09-06-2007
A few items:
I have updated Compiling a Memstick with a script that automates many of the steps required to initially load Tekkotsu.
- I have regulation robot soccer balls on loan from Peter Stone's lab. If any team wishes to use one, you'll have to sign one out. Send me an email or stop by office hours.
- I borrowed a memstick writer from Peter Stone's lab, which means that the number of memstick writers now equals the number of teams. Send me an email or stop by office hours to sign one out.
09-07-2007
A few groups have come to me concerning permissions. The default permissions on home directories allow reads from everyone in the class. To facilitate better permissioning, I created a group for each team, group names and members can be found under GroupSetup. Goto SharedFolderPermissions for a brief tutorial on how to set up permissions.
Additionally, a sign up sheet (Grading Sign-up Schedule) is now available for Assignment 1 evaluations. To sign up, just edit the page.
09-10-07
- This Friday my office hours will be from 10:00 am - 11:30am . Please email me if you cannot make these or my Thursday office hours but still wish to meet.
- I will only be available this Saturday and Sunday on a limited basis (only in the mornings). Please take this into account as you finish assignment 1.
- Clarifications for Assignment 1:
- Between evaluations you can send commands to your robot, change positions, etc. During an evaluation your robot's behavior must be autonomous (no using the joystick function).
- Some teams reported difficulty getting the Aibos to stand from certain positions. It is not necessary to provide a behavior that successfully stands from all starting positions. For this evaluation, standing from POUNCE.POS or another suitably non-standing position will be sufficient.
As for reading sensor values (Evaluation 1), it is sufficient to output sensor values to stdout (you need to telnet to 59000 to see these). I expect to see indications of button events as in the tutorial, and readings from any of the sensors defined in ERS7Info::sensorNames.
- For camera data it is sufficient to have ControllerGUI connect to your aibo and open the rawcam and segmentation windows.
- Please note the requirements for turning in an assignment from Lecture 1.
- Briefly:
- A memo describing the problem, the approach, and the results.
- The code.
- Each individual hands in a memo describing his/her particular role in accomplishing the assingment.
- The memos (items 1,3) are to be submitted in hardcopy to Dr. Kuipers and myself. You may email me your code. I prefer you create a .tar.gz of your project directory. You should include a README that, for each evaluation in the assignment, describes the behavior name and source file name of the behavior used for that evaluation. It is okay to use a single behavior to satisfy multiple evaluations.
E.G.
Evaluation # Behavior Name Source File 1 ReadSensors ReadSensors.h
- Briefly:
- No other teams are allowed in the lab during evaluations.
- As many have already noticed, a grading sign up schedule for 9-17 is available.
9-12-2007
- Since the built in behavior for pink ball tracking satisfies the second to last evaluation, I've decided to allow students to use the built in behavior. Just make sure that your team understands how that behavior works, so if I ask you a question about it during evaluation, you'll be able to answer it. I'd suggest spending the time you save on trying to learn kick functionality (the extra credit). Apologies to any teams that have already (re)-implemented this behavior.
- Since we only have a limited number of orange balls, I'll accept extra credit kicks of pink balls. The pink balls, based on feedback from last years class, are harder to kick properly. Orange balls will be available during my office hours. Teams can sign them out at that time.
9-17-2007
- A great job overall on evaluations today! If you didn't hand in your personal or team memo, please do so by class tomorrow.
9-19-2007
- I've put three orange robot soccer balls in the cabinet at the far left corner of the lab. The key to this cabinet is on top of the power (fuse?) box. Please do not take the robot soccer balls outside the lab, and please remember to lock them up and return the key to its hiding place if you are last in the lab.
9-20-2007
AustinVilla motion files are available! Go to /home/stober/public/av. The .lua file these originally came from is also in that directory. It is called sactions.lua. I have only tested some of these motions, but it should give you a good starting point for kicks in the new assignment. Let me know if you have any difficulties.
- Also, don't share these files outside of class.
I've uploaded a draft of Assignment 2 to the Assignments page. When the final version is available I will post that version and make another announcement.
9-23-2007
- A final version of Assignment 2 has been posted. It is the same as the draft version.
- Some more on write-ups for project proposals (Graduate Students Only):
The write-up should be approximately one full single-spaced page. It should describe the topic of the research project you are proposing to design. Explain what question you would like this to answer, and explain why the answer to this question is important, in the sense that it will make a difference to how someone else can reach their goals. At this stage, before having searched the literature, it is OK to state a goal, and for the plan to be to discover what part of that goal is already achieved somewhere in the literature, and what part has yet to be done. But you should make the case that it will make a difference if you succeed. And, like anything else you give anyone, make sure that this paper includes (a) your name (b) the date (c) a title or other short indication of what it is.
9-27-2007
- The rules for penalty kicks in Robot Soccer specify that the ball be place 1m from the center of the field in the direction of the opposing goal. Since a half field is 2700mm long, this means penalty kicks start from 1700mm from the front of the goal. You should use this as a guide for how your kicks will be evaluated. I reserve the right to test from shorter distances as well. The ball may be placed at longer distances for extra credit attempts.
9-28-2007
Some teams have reported problems with memsticks becoming read only. I've put up a new page detailing some things to try should this happen - MemstickReadOnly.
- I've put up some new reading assignments. Please read these before lecture on the specified days.
- Currently, three machines mount memsticks to /mnt/memstick instead of /media/disk (you will need to type mount /dev/sdb1 to mount the memstick on these machines). These are qbert, space-invaders, and bosconian. I will be configuring all the machines to use /mnt/memstick instead of /media/disk shortly.
10-1-2007
- I've put some colored tape down that should help with Aibo orientation. I'm open to suggestions for other landmarks that may make the current assignment more tractable (I have colored construction paper I could put up, for instance).
I've marked the point where penalty kicks are taken in RoboCup. This should be considered an upper limit on how far away the ball will be in attempting to score for regular credit (and a lower limit for extra credit).
- I've put up a grading sign-up schedule for Assignment 2. Please fill it out when you have a chance.
10-2-2007
- I've put down another mark that is 1m from the front of the goal. I think this is a reasonable minimum kick distance to receive full credit for the assignment. Your robot is allowed to "dribble" closer (this is required for extra credit), but it must not enter the penalty area.
One team has reported that the AustinVilla kicks don't seem to work, in part because the Aibo's rear legs don't seem to get much traction on the carpet. If you wish to use these kicks, be prepared to tweak some of the motion parameters.
- I'd like to see some evidence that you've tested a PID controller for walking up to the ball. If you decide to go with a P controller for the full task, that's okay, but you should justify it in your write-up and be prepared to demonstrate on your Aibo how your best tuned PID control compares to the P (or PI or PD) controller you decided to use.
10-4-2007
- Your Aibo is not allowed to shoot in the penalty box. For this assignment, your Aibo may pass through the penalty box while not shooting.
10-7-2007
One group reportedly cannot locate either their memstick or memstick writer. If you found any of these items, or if your group is willing to loan out yours, contact kpfile@gmail.com.
- Each group is required to submit a group memo and individual memos as in Assignment 1.
10-8-2007
- Great job on Assignment 2!
- Tomorrow we'll set aside some class time to head down to the lab. You'll have the chance to demo your robot behaviors for Dr. Kuipers and your classmates. If we have time, we may even have an informal longest kick competition.
- If you haven't already, you should give me your memstick writer tomorrow. We'll keep these in the lab (one per lab workstation) since the lab is now more secure.
10-9-2007
- The 3rd assignment is posted.
10-11-2007
- Last year's midterm has been posted on the Syllabus and Handouts page.
10-17-2007
- I've posted the pool play tournament schedule for 10-29. Let me know if your team has a scheduling conflict.
- Ben just asked me to post the following concerning the midterm:
Folks, A couple of late hints for the mid-term: - Be sure to review the wall-following control law. - Be sure to review the Kalman Filter equations. Cheers, BJK
10-19-2007
- In response to some requests, I've replaced the white penalty box with a yellow penalty box. Also, to aid in opponent identification, I'm allowing each team to construct a "uniform" for their opponent. The "uniform" must not impede the opponent's motors or sensors, and cannot be permanently affixed to the robot (scotch tape should be fine - no super glue). I've left some colored card stock in the drawer with the orange balls. You may use this cardstock or anything else for a uniform that would aid your Aibo in identifying the location of your opponent.
10-23-2007
For the next couple of lectures, the readings will include chapters from "Probabilistic Robotics" by Thrun, Burgard, and Fox. The relevant chapters will be listed on the syllabus and handouts page. Be aware that you may need to jump around a bit in the text to following the material in the order presented in class. Also, you may need to draw on Chapters 3 & 4 for some information.
10-25-2007
- I've added some additional cues for localization. The cues are squares of neon purple and pink cardstock placed at the left and right front corners of the penalty box. These are intended to aid in localization for the goalie. Please let me know if these will interfere with any already complete strategies ASAP.
- Please verify that your team can make your assigned pool play time slot on Monday!
- The neon pink cardstock cue proved to be too similar to the ball color. I've place two purple squares at the corners of the penalty box. I have also expanded the width of the yellow line by about 1cm on three sides.
10-26-2007
- I have not heard any complaints about the neon purple color cues at the corners of the penalty box, so as they will be included in the tournament.
Note for the 3sec blocking rule. The relevant section of the 2007 RoboCup rules is:
The vertical projection of the goalie to the goal line should not occupy more than 50 percent of the length of the goal mouth. Those robots that commit this action (intentionally) will be replaced immediately on the halfway line as per the standard removal penalty for 0 seconds (see Section 4.2). Robots may violate this rule for brief periods of time to, for example, block a shot. These brief periods of time should be no longer than 3 seconds and must not be done continually as per the holding rule (see Section 4.5).
I take this to mean that you cannot repeatedly move to and from the blocking position which covers 50% or more of the goal to avoid the 3 sec rule.
- Note the following:
A goal is achieved when the entire ball (not only the center of the ball) goes over the field-side edge of the goal line, i. e. the ball is completely inside the goal area4. The restart after the goal shall adopt the same rules as the kick-off.
This is more stringent than the criteria I applied on assignment 2.
- One more highlight from the rules:
If the attacker enters the penalty area then the penalty shot is deemed unsuccessful. If the goalie leaves the penalty area then a goal will be awarded to the attacking team.
10-27-2007
- I will be collecting team and individual memos, as well as collecting copies of team code as in previous assignments. If possible, please provide estimates of the amount of time spent on assignments 1,2, and 3 both as a team and individually. These estimates won't affect your evaluations in any way.
10-28-2007
- Since we'll be concluding our work with the Aibos shortly, I encourage you to include any thoughts or comments on how the last three labs could be improved in your lab memo. I'd also like you to include thoughts or comments on using Tekkotsu. As with the time estimates, this is not required, and your responses won't affect your evaluations in any way.
10-29-2007
Based on W-L and then goal differential, the following teams made the semis: RUR, Oreo, ManU, and RoboCop. Congratulations to every team. You folks did a great job on a difficult assignment!
- Class tomorrow will meet in the lab.
10-30-2007
- Assignment 4 has been posted on the Assignment page. Also, I've added a Graphics section to the Notes page, discussing various approaches to creating the graphics you'll need for the next assignment.
11-1-2007
- Could someone from each team send me a pointer to the directory for your Assignment 3 source? Please include a README with the names of the behaviors you used for defence and offense. No other documentation is necessary. Thanks!
11-2-2007
- I have added some additional trace files for testing. Your program needs to handle localized.data and localized2.data. Please include movies of map generation for both files when you hand in your assignment.
11-8-2007
I've included clarifications on Assignment 5 under the Notes page. This is based on notes from last year's class but covers the majority of questions I've been fielding via email regarding the assignment. Also, if you have pointers to GUI code examples for Java that would be relevant to the assignment, let me know or post them on the wiki (something similar to the opencv examples for C++ I provided).
11-9-2007
- You can ignore the reference to the 15 meter map size for corrected odometry. The actual map is larger than 15 meters.
- When you email me the directory for your project, please use the subject "CS344R/CS393R Project 4 Directory". This will make it easy for me to find your project.
- I've updated the graphics section on the Notes page with specific pointers to how to create movies. I encourage you to share insights on both GUI and movie creation with your classmates, preferably on the wiki.
11-10-2007
- Dr. Kuipers asked me to post the following.
To: kuipers@cs.utexas.edu
Subject: recruiting PhD students and undergrad summer interns
From: Jeffrey Mark Siskind <qobi@purdue.edu>
Date: Fri, 09 Nov 2007 13:20:40 -0500
-------------------------------------------------------------------------------
I am looking to recruit PhD students interested in the following general areas:
- Artificial Intelligence/Cognitive Science
- Computational Linguistics/Natural Language Processing
- Computer Vision
- Machine Learning
- Programming Languages and Compilers
- Robotics
I also am looking to recruit undergrads interested in an internship during
summer 2008. I intend this internship to be a mutual exploration of the
potential for working with me as a future PhD student.
Representative active research projects include:
- Automatic Differentiation
ftp://ftp.ecn.purdue.edu/qobi/popl2007a.pdf
- Programming Environments
http://www.ecn.purdue.edu/~qobi/papers/october-white-paper.pdf
- Robotic Imitation
http://www.ecn.purdue.edu/~qobi/papers/imitation-white-paper.pdf
- Stochastic Grammars for Images and Video
http://www.ecn.purdue.edu/~qobi/papers/pami2007.pdf
More information is available at http://www.ece.purdue.edu/~qobi
I am also starting a new collaborative project to study the acquisition of the
phonemic/allophonic distinction by infants. This will involve both experimental
work with subjects, corpus gathering and analysis, and computational modeling.
Interested students should contact me by email.
--
11-11-2007
- mencoder should now be installed on all the lab machines.
11-15-2007
- I have posted the official Assignment 5 handout. Sorry for the delay.
- I would like to start collecting locker keys on Tuesday. When you turn in your key, make sure your locker contains everything it contained at the beginning of the semester (Aibo, charger, battery, memstick, ball).
11-19-2007
- Please bring your locker key to lecture so that I may begin collecting them.
11-21-2007
- In /home/stober/public/code I've put some code for drawing random samples from a Gaussian. See gaussian.hh.
11-27-2007
I was asked whether the corrected odometry in localized.data was generated using "scan-matching" SLAM as in lecture 15 (pg. 4). The answer is yes, but with some caveats. The hypothesis scoring (i.e. particle weighting) code in our lab includes many extensions that reflect research more recent than when localized.data was produced, hence the current method of generating corrected odometry may behave differently now than when the trace was produced. In fact, it may be possible to produce odometry that is more correct than the corrected odometry in localized*.data. As with the previous assignment, everything will be graded qualitatively, so exactly matching corrected odometry to some prior corrected odometry is unnecessary (and probably impossible).
- My office hours this Friday will be from 3:30pm-5:00pm.
11-29-2007
- When you email me your project location, please use the subject "CS344R/CS393R Project 5 Directory".
- Note the following from the Assignment 5 Handout. This comparison will be an important part of Tuesday's discussion.
You should also produce image files, showing the map that is created if you use the raw odometry (O) without localization.
12-05-2007
- My Friday office hours are canceled. I will be holding office hours from 3:30 to 5:00pm on Thursday for those who wish to discuss grading issues for Assignment 5.