsoft Corporation, Mobile
Solutions, Mobile Applications and handheld
devices from Microsoft Windows Mobile,
http://www.microsoft.com/windowsmobile/defau
lt.mspx



The first choice was the Lifeview
Flycam
-
CF
6
. While the most expensive
of the lot it has far superior resolution
and support. The highest resolution
available for this camera is a full 1.3
megapixels. Of the cameras consid
ered
only this one is confirmed to have
Windows Mobile 5.0 drivers in
development. This is especially
important given the PDA chosen for the
robot uses this OS. Correspondence with
the company indicates that these drivers
may be ready before the end of thi
s
project and if so the full capability of the
hardware will be demonstrable.



The two other cameras were the
Pretec Compact Camera
7

and the Veo
Photo
-
Traveler
8
. Both are cheaper than
the FlyCam but offer a maximum
resolution of only 640X480. Also neithe
r
company has offered any indication that
they at this time intend to support
Windows Mobile 5.0. Given these two
points neither camera can offer much
scalability or adaptability going forward.


Robot Platform


While always expected that a
RoboSapien would

be used there was
some discussion about the version to be
used. A version1 RoboSapien was
already on hand for use a RS V2 could
be acquired easily enough. In the end it



6

LifeView Inc, FlyCam CF series,
http://www.lifeview.com/usa/html/products/IA/fl
ycamcf.h
tm

7

Pretec Electronics Corporation, Compact
Camera,
http://www.pretec.com/product/digital_imaging/
CompactCamera.htm

8

Veo International, Veo PhotoTraveler for
PocketPC,
http://www.veo.com/Traveler_PocketPc/default.
asp

was decided that the modifications could
be moved rather easily from one
platform to
the next and having the latest
version was unnecessary.

Software Review:



Here we will locate and assess all
software useful to the project. Included
in this are the mobile computer’s
operating system, the programming
environment, and infrared remote
soft
ware for the robot interconnection
and any useful code from the Freiburg
group.


Operating System &
Programming Environment



Given our use of the Dell Axim
x51 we are constrained to using
Windows Mobile 5.0. On this hardware
though it is considered to be

a significant
step forward from the previous OS
version. At this point it is not easily
possible to get a previous Windows CE


OS installed on this hardware. It follows
that all software we wish to use must be
compatible with Mobile 5.0. In order to
ensur
e the longevity of the platform
being developed here Mobile 5.0 is a
good choice due to its adoption by most
new models of PDA.



Microsoft has strongly supported
all its CE versions with SDKs and
Mobile 5.0 is no exception. Microsoft
offers a stand alone

editor and emulator
as well as an addition that works within
the MS.net development framework.






Windows CE Operating systems do not

have a
standardized naming convention. Versions 1
-
2
are sold under the Windows CE name, versions
3
-
4 are sold under the PocketPC name and
version 5 is sold as Windows Mobile 5.0


IR Remote Software



Using third party IR remote
software should save time when
implementing the control program.
Ultramote
9

produced by Cardonware has
been se
lected for this purpose. It boasts
several advantages over similar
software. First it is among the least
expensive. Second it is far more
customizable than other similar software
and thus more adaptable to our purpose.
Third, compatibility with the Axim x5
1
is fully assured. Finally this software
was used successfully by the Freiburg
group and we can use their code to
operate it without issue.



It is not without issue however;
using any remote software has the
limitation of not being infinitely
adaptable.

While all the functionality of
the RoboSapien V.1 should be accessible
with this software it may not be possible
to interface with the entire functionality
of a more current model of RoboSapien.


Freiburg Group Code



The code left over from the
Freiburg

University group when creating
a soccer playing RoboSapien is
extremely useful. The code base is
properly encapsulated and very well
documented. Even better is that it works
with proven and readily available MS
development software. The largest
benefit is

that the code provides basic
access to almost the entire functionality
of the RoboSapien and photo
-
capture



9

CardonWare, Ultramote Universal Remote
package for Pocket PCs,
http://www.u
ltramote.com/

code for a couple of cameras. One of
these cameras is the Lifeview FlyCam
-
CF. Obviously this will make the
upcoming development steps much
easier if

this code can be used.


Challenges:



Thus far only one serious issue
has surfaced that will cause immediate
problems. This issue is that of
compatibility of the FlyCam
-
CF camera
with Windows Mobile 5.0. At this point
we lack the drivers to operate the ca
mera
properly. While they are in development
Lifeview has stated that they could come
out anytime within the next one to two
months. Obviously given the limited
timeframe of this project it may be
impossible to demonstrate the vision
capability of the plat
form by the
presentation date. This would be a
serious failing given the goal of using
external sensors to allow the robot to
interact with the outside world.



Therefore instead of idly waiting
for the drivers to be released an attempt
will be made to clo
se the control loop by
integrating input from the PDA
microphone. The specifics of this have
not yet been decided but a robot activity
based on claps or whistles should be
possible with a reasonable amount of
effort. In the best case this could lead to
bei
ng able to demonstrate behavior
based upon both sound and vision.


Conclusion:



The project is moving ahead very
well despite the compatibility problems
with the camera. It was and remains the
only foreseeable problem of any
magnitude. All needed material
s have
been acquired and the next steps in
development can be taken immediately.
The hardware modification phase has
already begun and is well ahead of
schedule. As long as no further serious
issues arise there should be no problem
in completing the projec
t within the
term.



Appendix: Tables and Photographs


1. Tables

Figure 1.1 PDA comparison chart

Manufacturer

















Model

Processor

CF Version

OS

RAM

Mass

Price

Toshiba

















E830

Intel PXA 520MHz

II

PPC 2003

128MB

200g

$600

















Hewlett
Packard

















IPAQ hx 2190

Intel PXA 312MHz

II

Win Mobile 5

64MB

164.4g

$400



IPAQ hx 2490

Intel PXA 520MHz

II

Win Mobile 5

64MB

164.4g

$550



IPAQ hx 2750

Intel PXA 624Mhz

II

PPC 2003se

128MB

164.4g

$650

















Dell

















Axim X51
-

416

Intel PXA 416MHz

II

Win Mobile 5

64MB

167g

$300



Axim X51
-

520

Intel PXA 520MHz

II

Win Mobile 5

64MB

167g

$400



Axim X51v

Intel PXA 624MHz

II

Win Mobile 5

64MB

175g

$500

















Fujistu
-

Siemens

















Loox 720 *

Intel PXA 520MHz

II

PPC 2003

128MB

170g

N/A

*No longer in production



Figure 1.2 Camera Comparison Chart

Manufacturer

Model

Resolution

Compatibility

Price

LifeView

FlyCam
-
CF

1.3MP

PPC 2000
-
2003se (WM5 in development)

$130

Pret
ec

Compact Camera

640X480

PPC 2002

$90

Veo

Photo Traveler

640X480

PPC 2002 Palm OS

$100



2. Photographs


Figure 2.1 Dell Axim x51 and Lifeview FlyCam
-
CF 1.3 hardware.



Figure 2.2 Axim and FlyCam mounted on RoboSapien



Bibliography:


1. Behnke, Mul
ler, Schreider, “Playing Soccer with RoboSapien”, University of Freiburg,
Institute for Computer Science, 2005.


2. Toshiba Corporation, e830,
http://www.toshiba.ca/web/product.grp?lg=en&section=1&group=22&product=3690&ca
tegory=


3. Hewlett
-
Packard, Handhe
lds,
http://h10010.www1.hp.com/wwpc/ca/en/sm/WF02a/12139188
-
12139190
-
12139190.html?a=/


4. Dell Computers, Dell Axim Handhelds,
http://www1.ca.dell.com/content/products/category.aspx/handhelds?c=ca&cs=CABSDT1
&l=en&s=bsd


5. Microsoft Corporation, Mobile So
lutions, Mobile Applications and handheld devices
from Microsoft Windows Mobile,
http://www.microsoft.com/windowsmobile/default.mspx


6. LifeView Inc, FlyCam CF series,
http://www.lifeview.com/usa/html/products/IA/flycamcf.htm


7. Pretec Electronics Corpor
ation, Compact Camera,
http://www.pretec.com/product/digital_imaging/CompactCamera.htm


8. Veo International, Veo PhotoTraveler for PocketPC,
http://www.veo.com/Traveler_PocketPc/default.asp


9. CardonWare, Ultramote Universal Remote package for Pocket PCs
,
http://www.ultramote.com/