<<
, >>
, Title
, Contents
, Index
8. MINUTES FROM MEETINGS
These
are impromptu meetings which are called when it is urgent to discuss certain
issues.
|
venue: | UWSN KUG25 (Big Shoe Box) |
|
time: | 29 July 1997, 10:00am to 11:00am |
|
present: | David, Michael, Mik and Stephen |
Put
off deciding this until next week, when we can all be here and we're all more
sure of our tute times, etc.
The
following list shows what needs to be done to finish the project. They are
shown with most important things first.
1. Socket interface semantics
2. Environment simulator
3. Real environment window
4. BOTSsim window (main form)
5. Robot ranking (done in environment)
6. Saving and loading runs
7. Rule set window
8. Super TED (like Big TED only better)
9. Testing/debugging
10. Cosmetic stuff (eg: multiple bot icons)
11. Tweak help and user doccos
The
following ideas were suggested but not discussed much:
- Implement rule set editing feature
- Simulate more realistic sensors
- Go further down the user wish list[2]
- Implement full gerbil support[3]
- Standardise user interface
- Speed the whole thing up
- Use HTML for help
CVS,
the version control system that replaces tokens,will need to be discussed
further. In general we are in favour of it.
This
is in our timetables. Please be there, so we can find out details about the
following:
- Who our supervisor is
- Who our client is
- We need to discuss who does what on the wish list[4]
- Discuss Mark's CVS idea3
- Steve will unlock our account[5]
- Steve will get our final report for Autumn back from George
- We'll all be at that meeting tomorrow
|
venue: UWSN KUG50 |
|
time: | 3 September 1997, 12:30pm |
|
present: | Colin, David, Michael, Mik and Stephen |
- Once again new code was overwritten by old code, basically undoing some of
the work we did yesterday.
- There was some discussion about whether the problem was the procedure
(reading e-mail, posting e-mail, downloading, doing work, uploading, posting
e-mail) or whether people just weren't following the procedure.
- It was probably a bit of both, although these sorts of problems could be
substantially reduced through the use of CVS, discussed below.
- Until CVS is implemented, we will adhere to the following, slightly
revised, procedure:
- Check e-mail to see if anybody is working on the files you want to work
on. If not:
- Post e-mail to the list saying which files you intend to be working on.
- Download the files from the file server.
- Make the changes to the files you are working with, and no other
files.
- Upload only the files you have changed--do not do a multiple
file upload!
- Post e-mail to the list saying you have finished working on the files, and
detail any changes you have made.
Before
CVS can be implemented, we need:
- A CVS administrator--Mark was unanimously appointed.
- Some support utilities--Mark will ensure these are in place.
- Decisions about how the project can be split. This is necessary so
programmers can "check in" and "check out" various components of the program.[6] The following draft has been proposed:
- TED
- brain
- Genetic Programming
- Evaluation
GUI
- Environment window
- Bot window
- Hits histogram
- Run information dialogue
- Help/about
environment
There was discussion about how
urgent it was to implement CVS, which lead to the next agenda item.
Although
implementing CVS will speed up the project (because we won't be reinventing the
wheel as much), it is probably better to wait until next week to do this. There
are several reasons for this:
- This gives Mark time to become more familiar with the product
- This gives Mark time to chase up the required support software
- It is more natural to revise the components of the project at the start of
version 2, rather than in the final hours of developing version 1
- We need drive space to support CVS, which we can't easily free up until
after we move to the new project lab next week
- There are more important things to do first
After a brief
discussion, the current priorities were deemed to be:
1. Report IV--Mark still needs to do some work for this, especially the current
Gantt Chart
2. Version 1--Some of the changes made yesterday need to be redone because of
the overwritten code
3. Implementation of CVS--This can be done next week, after Report IV and
Version 1 are finished, and everyone has their nasty exams out of the way.
These
are weekly meetings where we discuss groupwide issues. They mostly due to start
at 9:15am every Tuesday, and every group member is supposed to be present at
them.
This
meeting was cancelled because of lecture clashes.
|
venue: | UWSN KY238 (Project Room) |
|
time: | 5 August 1997, 9:15am |
|
present: | Colin, David, Mark, Michael, Mik and Steve |
- Environment
- "about 78% done"
- still need to do: block drop, fitness level update, simulation, and a few
niggly things.
GUI
- 80% done
- waiting for other components (Environment, Comms)
- still need to do:
- Multiple Environment windows
- Bot Windows
- BOTSsim window
- saving & loading run info, brains, runs...
Brain
- "100% coded"
- needs testing
- waiting for all other components
- needs support for getting & setting rule sets
TED
- "97% done"
- fix loss of toolbar bug due to scrolling.
Everyone
will do coding as they see fit to ensure the project is complete next fortnight.
|
venue: UWSN KY238 (Project Room) |
|
time: | 12 August 1997, 9:15am |
|
present: | Colin, David, Mark, Michael, Mik and Steve |
- To date nobody has been to any of these meetings because of the timetable
change and other commitments. Apparently last week's meeting was cancelled.
- Discussed whether we know what's been happening at these meetings. We have
no idea. Mik said "so for all we know the subject's been cancelled"--Colin said
"no, because George would know that".
- Steve can go to this week's project meeting (13 August).
- Mik will be able to go to project meetings from 20 August to 17 September.
- Reviewed what happened at the last client meeting--see section 8.4.1 on
page 163.
- Everyone can go to this afternoon's client meeting except Mik.
- Steve will take minutes at this afternoon's client meeting.
Vigorous
discussion about what to do, in order, to finish the project. Dates and the
people responsible have been added for items up to the next group meeting.
#
|
Description
|
Person responsible
|
Should be done by
|
1.
|
Start
Run coding--Run Brain (GUI à GUI)
|
Steve
|
12th
|
2.
|
Start
Run coding--Init Brain procedure (GUI à Environment à Brain)
|
Colin,
Steve
|
12th
|
3.
|
Clone
worlds environment code
|
Michael
|
13th
|
4.
|
Start
Run coding:
* Clone Worlds
* Assign Bots
* Evaluate Bots
|
Colin,
Michael, Steve
|
14th
|
4.
|
Robot
World parser message
|
Colin
|
14th
|
5.
|
Drop
block environment function
|
Michael
|
15th
|
6.
|
Fitness
level update
|
Michael
|
15th
|
7.
|
Comms
debug window
|
Steve
|
15th
|
8.
|
Support
for getting and setting rule set
|
David,
Mark
|
15th
|
8.
|
Rule
set parser messages
|
Colin
|
15th
|
9.
|
Rule
set and bot windows
|
Mik
|
17th
|
10.
|
Parser
error message
|
Colin
|
19th
|
11.
|
Capability
for multiple environment windows
|
|
|
12.
|
The
real environment window
|
|
|
13.
|
Loading
and saving runs
|
|
|
14.
|
Parser's
okay message
|
|
|
15.
|
Hit-o-gram
|
|
|
16.
|
Robot
ranking
|
|
|
17.
|
BOTSsim
(main) window
|
|
|
18.
|
TED
scroll bar bug fix
|
|
|
19.
|
Testing
and debugging
|
|
|
20.
|
Update
online help
|
|
|
21.
|
Update
user documentation
|
|
|
We
discussed Report IV, and when it's due.
We will ask George this afternoon if we can hand in report IV, together with
the completed (original) project, at the end of week 5.
|
venue: UWSN KY238 (Project Room) |
|
time: | 19 August 1997, 9:15am |
|
present: | Colin (Late), David, Michael, Mark (Late), Mik, Stephen. |
We'll
aim to be here at 9:00am.
Get
drinks, have toilet breaks, and deal with students before 9:15am, and
stay at the meeting venue until 10:50am. Doing other things is okay, as
long as we're all available to be yelled at.
We'll take breaks occasionally, but at the same time.
Let's
take our breaks at the same time, and not go for longer that half an hour
without a break.
- Everyone posts agenda items to the sad account by noon on the day before
the meeting
- Mik will prepare the agendas in text format and put it in the
/documentation/agendas directory
- People can look at the agendas online or whatever, and even print them out
if they're still in the paper age
- Carolyn says we need to have Gantt Charts for every meeting.
- Discussion about whether George wants a "to do" list or a Gantt Chart.
Since a "to do" list is just a subset of a Gantt Chart, we'll stick to doing
those.
- We'll convert the "to do" list into a MS
Project(TM) file, and Steve will update it every
week after meetings.
- Needs to be on the Web at the end of week 6.
- We'll aim to have doccos done by the end of week 5.
- Mik will put it together over the week 5 weekend.
- Colin will convert it to HTML early in week 6.
- Mik will clean up the HTML code on Wednesday on week 6.
The
web page needs:
- General cleaning up
- General updating
- Warnings before all the naughty bits
- Tone down Franz slander
- Report IV
Mik will do all this in time with Report 6.
- Ravi has posted e-mail to our group about backing up several times over
the last few weeks.
- All we need to do is make sure a copy of our code is on our accounts,
which should be saved.
- David will check all the ACL stuff is backed up, and will tell Ravi to go
for it.
These
are subject to change when Mark works out CVS, which should be done in time for
the new project.
- Make a reasonable effort to find out if somebody else is working on that
code, ie: check your mail
- Post a message to the list about what changes you are going to make
- Download original code
- Make your changes
- Upload your changes
- Post a message to the list that you have done the changes
Discussion
about what to do, in order, to finish the project. Dates and the people
responsible have been added for items up to the next group meeting.
#
|
Description
|
Person responsible
|
Was Due
|
Is it Done?
|
Now Due
|
0.
|
Hooking
up for testing purposes.
|
most
of us
|
|
|
asap
|
1.
|
Start
Run menu function--Run Brain (GUI à GUI)
|
Steve
|
12th
|
Yes.
|
|
2.
|
Start
Run menu function--Init Brain procedure (GUI à Environment à
Brain)
|
Colin,
Steve
|
12th
|
Stalled
due to bug.
|
|
3.
|
Clone
worlds environment code
|
Michael
|
13th
|
Yes.
|
|
4.
|
Start
Run menu function:
* Clone Worlds
* Assign Bots
* Evaluate Bots
|
Colin,
Michael, Steve
|
14th
|
Mostly:
* Yes
* No.
* Yes.
|
19th
|
4.
|
Robot
World parser message
|
Colin
|
14th
|
Need
step 4.
|
|
5.
|
Drop
block environment function
|
Michael
|
15th
|
Yes.
|
|
6.
|
Fitness
level update
|
Michael
|
15th
|
Yes.
|
|
7.
|
Comms
debug window
|
Steve
|
15th
|
Yes.
|
|
8.
|
Support
for getting and setting rule set
|
David,
Mark
|
15th
|
Yes.
|
|
8.
|
Rule
set parser messages
|
Colin
|
15th
|
Need
step 9.
|
20th
|
9.
|
Rule
set and bot windows
|
Mik
|
17th
|
No.
|
19th
|
10.
|
Parser
error message
|
Colin
|
19th
|
No.
|
20th
|
11.
|
Capability
for multiple environment windows
|
|
|
|
|
12.
|
The
real environment window
|
Steve
|
|
|
22nd
|
13.
|
Loading
and saving runs
|
Michael
|
|
|
22nd
|
14.
|
Parser's
okay message
|
Colin
|
|
|
22nd
|
15.
|
Hit-o-gram
|
Steve
|
|
|
22nd
|
16.
|
Robot
ranking
|
Mik
|
|
|
22nd
|
17.
|
BOTSsim
(main) window
|
Steve
|
|
|
22nd
|
18.
|
TED
scroll bar bug fix
|
|
|
|
|
19.
|
Testing
and debugging
|
All
of us
|
|
|
2nd
|
20.
|
Update
online help
|
Steve
|
|
|
2nd
|
21.
|
Update
user documentation
|
Steve
|
|
|
2nd
|
|
venue: UWSN KY238 (Project Room) |
|
time: | 19 August 1997, 9:15am |
|
present: | Colin, Mark, Mik, Stephen (Late). |
- Nobody has done the Gantt chart this week. Mark will update this week's to
do list to a Gantt chart, and we will decide next week how it will be updated
after future meetings.
- Ravi's backup concerns have been dealt with.
- Mark and Colin will definitely be there. Mik may be there. Steve
probably won't be there.
- Rocky asked us to make slides of pseudocode, etc., GA, etc. Colin and Mark
will do this.
- We told Mark off for this.
- We've agreed to allow the user to set the seed for the random number
generator, which is better than hard-coding it.
- Some debate about the name of the seed variable--agreement on
seed in Delphi and *seed* in ACL.
- Mark will update the protocol on the brain side.
- Steve will update the runinfo dialog (sic).
We
need the following:
- Bug list--everyone make their own and we'll thrash it out next week
- Mark will write something about brain speed
- About the three components:
- Progress since last semester, as a list of chronological, dated points
- What stuff still needs to be done (by the following week), as a single
paragraph
David will write about the brain
- Michael will write about the environment
- Mik will write about the GUI
Steve
has some cool teddy bear icons. He will implement the best one in time for open
day.
We
convinced Colin to enable error handling at the environment/GUI end, even
though they're not actually sent.
- David said in his e-mail that prefers to use sequence numbers, to handle
two processes working at once.
- Colin prefers to use semaphores.
- Mark says that two processes should run happily together anyhow.
- No conclusion has been reached at this point.
- Mark will update this week's to do list to a Gantt chart.
- Colin and Mark will make slides for Open Day.
- Mark will update the protocol on the brain side w.r.t. random number seeds.
- Steve will update the runinfo dialog (sic) w.r.t. random number seeds.
- Mark will write something about brain speed for the report.
- David will write about the brain for the report, following the convention
documented above.
- Michael will write about the environment, following the convention
documented above.
- Mik will write about the GUI, following the convention documented
above.
Vigorous discussion about what to do, in order, to finish the
project. Dates and the people responsible have been added for items up to the
next group meeting.
#
|
Description
|
Person responsible
|
Was Due
|
Is it Done?
|
Now Due
|
0.
|
Hooking
up for testing purposes.
|
most
of us
|
asap
|
Yes.
|
|
4.
|
Start
Run menu function:
* Clone Worlds
* Assign Bots
* Evaluate Bots
|
Colin,
Michael, Steve
|
19th
|
Yes.
|
|
4.
|
Robot
World parser message
|
Colin
|
|
Yes.
|
|
8.
|
Rule
set parser messages
|
Colin
|
20th
|
Yes.
|
|
9.
|
Rule
set and bot windows
|
Mik
|
19th
|
Yes.
|
|
10.
|
Parser
error message
|
Colin
|
20th
|
No.
|
26th
|
12.
|
The
real environment window
|
Steve
|
22nd
|
Mostly.
|
26th
|
13.
|
Loading
and saving runs
|
Michael
|
22nd
|
No.
|
31st
|
14.
|
Parser's
okay message
|
Colin
|
22nd
|
Need
#13.
|
2nd
|
15.
|
Hit-o-gram
|
Steve
|
22nd
|
Mostly.
|
2nd
|
16.
|
Robot
ranking
|
Mik
|
22nd
|
Started...
|
26th
|
17.
|
BOTSsim
(main) window
|
Steve
|
|
|
22nd
|
18.
|
TED
scroll bar bug fix
|
|
|
|
|
19.
|
Testing
and debugging
|
All
of us
|
|
|
2nd
|
20.
|
Update
online help
|
Steve
|
|
|
2nd
|
21.
|
Update
user documentation
|
Steve
|
|
|
2nd
|
22.
|
Concept
of a generation (time steps)
|
Steve
and Colin
|
|
|
26th
|
23.
|
Sending
fitnesses
|
Michael
|
|
|
29th
|
24.
|
Automation
of animation
|
Steve
and Colin
|
|
|
26th
|
25.
|
Ability
to play midi files
|
All
of us
|
|
|
|
|
venue: UWSN KY238 (Old Project Room) |
|
time: | 2 September 1997, 9:15am |
|
present: | Colin, David, Michael (late and wet), Mik, Steve
(late). |
|
Apologies in advance from Mark for not turning up |
|
Apologies in advance from Steve for being late |
- Mark has not done the Gantt chart. We expect him to do so this
afternoon so it can be included within Report IV.
- We are unsure about whether Michael has written an environment report,
since he hasn't turned up yet.
Everything else on last week's To Do list
has been done.
This
item has been postponed until next week, since Mark is not here, and the rest
of us have limited knowledge of CVS.
Component
|
To be reviewed by
|
Bot
Form
|
Colin
|
Comms
Debug
|
Colin
|
Run
Information Dialogue
|
Colin
|
About
box (and subordinates)
|
David
|
Show
Environment
|
David
|
TED
|
David
|
Evaluation
|
Mark
|
Run
Information Functions
|
Mark
|
Worldsize
|
Mark
|
Hit-o-gram
|
Michael
|
Main
Form
|
Michael
|
Parser
|
Michael
|
Bot
Box
|
Mik
|
Comms
|
Mik
|
Environment
Window
|
Mik
|
Environment
|
Steve
|
Fasteval
|
Steve
|
Geometry
|
Steve
|
- First reaction: "great idea!"
- Conversion between HTML and DOC/RTF is fiddley either way, and maintaining
two separate files is a pain. So maybe it's not such a great idea.
- Netscrape is clever enough to load Word when it sees an RTF file--Mik
suggested that we just link to RTF files, and posted this suggestion to Mark
via e-mail.
Deficiency:
|
Possible Remedy:
|
Who will fix it?
|
ID
zero is assigned to multiple bots.
|
Perhaps
only bots in one world are being assigned IDs, since when we have two worlds
with three bots each, three of them have ID zero. OTOH, maybe the last bot in
each world isn't being assigned a number.
|
Colin.
|
The
fitness function needs improvement
|
|
Nobody.
|
The
hit-o-gram does not display zero hits properly
|
No
idea.
|
Steve.
|
Loading
and saving templates causes crash
|
|
Michael.
|
General
corruption of bot IDs
|
Trace
through the code that assigns bot IDs
|
Colin.
|
You
can't restart the simulation after stopping it (as opposed to pausing it)--it
says "cannot hook up with brain component"
|
Perhaps
the brain is not being closed properly after stopping, or maybe it's happening
too fast.
|
Colin.
|
The
hit-o-gram only shows hits--drops and pickups would be a nice addition.
|
Using
tabs.
|
Steve.
|
Mark
was volunteered to write a couple of paragraphs about our presentation at Open
Day. He'll do this this afternoon, so it can be included in Report IV.
Mark
didn't do it. Steve will try to do it today, and if not, will get Mark to
finish where he left out.
Mik
will e-mail these clowns and set up a meeting time.
Version
1.00 will be finished very soon. The bugs that need to be fixed are noted
above, and the other things to do are listed in the To Do list below.
Michael's
telephone number has been written to Report IV.
#
|
Description
|
Person responsible
|
Was Due
|
Is it Done?
|
Now Due
|
10.
|
Parser
error message
|
Colin
|
26th
|
Yes.
|
|
12.
|
The
real environment window
|
Steve
|
26th
|
Yes.
|
|
13.
|
Loading
and saving runs
|
Michael
|
31st
|
No.
|
2nd
|
14.
|
Parser's
okay message
|
Colin
|
2nd
|
Need
#13.
|
2nd
|
15.
|
Hit-o-gram
|
Steve
|
2nd
|
Not
sure.
|
2nd
|
16.
|
Robot
ranking
|
Mik
|
26th
|
Mostly.
|
2nd
|
17.
|
BOTSsim
(main) window
|
Steve
|
22nd
|
Yes.
|
22nd
|
18.
|
TED
scroll bar bug fix
|
|
|
|
|
20.
|
Update
online help
|
Steve
|
2nd
|
Yes.
|
|
21.
|
Update
user documentation
|
Steve
|
2nd
|
No.
|
2nd
|
22.
|
Concept
of a generation (time steps)
|
Steve
& Colin
|
26th
|
Yes.
|
|
23.
|
Sending
fitnesses
|
Michael
|
29th
|
Yes.
|
|
24.
|
Automation
of animation
|
Steve
& Colin
|
26th
|
Yes.
|
|
|
venue: UWSN KY238 (Old Project Room) |
|
time: | 9 September 1997, 9:15am |
|
present: | Colin (late), David, Mark (late), Michael (late), Mik,
Steve |
- Mik has not yet contacted group 9707--would like to bring this up as a
meeting item first.
- It works on Mark's machine at home.
- With some work, Mark and David should be able to get CVS working on our
machines.
- CVS will also be beneficial for people working with our project in future.
- What we need to know:
- The commands are:
- Commit--checking in
- Check out--can do modules or files, but Mark says it's better to check out
modules only
- Update--updates checked out source with recent commitments. If somebody
else has been working on the same code as you, it will complain, but not
overwrite your changes with somebody else's.
We need to divide the project
up into modules
- This is better done after we have been given our new direction by George.
- It's better if the modules are in separate directories.
- Mark, Steve and Colin have already done their code review.
- It was emphasised the we are to write comments about the only, and
not change the code or the comments within it.
- Everyone will do their code review by next week, and write their comments
to /documentation/internal qa directory.
How
should we do it?
- Ask them for:
- scope
- objectives
- logical design
- user documentation
We make up a test plan
- We can go away and make up our own
- Then we'll get together and make a big evil one
After that we ask
them for: We crash their code
- We tell them to fix it
- We laugh
- We'd like Michael (and everyone else for that matter) to write .TED files
as text (or .INI style files). This will hopefully fix any saving problems, and
gives us the ability to debug such things.
- Much discussion about the best way to do this.
- Everyone will have a think about it, and post their ideas to the list.
- When we've got a new direction, we should redefine the modules, and review
the level of abstraction we have.
- Carolyn wants an ERD for tomorrow evening's project meeting.
- ERDs are not relevant to our project, because our project does not
manipulate much data.
- Steve and Mik will explain why in the presentation, and show a DFD instead.
#
|
Description
|
Person responsible
|
Was Due
|
Is it Done?
|
Now Due
|
13.
|
Loading
and saving runs
|
Michael
|
2nd
|
No.
|
Yesterday.
|
14.
|
Parser's
okay message
|
Colin
|
2nd
|
Need
#13.
|
|
15.
|
Hit-o-gram
|
Steve
|
2nd
|
Almost.
|
9th.
|
16.
|
Robot
ranking
|
Mik
|
2nd
|
Almost.
|
9th.
|
21.
|
Update
user documentation
|
Steve
|
2nd
|
Yes.
|
|
Deficiency:
|
Possible Remedy:
|
Who will fix it?
|
Is it fixed?
|
Further comments
|
ID
zero is assigned to multiple bots.
|
Perhaps
only bots in one world are being assigned IDs, since when we have two worlds
with three bots each, three of them have ID zero. OTOH, maybe the last bot in
each world isn't being assigned a number.
|
Colin.
|
Yes.
|
It
was a parser bug.
|
The
fitness function needs improvement
|
|
Brain
team.
|
Yes.
|
Still
being improved.
|
The
hit-o-gram does not display zero hits properly
|
No
idea.
|
Steve.
|
Yes.
|
Yay
Steve and Colin.
|
Loading
and saving templates causes crash
|
|
Michael.
|
No.
|
Change
it to text... we'll trash it out in e-mail.
|
General
corruption of bot IDs
|
Trace
through the code that assigns bot IDs
|
Colin.
|
Yes.
|
See
above.
|
You
can't restart the simulation after stopping it (as opposed to pausing it)--it
says "cannot hook up with brain component"
|
Perhaps
the brain is not being closed properly after stopping, or maybe it's happening
too fast.
|
Colin.
|
Yes.
|
Possibly
caused by paging--it's less of a problem when there's more RAM.
|
The
hit-o-gram only shows hits--drops and pickups would be a nice addition.
|
Using
tabs.
|
Steve.
|
Done.
|
Nice
work, Steve.
|
Bots
can drop one block an infinite number of times (without picking it up each time.)
|
Either
it's not dropping properly, or the pickup counter is not implemented.
|
Michael.
|
|
|
|
venue: UWSN KY108 (New Project Room) |
|
time: | 16 September 1997, 9:15am |
|
present: | Colin (late), David (late), Mark (late), Michael, Mik,
Steve |
- Mik was unable to finish robot ranking because he couldn't get into the
project lab.
- Mark left his internal QA at home. Last week he said the dog ate it.
Michael just didn't do it.
- "Where's our smeggin' second 'poota??"
- We don't have one, but it doesn't really matter.
Mark can't be a
vampire by day anymore, but he can see the full moon at night, and at least
we're near the Coke machine.
After
some discussion about what we should be aiming for, we agreed that the central
focus should be a high distinction. To achieve this, we need:
- Good user interface and documentation
- Discussion about the target audience. We agreed that we should aim for two
target audiences:
- A genetic programming researcher
- A user who is developing real-life bots to run around a real-life
factory
Our final documentation should include a "where to now?"
section
Bug-free code
- We should allow plenty of time for bug fixing and internal QA
To
explore new frontiers in genetic programming
- This not only brings on HDs, but adds to the name of the University, and
gives us a sense of achievement
A professional web page
- Again, improves the University's profile, and makes it more likely for
somebody in Sweden or something to get something out of our code
- Mik and Michael will work on this
We discussed the new aim of
the project. The program should be able to produce the following things for the
following users:
- For the factory bot designer:
- An optimal rule set
- Perhaps more--to be discussed
For the GP researcher:
- An optimal rule set
- Statistics about simulations, perhaps in a .CSV file so we can use Excell
or whatever to make pretty graphs.
- Perhaps more--to be discussed
- We have reviewed the project's module abstraction. The major problem that
we observed stemmed from the lack of a CBEP module to handle all of the CBEP
specific routines. Before we had a number of modules directly using the
Communications Routines from the Comms module, this was a complex and possibly
disastrous situation.
- To combat this complexity and the potential for danger we developed a CBEP
module. In the CBEP module we placed all of the routines that implemented parts
of the Environment/GUI side of the Complex Brain Environment Protocol (CBEP).
The CBEP module now includes routines for the following:
- Getting and Setting Rule Sets for particular Bots.
- Setting the GP Variables.
- Sending a Bot's Sensor information and receiving the Bot's corresponding
Action.
- Setting the Bots' Fitness Values to the Brain.
- Initialising the Rule Sets for Bots.
These changes mean that
runinfofunc.pas and parser.pas are no longer required.
- There are of course other problems involving the abstraction of our
modules in BOTSsim. These problems center around the Evaluation object and it's
methods. These problems will be addressed if we have the time.
This
has been deferred to tonight's meeting.
Although
we can see some minor benefits, we don't want a database now for the following
reasons:
- They're inefficient--the simulations are slow enough without having to do
any database insertions, and there's no way to get around that really.
- Our results are flat files
- We only have one table (according to Mark)--there are insufficient
relations in the data to make it worthwhile
This
has been deferred to tonight's meeting.
Deficiency:
|
Possible Remedy:
|
Who will fix it?
|
Is it fixed?
|
Further comments
|
Loading
and saving templates causes crash
|
|
Steve
and Colin.
|
No.
|
We
still need to thrash out how to do it.
|
Bots
can drop one block an infinite number of times (without picking it up each time.)
|
Either
it's not dropping properly, or the pickup counter is not implemented.
|
Steve
will look at the code and wee what he can do.
|
No.
|
|
If
Bot.Carrying <> NIL Then Bot.Carrying.Carried := False causes an
access violation.
|
Dangling
pointer problem.
|
Mark.
|
No,
but Colin has done a quick fix.
|
|
When
picking up and dropping, picking up is rewarded by the system but incorrect
dropping is not punished.
|
|
|
Yes.
|
|
When
a bot tries to pick up a block that's not in the DA, neither its fitness nor
its pickup count increases.
|
|
Michael.
|
No.
|
Michael:
"I thought I fixed that!"--it probably got overwritten.
|
There
are two types of fitness: raw fitness, and refined fitness. Raw fitness follows
a higher-is-better convention, and is displayed throughout the program. Refined
fitness, calculated at the end of each generation, follows a lower-is-better
convention, and is used within the brain and is printed to log files.
The fitness function will be reviewed shortly.
We
will save the following in a text format, following an object-orientated
approach. The following will be saved:
- Environment templates (*.TED) (Template Editor Designs)
- Run information variables (*.RIV)
- Rule sets (*.SEX) (S-EXpressions)
We
will have a special group meeting tonight at 7:00pm to finalise project
direction, abstraction and CVS setup.
- BOTS project direction--our suggestions:
- Optimising fitness function
- For the factory bot designer:
An optimal rule set
- For the GP researcher:
- An optimal rule set
- Statistics about simulations, perhaps in a .CSV file so we can use Excell
or whatever to make pretty graphs.
Is it okay if we deliver just (a
really good version of) what we promised last semester?
- Does he want a Report V? If so, by when?
- We need updates of the following utilities to implement CVS:
- GNU diff utilities
- A "decent version" of RCS
#
|
Description
|
Person responsible
|
Was Due
|
Is it Done?
|
Now Due
|
13.
|
Loading
and saving runs
|
Michael
|
1984.
|
No.
|
"Some
time this week"
|
14.
|
Parser's
okay message
|
Colin
|
|
Need
#13.
|
|
15.
|
Hit-o-gram
|
Steve
|
9th.
|
Almost.
|
"Some
time this week"
|
16.
|
Robot
ranking
|
Mik
|
9th.
|
Almost.
|
a.s.a.p.
|
|
venue: UWSN KY108 (New Project Room) |
|
time: | 23 September 1997, 9:15am |
|
present: | Colin (late), David (late), Mark (late), Michael
(late), Mik, Steve |
- Mik hasn't finished robot ranking yet, but he's really working on it.
- Mark left his internal QA at home again. Michael forgot about it.
- Bots should be rewarded (refined fitness deducted) for the following
thing:
- Running towards a block
- Picking up the block
- Running towards a bin
- Dropping block in bin
Better bots have lower refined fitness.
- This information will be presented to George, along with previous fitness
functions tried, in the hope of George coming up with a better fitness function.
- David started typing a report on this subject, but NT ate it.
- Target users:
- For the bot factory designer: For the GP researcher:
- An optimal rule set
- Best fitness per run
- Average fitness per run
- Diversity: how many different rule sets per run
- Incest: The probability of bots crossing over with another bot with a
similar rule set
- Fitness histogram: fitness per bot per generation (3D graph)
- Hits histogram: number of hits per bot over one generation (shown in
realtime)
- Create a new component called the Statistics component which generates the
above using a log, and gives the option to have the above in .CSV or .GNU, or
something that some NT graphics program likes to read.
This
warrants further investigation, so that GP research users can change it. The
brain team will look into this.
Creating
the stats component GUI, standardising interface, etc.
We
need to see what George wants now that he's back, and we'll create a schedule
after the break.
We
will discuss this properly after we've completed internal QA.
Alf
is currently looking into this.
This
will be reviewed at the bot meeting after the break, by which time Michael and
Mark will have done theirs.
The
other group has no documentation to give us yet. We'll hassle them again after
the break.
It
was noted that we need to give a presentation after the break. We will know
tomorrow night when it is. Since Colin, Mark and David's Wednesday night
lecturer is George, they will hopefully be accommodated for.
Our
only meeting over Spring will be playing Robo Rally some time.
Deficiency:
|
Possible Remedy:
|
Who will fix it?
|
Is it fixed?
|
Further comments
|
Loading
and saving templates causes crash
|
|
Steve
and Colin.
|
Yes!!!
|
Hooray!
|
Bots
can drop one block an infinite number of times (without picking it up each time.)
|
A
side-effect of the next bug:
|
|
|
|
When
a bot tries to pick up a block that's not in the DA, neither its fitness nor
its pickup count increases.
|
|
Michael.
|
Yes,
but not tested.
|
Michael:
"I thought I fixed that!"--it probably got overwritten.
|
If
Bot.Carrying <> NIL Then Bot.Carrying.Carried := False causes an
access violation. Has to be fixed properly!
|
Dangling
pointer problem.
|
Mark.
|
|
|
Resetting
and starting a new run often causes "Could Not Hook Up Brain Component".
|
Try
hooking up ten times, and then report an error.
|
Steve
and Colin.
|
|
|
Mik
will ring everyone else next time to make sure they turn up on time--be
here at 9:00am!
- BOTS project direction--our suggestions
- Fitness function
- Is it okay if we deliver just (a really good version of) what we promised
last semester?
- Does he want a Report V? If so, by when?
- We need updates of the following utilities to implement CVS:
- GNU diff utilities
- A "decent version" of RCS
#
|
Description
|
Person responsible
|
Was Due
|
Is it Done?
|
13.
|
Loading
and saving runs
|
Michael
|
"Some
time last week"
|
No,
but leading and saving templates works real good now!
|
14.
|
Parser's
okay message
|
Colin
|
|
Yes.
|
15.
|
Hit-o-gram
|
Steve
|
"Some
time last week"
|
Yes.
|
16.
|
Robot
ranking
|
Mik
|
a.s.a.p.
|
Just
about.
|
|
venue: | UWSN KY108 (New Project Room) |
|
time: | 14 October 1997, 9:15am |
|
present: | Colin, David (early), Mark, Michael (late),
Steve |
|
| Mik sent his apologies. |
- Loading & saving Runs will be split up.
- Loading & Saving Templates--Done.
- Loading & Saving Run Information--Stephen.
- Loading & Saving S-Expressions (Rule Sets)--Colin.
MikZ has done
as much of Robot Ranking as he can. Bot IDs or fitnesses are incorrectly
assigned.
- The Brain Team
- Testing Greedy Over-Selection.
- Waiting for more RAM which was promised to us.
The Environment
Team
- Michael will profile the Environment.
- Most if not all Environment bugs have been fixed.
The GUI Team
- Saving & Loading Runs (Templates, Run Info, S-Expressions) needs to be
done. This includes adding buttons, menu items, writing the I/O code and so on.
- Standardise the Interface, this probably already done, but we'll have a
look at it anyway .
- What the hell is going on ?
- The team we are performing QA on has not provided us with any
documentation, executables or anything at all.
- We will E-mail them and allow one more week before we really get pissed off.
- When are we doing this ?
- Stephen will find out when he attends the meeting on Wednesday night.
What do we have to present on ?
- Stephen will find out when he attends the meeting on Wednesday night.
Which group members are going ?
- Stephen, Mark, MikZ, Colin.
- What we need to do for it:
- Concrete and understandable justification for more RAM (Brain Team).
- Progress summaries for all teams--David & Colin, Stephen,
Michael.
When is it due.
- Next week probably. We will consult with George.
- Need to be finished by Week 13.
- Testing the integration of all components--Black Box.
- We cannot demonstrate the project to George today because of a lack of RAM.
- We decided against implementing CVS because:
- It is not worth it at this late stage.
- Mark understands it best and is currently overloaded with other
tasks.
- David also could do it but he is busy as well.
- We have PERL for NT. We need GNU Plot for NT or something similar, we need
to ask George to obtain this for us.
Michael
still has not done his Internal QA. He will try and complete this today.
We
will still discuss this properly after we've completed internal QA.
The
other group still has no documentation to give us yet. See above.
It
was noted that we need to give a presentation after the break. We will know
tomorrow night when it is. Since Colin, Mark and David's Wednesday night
lecturer is George, they will hopefully be accommodated for.
Deficiency:
|
Possible Remedy:
|
Who will fix it?
|
Is it fixed?
|
Further comments
|
If
Bot.Carrying <> NIL Then Bot.Carrying.Carried := False causes an
access violation. Has to be fixed properly!
|
Dangling
pointer problem.
|
Mark.
|
No.
|
|
Resetting
and starting a new run often causes "Could Not Hook Up Brain Component".
|
Try
hooking up ten times, and then report an error.
|
Steve
and Colin.
|
Yes.
|
We
added a Retry Count J
|
Mik
will ring everyone else next time to make sure they turn up on time--be
here at 9:00am!
- Let George know about Loading and Saving Runs.
- Inform George about Team progress, see above.
- Let George know about our dissatisfaction with our QA team.
- Talk about our presentation, see if George has any inside information.
- Ask George about Report 5.
- Extra RAM - related to our Demo and Testing.
- Let him know we've given up on CVS.
#
|
Description
|
Person responsible
|
Was Due
|
Is it Done?
|
1
|
Loading
and saving Run Information.
|
Stephen.
|
14th
|
|
2
|
Loading
and saving S-Expressions.
|
Colin
|
14th
|
|
3
|
Robot
ranking
|
Mik
|
14th
|
|
4
|
Testing
Greedy Over-Selection.
|
Brain
Team.
|
14th
|
|
5
|
Profile
the Environment.
|
Michael.
|
21st
|
|
6
|
Find
out Presentation Information.
|
Stephen
|
15th
|
|
7
|
Team
Progress Reports.
|
David
& Colin--Brain, Stephen--GUI, Michael--Environment.
|
18th
|
|
8
|
Update
Gantt Chart.
|
Mark.
|
21st
|
|
|
venue: | UWSN KY108 (New Project Room) |
|
time: | 21 October 1997, 9:15am |
|
present: | Colin, David, Mark, Michael, Mik, Steve |
- Mik said last week that he'd fix the robot ranking problems this week. He
repeated this today.
- The following table explains the rest...
#
|
Description
|
Person responsible
|
Was Due
|
Is it Done?
|
1
|
Loading
and saving Run Information.
|
Stephen.
|
14th
|
Yes.
|
2
|
Loading
and saving S-Expressions.
|
Colin
|
14th
|
No.
|
3
|
Robot
ranking
|
Mik
|
28th
|
|
4
|
Testing
Greedy Over-Selection.
|
Brain
Team.
|
14th
|
Yes.
Gave worse results.
|
5
|
Profile
the Environment.
|
Michael.
|
21st
|
No.
Steve will do it by next week, and "Michael will stick his head in a pig"--Mark.
|
6
|
Find
out Presentation Information.
|
Stephen
|
15th
|
Yes,
but we don't know our target audience.
|
7
|
Team
Progress Reports.
|
David
& Colin--Brain, Stephen--GUI, Michael--Environment.
|
18th
|
Steve's
done it. Nobody else has. Grrrr.
|
8
|
Update
Gantt Chart.
|
Mark.
|
21st
|
No.
|
- Steve has done his bit, but nobody else.
- We spent a few minutes reviewing e-mail to clarify what needed to be done.
- Since George hasn't marked our Report IV, it's fair to aim to have Report
V ready by this time next week.
We
revised the Gantt chart, to include the entire project.
David
has not done his yet. It was noted that two months ago he said he had. We want
him to have it done by 24 October 1997.
We
have set a deadline for their QA submission. We have e-mailed notice of this
deadline to them. The deadline is 9:15am on 28 October.
Instead
of having the GUI constantly polling the brain for information, we can use
blocking I/O and have the operating system wake up the GUI when the brain
responds. This dramatically reduces CPU load, and should speed up the
program.
Michael will implement this Friday.
- It makes the output more user-friendly--definitely worth marks.
- It is easy to implement and shouldn't take up much time--most of the work
is already done.
- Icing on the cake.
- Clearly delivers more than what we promised last semester.
- Porting PERL scripts to NT--may not be transparent.
- Launching Excell
- Explanation of statistics needed asap!!
- Exactly how to represent the data as CSV files--Mik will look into this.
- Our presentation is on 5 November (same date as KBS is due)--a fortnight
from tomorrow.
- The time limit is twenty minutes. This does not include time for
questions, but does include set-up time.
- Everyone. George will provide a catchup tute for Colin, Dave and Mark.
It
is a progress report, informing everyone what's changed since the end of last
semester.
- Objective of this presentation (1 minute)
- Problem outline--objective, wish list, etc. (4 mintues)
- Progress made (5 minutes)
- Brief history of our project this semester
- Recent achievements
- What we have left to do
Problems encountered (5 mintues)
- Trying out different rule sets
- Out of memory--includes brief explanation of trees, S-expressions,
etc.
Changes in client's wishes (1 minute)
- None--say how nice the client is, how he's helped us with ideas, etc.
Quick demo (4 minutes)
- Show bots nooping
- Show bots after n generations, just picking up blocks
- Show fudged results
- Tuesday 25 November at 11:30am seems to suit all of us. If anything
changes, we'll let George know immediately.
- Steve will find out from Carolyn the purpose and format of the final
presentation.
Deficiency:
|
Possible Remedy:
|
Who will fix it?
|
Is it fixed?
|
Further comments
|
If
Bot.Carrying <> NIL Then Bot.Carrying.Carried := False causes an
access violation. Has to be fixed properly!
|
Dangling
pointer problem.
|
Mark.
|
No.
|
|
Discussion
on how to improve group meetings. Next week we'll try being more flexible in
what we discuss at what time. We'll periodically check whether we've covered
everything, and cover issues involving the whole group first.
- Can we get the Microsloth GUI standards documentation?
#
|
Description
|
Person responsible
|
Now Due
|
2
|
Loading
and saving S-Expressions.
|
Colin
|
28th
|
3
|
Robot
ranking
|
Mik
|
28th
|
5
|
Profile
the Environment.
|
Steve.
|
28th
|
6
|
Find
out target audiences, purposes and formats for both presentations.
|
Stephen
|
22nd
|
7
|
Team
Progress Reports.
|
David
& Colin--Brain.
Michael--Environment.
|
24th
|
8
|
Update
Gantt Chart.
|
Mark.
|
21st
|
9
|
Blocking
I/O
|
Michael.
|
24th
|
10
|
Internal
QA
|
Dave.
|
24th
|
11
|
Feasibility
of BOTSstat
|
MikZ,
Dave
|
28th
|
|
venue: | UWSN KY108 (New Project Room) |
|
time: | 28 October 1997, 9:15am |
|
present: | Colin, David, Michael (late), Mik,
Steve |
|
| Apologies from Mark. |
Everything
was done, except:
- Michael has not implemented blocking I/O. "Oh! Bugger! I'll do it tonight."
- Neither Steve nor Michael have profiled the environment. Michael does not
have the appropriate software to do it. We've decided to abandon the idea of
doing it, since we no longer have time to optimise any code that's discovered
to be inefficient anyhow.
- David has done little research on the feasability of adding options to his
Perl scripts.
#
|
Description
|
Person responsible
|
Was Due
|
Is it Done?
|
2
|
Loading
and saving S-Expressions.
|
Colin
|
14th
|
Yes.
|
3
|
Robot
ranking
|
Mik
|
28th
|
Yes.
|
5
|
Profile
the Environment.
|
Michael.
|
21st
|
No.
Michael said he'd do it yet, but he's not here yet to say whether or not he
has. Maybe his head is stuck in the pig.
|
7
|
Team
Progress Reports.
|
David
& Colin--Brain, Stephen--GUI, Michael--Environment.
|
18th
|
Michael
still hasn't done his.
|
8
|
Update
Gantt Chart.
|
Mark.
|
21st
|
Yes.
|
Now
that internal QA has been completed, the authors of the various modules will
fix the problems raised. Not only will this fix problems with our software, but
it gives our programmers constructive criticism about their programming styles.
Module/unit/form
|
Reviewed by
|
Will be fixed by
|
Fasteval
unit
|
Steve
|
Colin
|
mainfrm.pas
|
Michael
|
Colin
|
Evaluation.pas
|
Mark
|
Michael
|
COMMS.PAS
|
MikZ
|
Michael
|
Environment
unit
|
Steve
|
Michael
|
Geometry
unit
|
Steve
|
Michael
|
showanenv.pas
|
David
|
Mik
|
botform
|
Colin
|
Mik
|
Runinfofunc.pas
|
Mark
|
nobody--it
no longer exists... see notes on module abstraction.
|
Parser.pas
|
Michael
|
nobody--it
no longer exists... see notes on module abstraction.
|
ted.pas
|
David
|
Steve
|
CommsDebugDlg
|
Colin
|
Steve
|
RunInformation
|
Colin
|
Steve
|
runinformation.pas
|
Mark
|
Steve
|
bot
properties dialogue
|
MikZ
|
Steve
|
ENVIRONMENT
BOX
|
MikZ
|
Steve
|
runinformation.pas
|
Michael
|
Steve
|
hits.pas
|
Michael
|
Steve
|
That
group has failed to meet our deadline. We now do not have enough time to do QA
on their project, and we will report this as such. Everyone who has spoken to
them e-mailed them, or whatever, will document this.
Mik
demonstrated that Excel is quite capable of producing the types of graphs we
need.
Some discussion about methods of passing parameters from BOTSstat to Dave's
Perl scripts. The easiest way is for us to simply transfer them via a
configuration file.
The mirror messages is a more pressing issue for Dave at the moment, although
Dave will try to find time to do his side of BOTSstat over the next week or so.
Mik will also see if he can make the necessary changes to Dave's Perl scripts.
Our
presentation is on 5 November (same date as KBS is due)--a week from tomorrow.
It is a progress report, informing everyone what's changed since the end of
last semester.
We discussed who was going to do what part, and finalised time limits.
- Objective of this presentation--Mik, 1 minute.
- Problem outline--objective, wish list, etc.--Steve, 4 minutes.
- Progress made
- Brief history of our project this semester--1 minute
- Recent achievements--Mark, 2 minutes
- What we have left to do--Dave, 2 minutes
Problems encountered--Mark,
5 minutes
- Trying out different rule sets
- Out of memory--includes brief explanation of trees, S-expressions,
etc.
Changes in client's wishes--Michael, 1 minute
- None--say how nice the client is, how he's helped us with ideas, etc.
Quick demo--Colin and Steve, 4 minutes
- Show bots nooping
- Show bots after n generations, just picking up blocks
- Show fudged results
We will meet at 2:00pm this Saturday for
our first walkthrough.
- The presentation will be on Tuesday 25 November at 11:30am.
- The target audience is our client, and project supervisors.
- The general purpose should be to inform the audience about the project
status. A demonstration should play a big part in the final presentation,
probably around 5 minutes.
Deficiency:
|
Possible Remedy:
|
Who will fix it?
|
Is it fixed?
|
Further comments
|
If
Bot.Carrying <> NIL Then Bot.Carrying.Carried := False causes an
access violation. Has to be fixed properly!
|
Dangling
pointer problem.
|
Mark.
|
No.
|
|
Things
to add to the user documentation:
- System requirements;
- Introduction to what the program does.
- Everyone will prepare for next week's presentation by 1/11/1997.
- Everyone will read the internal QA written about their code, as specified
in the table above, by 4/11/1997. Where possible they will fix the problems.
- Everyone will document the contact (both verbal and e-mail) they've had
with SAD9707 by 4/11/1997.
- Michael will implement blocking I/O by 29/10/1997.
- Mik will write a standard for BOTSstat óDavePerl communication file
by 30/10/1997.
- David will do his best to change his Perl scripts to work with that file
by 4/11/1997.
- Mark will fix the dangling pointer problem in the environment by 4/11/1997.
- Colin will write the system requirements for BOTSsim by 4/11/1997.
|
venue: UWSN KY108 (New Project Room) |
|
time: | 4 November 1997, 9:00am |
|
present: | David, Colin (late), Mark (late), Michael, Mik, Steve |
Everything
was done, except:
- Everyone except Colin and Michael have started making changes reflecting
the Internal QA on their code. Mik has finished all of his, and Steve only has
TED to go.
- Michael still hasn't implemented blocking I/O.
- David still needs to sort out some things with Mik before he can change
his Perl scripts to work with BOTSstat.
- Mark still hasn't fixed the dangling pointer problem in the environment.
Since it is a minor problem, we've decided to put off fixing it. It should be
fixed by 18/11/1997.
- Colin hasn't written the system requirements for BOTSsim. He claims to not
know them. This was discussed as a group--see below.
That
group tried to give us some documentation last Thursday, but since it was past
the deadline, we declined. This has been documented in Report VI.
Mark
will write about this by noon today.
Steve
will write about these by noon today.
Mark,
with the help of the rest of us, will get this done by 1:00pm today.
These
have not been written, but as a group we remembered what happened and wrote the
minutes.
The
loading and saving of all data has now been implemented. Mik and Steve will
change the main window so the currently loaded files are shown on the status
line.
- We have done a number of walkthroughs, so preparation is progressing.
- For the demonstration, this is needed:
- Colin and Mark will ask Rocky (and probably Ian after that) about having
our machine available for the presentaiton.
- All future tests of BOTSsim will be done in 800x600 mode, to make sure it
can be presented tomorrow night.
- Mark and Colin will do a demonstration walkthrough (in 800x600 mode) this
afternoon.
Deficiency:
|
Possible Remedy:
|
Who will fix it?
|
Is it fixed?
|
Further comments
|
If
Bot.Carrying <> NIL Then Bot.Carrying.Carried := False causes an
access violation. Has to be fixed properly!
|
Dangling
pointer problem.
|
Mark.
|
No.
|
This
is a minor problem, so we've given Mark until the 18th to fix it.
|
Show
Environment button doesn't work in the Bot window.
|
Uncomment
and debug.
|
Steve
and Mik.
|
|
|
- Discussion about whether the dangling pointer problem is worth worrying
about. Apparently it only causes memory leaks in obscure circumstances, so we
can afford to put off fixing it until 18/11/1997.
- The minimum system requirements are Windows NT, 16MB of RAM (64MB
recommended), standard VGA card and monitor, 38MB of free disk space for
program storage, 500MB of disk space recommended for a run.
- Talk of support for a rule set-free log file, to reduce the space required
for a run. Colin will look into this.
- Everyone will make changes reflecting the Internal QA on their code by
11/11/1997.
- Michael will implement blocking I/O before we kill him.
- Mark will write a couple of paragraphs about mirror messages by noon
4/11/1997.
- Steve will write a BOTSs about file formats by noon 4/11/1997.
- Colin and Mark will ask Rocky (and probably Ian after that) about having
our machine available for the presentaiton by 7:00pm 4/11/1997.
- Mark and Colin will do a demonstration walkthrough (in 800x600 mode) by
7:00pm 4/11/1997.
- The brain team will tests BOTSsim in 800x600 mode by 7:00pm 4/11/1997.
- Steve will implement the Display of Multiple bot windows by 11/11/1997.
- Steve and Mik will change the main window of BOTSsim to show currently
loaded files by 11/11/1997.
- Steve and Mik will implement the Show Rule set button in the bot
window by 11/11/1997.
- Colin will estimate the amount of time required to implement a rule
set-free log file option by 11/11/1997.
- David will make the changes to his Perl scripts to implement BOTSstat by
18/11/1997.
- Mark will fix the dangling pointer problem by 18/11/1997.
- Mik will finish the GUI component of BOTSstat by 18/11/1997.
|
venue: | UWSN KY336 (George's Room) |
|
time: | 5 August 1997, 3:00pm |
|
present: | Colin, David (late), Mark, Michael,
Steve |
- Discussed what George wants us to do now, which is to just finish the
project
- Light discussion about what to do when the original project is
finished--final decision deferred until then
- George wants a To Do list by next week
|
venue: | UWSN KY336 (George's Room) |
|
time: | 12 August 1997, 3:00pm |
|
present: | Colin, David, Mark, Michael, Steve |
- The majority of the meeting discussed our progress to date and where we
were heading this semester once we finished Version 1.00 of the BOTS Project.
- The To Do list was discussed in depth with all group members taking part
in clarifying things for the supervisor. Our explanation of the project helped
George come to grips with when the project would be ready.
- Brief discussion about when George expects report IV. The final decision
was that we would give it to him in person upon his return to Australia on the
Thursday in Week 7. We also agreed to publish the report on the World Wide Web
to give George a chance to view it while overseas. The WWW publication was due
midway through Week 6.
|
venue: | UWSN KY336 (George's Room) |
|
time: | 19 August 1997, 3:00pm |
|
present: | Colin, David (late), Mark, Michael,
Steve |
- Talked about our progress with getting the Brain to communicate with the
Environment/GUI. This involves getting the socket and it's messages flowing
back and forth so that full testing can be done.
- Noted that everything hinges on getting communications stable--we mean
everything.
- George requested something concrete in Week S6, ie: a demonstration of the
software. This should be provided at the same time as the next report that we
give him.
- The next time we see George will be on Thursday 11 September 1997. It has
already been noted that Stephen and MikZ will likely be unable to attend.
There
were no client meetings during these weeks. The client was unavailable because
of travel commitments and personal hardship.
|
venue: | UWSN KY336 (George's Room) |
|
time: | 23 September 1997, 3:00pm |
|
present: | Colin, David, Mark, Michael (asleep), Mik, Steve |
Functions
tried so far:
Function
|
Result
|
Raw
Fitness
|
Punish
good bots
|
Raw
Fitness / Distance
|
Punish
good bots
|
1/(((Raw
Fitness + 1) / (Distance + 1)) + Raw Fitness + 1)
|
Does
not run towards block, picks up well
|
Distance
/ (Raw Fitness + 1)
|
Punishes
Picking up Poor at Running towards block
|
Distance
/ (Raw Fitness + 1) ^ 2
|
Punishes
Picking up Poor at Running towards block
|
Distance
+ 200--(Diaganal * Raw Fitness)
|
Does
not pick up
|
- Discussion about rewarding bots based on how far they've moved before
dropping a block.
- Discussion about rewarding bots based on how far the blocks actually move.
- Discussion about rewarding bots in real time, and not at the end of each
run.
- George gave us Koza's Genetic Programming for Mark to get some
ideas from.
Yes.
Week
10.
- GNU diff utilities
- A "decent version" of RCS
George
wants a demonstartion of BOTSsim in week 10.
|
venue: UWSN KY336 (George's Room) |
|
time: | 14 October 1997, 3:00pm |
|
present: | Colin, David, Mark, Michael, Steve |
We
discussed the current issues:
- Our dissatisfaction with the team we need to do QA on--George told us to
document every time we nag them for their documentation, etc. If they don't
deliver, it will reflect badly on them, not us.
- Our presentation--George knows no more about it than we do.
- Report V--George wants it asap.
- To increase the number of rule sets that can be generated per generation,
we need extra RAM in our machine--this has been done.
We
briefed George on our progress since the last client meeting, highlighting the
fact that runs can now be loaded and saved. We also informed George that we
have decided not to implement CVS, because of time restrictions.
|
venue: | UWSN KY336 (George's Room) |
|
time: | 21 October 1997, 3:00pm |
|
present: | Colin, David, Mark, Michael, Steve |
- We mentioned that we are still experiencing difficulties in performing QA
on this group, because they have given us no documentation or code to work with.
- George told us to set a deadline for this group, and notify them of it.
- George returned Report IV to us.
- He was very happy with it, especially the summaries, but said Report V
should be thinner.
- George wants a demonstration of BOTSsim next week.
- George asked if we could have BOTSsim select bots that have desirable
individual abilities, such as dropping blocks into bins, running towards
blocks, etc., and combine the genetic material from these bots. This would
involve complimenting bots' genetic material during the crossover phase, thus
resulting, in our example, with a bot that can run towards a block and drop
blocks into bins.
- We declined to implement this, because:
- It is too big a fudge job,
- Other desirable characteristics in bots would probably not be developed,
such as bots giving way to each other, and
- Time constraints.
|
venue: | UWSN KY336 (George's Room) |
|
time: | 28 October 1997, 3:00pm. |
|
present: | Colin, David, Michael, Steve. |
- We showed George BOTSsim in its current state.
- Bots still aren't dropping in bins.
- We discussed Mark's "mirroring" idea to help bots achieve their goal.
- George was satisfied by the demonstration.
- George agreed that we could hand in Report V tomorrow evening.
- Because we have recently printed reports and presentation slides, we are
low on printer credits.
- George told us to e-mail him to remind him to organise some more.
[2] Of course, we don't know who our user will
be at the moment.
[3] We were never really sure what Mark meant by
this.
[4] This is only if time permits, since Colin
and Mark have to rush off to some 6:00pm lecture at Westmead.[5]
This has already been done. [6]
We will revise the basic components to the project, ie: brain, GUI,
environment, TED, and possibly help, for Version 2.
<<
, >>
, Title
, Contents
, Index