news

Links

2010-10-28T06:14:53Z

Hi all,

I've had a few people ask me about links to various things. There are two easy ways to find things in this wiki: the 'RecentChanges' link at the top of each page lists recent changes to the wiki :). The 'Site Map' in the 'Quick Links' box on the right hand side lists every page in the wiki. :)

Having said that, the notes from the overview lecture in the last week of class are here, the RL notes are here, and the sample exam is here. Malcolm's planning slides are here.

Will :-}

Exam 2010

2010-10-19T05:23:00Z

Just so everyone is aware. The exam is scheduled for the evening of Tuesday 2nd November, 17:45 - 21:00 (5:45pm - 9:00pm). It will be held in K17_B02. It will be open book, calculator, notes, laptop, internet. The rule is no communication with another person.

Assignment 2 notes 2010

2010-10-18T23:33:55Z

Hi all,

I've had some people ask some questions about assignment 2 this year. Here are the answers so that everyone is on the same page:

Goals: will they be there or not? I have no strong opinion either way. I'm leaning towards them being there, but if that causes a problem for someone's localisation then I'll reconsider. I would like it consistant - I don't want to be taking them on and off the field for each team's run.
Beacons: Let's have them in the same positions as assignment 1. You are allowed to use your solution to assignment 1 in assignment 2, so you should have reasonable localisation.
Usage of functions: In previous years the early assignments already had you implementing ball tracking and grabbing, so I didn't allow use of the built-in code. This year I'm happy for you to re-use the grab. I'd prefer if you implemented your own ball tracking (but look at the code that is there). I'm happy to talk about this in the review lecture tomorrow.
Start location: I'll start you pointing across the centre-line. It will not be to far from the centre line. I don't start timing until you cross the centre line (unless your robot seems to be completely confused, in which case I'll ask you to re-start it).
Stopping criterion: I'll stop timing when the ball crosses the centre-line even it if then proceeds to go out. I've had close calls before - It matters if the ball goes out just before or just after it crosses the centre line.

Will

Will's Solution to Homework 2

2010-09-17T05:26:19Z

... is at http://www.cse.unsw.edu.au/~cs3431/robotModel.py or ~cs3431/public_html/robotModel.py.

it will be there until assignment 1 is due.

Robots in War talk 2010

2010-09-13T06:35:40Z

Sydney Opera House & St James Ethics Centre is presenting The Festival of Dangerous Ideas on October 2-3 2010 which will feature key local and international thought leaders (Geoffrey Robertson QC, Famed American criminal lawyer, Alan Dershowitz, SMH Foreign Correspondent Paul McGeough, BBC Broadcaster Tariq Ali, and many more) who will present their dangerous ideas on art, climate change, economics, justice and more big issues along with things that matter in everyday life, with an edge.

As part of the Festival of Dangerous Ideas, Peter W. Singer in "Wired for War: How Human Soldiers are Being Replaced by Machines" explores the increasing use of autonomous robotics in warfare and the ethical, political and legal implications of this revolution. For more information, visit http://www.sydneyoperahouse.com/whatson/pw_singer.aspx.

Singer is Director of the 21st Century Defense Initiative at Washingtons Brookings Institution and served as coordinator of the Obama-2008 campaigns defense policy task force. To watch Singer on The Daily Show, please visit - http://www.thedailyshow.com/watch/thu-january-29-2009/p-w--singer

Would you be able to include some information on the event in an email/newsletter to your fellow staff or students? I can organise an e-card to be sent out to you that can be forwarded if this makes things easier. We also have programs on the Festival that I can send out to you (boxes of 100 or so).

To find out more on the Festival, please visit http://www.sydneyoperahouse.com/ideas

Assignment 1 2010 Notes

2010-09-13T01:22:27Z

Hi all, I've just updated the assignment 1 sheet with the due date we agreed in class just before the break: The lab of week 9.

I've also updated the base robot code you are allowed to use. It no longer crashes when the matrix library is used (I've added a cut-down matrix library to the default python code). I've also added a function to the robots, VisionLink.getMotionDelta(). To get these changes you will need to git pull to update your code (or git clone an entirely new trunk). I've only updated the main branch - if you're using one of the other git branches then you'll want to merge in my changes.

The getMotionDelta() function returns a tuple (fwd,left,turn,count), where fwd,left and turn would be measured in centimetres (for distances, not sure of the angular units) if the robot was correctly calibrated (which may or may not be the case), and count is a count :). These numbers represent what the walking module thinks it has done since the last vision frame (i.e. the last time the python behaviours were called). The count may not be useful, but I put it in for completeness - it represents the number of updates that the motion module has done since behaviours were last called, all those updates are added to give the other three numbers.

Cheers,

Will :-}

Homework 2 2010 notes

2010-09-01T12:53:49Z

Hi all, People seem to be getting worried by observationSample(). You shouldn't worry too much about this function as it is only used by the test harness to generate observations, and not by any of the filters. It needs to pick a beacon to return the observation to - I chose the beacon with least angle relative to the robot in my implementation. You could pick randomly if you like.

expectedObservation() has also generated some questions. Unlike observationSample(), expectedObservation() is passed an observation. From this you can get the beacon ID. You can then generate the expected observation for that particular beacon.

Will :-}

Homework 2 2010 update 1

2010-08-30T07:35:13Z

Hi all, I've updated the source files for Homework 2 to fix a bug or two. There are now some extra functions for you to implement. These will be easy if you understand what is going on. I've also updated the homework sheet to clarify a few things: there is now a 'notes' section at the end of the assignment that may help.

I've also added a new python file test.py. This file can be run using python test.py. It uses the motionSample and (not suppled) observationSample functions to generate a series of motions and observations with the correct mean and variance from a list of supplied actions. The actions given would move the robot in a square (like the first lab), but noise in the execution of those actions leads the robot to only approximate that shape. Test.py uses a kalman filter to track the actual location as the robot walks around. I've also included the output for five different runs of this program.

I'm giving a general extension on this Homework until Friday. I'm very sorry for the bugs. :(

Will

Homework 2 2010 bug

2010-08-30T05:36:44Z

Sigh,

I've found a bug in Homework 2. I'll release some changes later today. Because of this I'm going to extend the assignment until Friday. More details coming...

Will

Homework 2 2010 released

2010-08-21T03:06:11Z

As noted yesterday in lectures, Homework 2 has been released. It is available as Homework2.pdf. This homework is due by the start of Wednesday lecture in week 7 (1 September). I've put the mpmath library in with the python files for the homework.