Post seminar presentations
I would like to bring to your attention that there are two super-relevant talks in the CS seminar that may be of interest to you.
The first is Ofir Gordon, titled "Toward Understanding the Hardness of Multi-Agent Path Finding”.
The talk will be held via zoom.
Time and Zoom Link: 02/09/2021, 14:00, https://technion.zoom.us/j/7569145171
You can find the complete details in the following link: https://www.cs.technion.ac.il/events/2021/10099/
The second talk is by by Nir Greshler, titled "Cooperative Multi-Agent Path Finding".
The talk will be held via zoom.
Time and Zoom Link: 05/09/2021, 14:30, https://technion.zoom.us/j/3454121052
Abstract is below:
In this research, we introduce and study the Cooperative Multi-Agent Path Finding (Co-MAPF) problem, an extension to the classical MAPF problem, where cooperative behavior is incorporated. The classical MAPF problem is inherently cooperative since each agent has to arrive at its goal, without colliding with other agents. However, in many real-world applications, agents that operate in a shared environment are often heterogeneous and may have a different set of abilities and restrictions. Therefore, in the Co-MAPF framework, achieving goals and completing tasks may not depend only on avoiding collisions between agents, but also on actively coordinating their actions. Simply put, we may want agents not just to not interrupt each other, but also to help each other achieve their goals. We term this a truly cooperative setting. In this setting, a group of autonomous agents operate in a shared environment and have to complete cooperative tasks while avoiding collisions with the other agents in the group. To complete cooperative tasks, agents must collaborate and coordinate their high-level decisions. This introduces a significant computational challenge on top of path planning and collision avoidance. This extension naturally models many real-world applications, where groups of agents are required to collaborate to complete a given task. To this end, we formalize the Co-MAPF problem and introduce Cooperative Conflict-Based Search (Co-CBS), a CBS-based algorithm for solving the problem optimally for a wide set of Co-MAPF problems. Co-CBS uses a cooperation-planning module integrated into CBS such that cooperation planning is decoupled from path planning. We suggest two improvements to Co-CBS that significantly improve its success rate. We also address the Task Assignment (TA) problem, which is NP-hard in this context. We propose to formulate the TA problem as a Multi-Index Assignment Problem (MIAP), use an off-the-shelf algorithm to solve it, and show how to integrate it into Co-CBS. Finally, we present empirical results on several MAPF benchmarks demonstrating our algorithms properties.
|פורסם ב- 24/8/2021, 15:25:41 Created on 24/8/2021, 15:25:41 Создано 24/8/2021, 15:25:41 تم النشر ب- 24/8/2021, 15:25:41|
Final report and final grades
In general, the reports were great and I enjoyed reading them.
The grades (as expected), were high with most students getting the maximum 30 points and the average grade being 28.875 out of 30.
The reasons that points were reduced can be found in the following legend:
(1) Key takeaway oversimplstic - it cannot be something like "paper is not realistic enough". That is not a takeaway or an insight.
(2) Project not connected to the paper's topic
(3) Evaluation metrics in the project are not clearly defined e.g., "will it serve it's purpose" is vauge or "Try to do something beyond reproducing the article results,"
(4) No key takeaway or insight, only missing details / extensions.
For each of the aforementioned reason, two points out of a total of 30 were reduced.
If you did not get a perfect score (30), the exact reason is given below with the first column giving the lst two digits of your ID and the second column the explanation (use legend) above.
Finally, the final grades will be submitted shortly and include a small 1-point factor.
|פורסם ב- 24/8/2021, 15:18:24 Created on 24/8/2021, 15:18:24 Создано 24/8/2021, 15:18:24 تم النشر ب- 24/8/2021, 15:18:24|
Final report extension
|The submission of the final report is extended by a month - oficial due date is 22.8.21.|
Good luck with your exams,
|פורסם ב- 1/7/2021, 16:01:44 Created on 1/7/2021, 16:01:44 Создано 1/7/2021, 16:01:44 تم النشر ب- 1/7/2021, 16:01:44|
Reminder - make up lesson tomorrow
This is a reminder that tomorrow we will have a virtual make up lesson at 12:30 for one hour.
Please do everything you can to join the lesson.
|פורסם ב- 29/6/2021, 21:55:55 Created on 29/6/2021, 21:55:55 Создано 29/6/2021, 21:55:55 تم النشر ب- 29/6/2021, 21:55:55|
Final report guidelines
|Here are some guidelines for the final project (remember, it is 30% of your grade and should be submitted by 22.7).|
Please pick a paper / topic that you did not present
(1) Summarize the paper in your own words - This is supposed to be in high level but *use your own words*. Assume the reader is knoledgable in CS and AI and know the definition of MAPF and nothing more. The summary should be as sewlf-contained as possible. As a rule of thumb,a one-page summary suffacies.
(2) Following the summary, try and pinpoint what are the key takeaways and insights and note on any missing details or interesting extensions. This should be a couple of paragraphs long but shows the difference between repeatng the paper's some "deeper understanding"
(3) Design in detail a project around the paper. This should be roughly one page long snd should include the following points
(-) What are the outcomes (implement the approach is not a very good project)?
(-) What are the expected results?
(-) What tools are required to implement the project? If you are aware of open source implementations that could be used to start working on the project that would be great. You can check out in mapf.info if something exists.
As a sanity check, ask yourself if you would be excited to do this project (is it feasible? challenging? interesting? Is it clear what needs to be done and how?)
Each part will be given 10 points out of the total 30 points the project accounts toward your final grade.
|פורסם ב- 13/6/2021, 10:49:16 Created on 13/6/2021, 10:49:16 Создано 13/6/2021, 10:49:16 تم النشر ب- 13/6/2021, 10:49:16|
Technion Robotics Seminar
Guy Hoffman will be talking today at the Technion Robotics Seminar.
Details can be found at https://tasp.technion.ac.il/news-events/.
He is an amazing speaker (and researcher) with a top-ten most-popular TED talk for 2014.
He works on robot design and I'm sure the talk will be a blast.
See you (virtually) today at 15:30.
|פורסם ב- 2/6/2021, 07:26:35 Created on 2/6/2021, 07:26:35 Создано 2/6/2021, 07:26:35 تم النشر ب- 2/6/2021, 07:26:35|
Class physically taking place today
Class will be held today as usual.
Looking forward to seeing you,
|פורסם ב- 31/5/2021, 07:21:35 Created on 31/5/2021, 07:21:35 Создано 31/5/2021, 07:21:35 تم النشر ب- 31/5/2021, 07:21:35|
No class next week
First, I want to wish you all a happy, safe and calm holiday - may we see the flames going down and tranquility return to our homes.
Second, we will not be meeting next week as there will be no AC at Taub building.
See you on May 31st in class.
|פורסם ב- 16/5/2021, 09:46:44 Created on 16/5/2021, 09:46:44 Создано 16/5/2021, 09:46:44 تم النشر ب- 16/5/2021, 09:46:44|
Updated grades for mini project
In grading the final project I deduced points for implementations that did not produce all the results within a time limit.
This was done since I can't run your code on my computer forever (classic decidability problem) so I had to add a timeout.
Having said that, I told you in class that efficient code wasn't a requirement and this may have caused students to settle on highly sub-optimal implementations.
I should have stated in advance that I will test your code with some reasonable timeout (remember that I took the average running time among the 5 fastest implementations and multiplied this by 5) and I apologize for not doing so.
Thus, I added 4 points to every project for which points were deducted due to a timeout.
|פורסם ב- 10/5/2021, 14:15:00 Created on 10/5/2021, 14:15:00 Создано 10/5/2021, 14:15:00 تم النشر ب- 10/5/2021, 14:15:00|
Mini project graded
|Project grades are available.|
Some notes on the grading process:
Everyone who submitted the exercises, had running code with reasonable answers to the dry part got 50 points.
For the remaining 50 points, I took the average of the 5 fastest submissions (all of whom got all the answers correct) and multiplied this running time by five. This value (around five minutes) was used as a timeout. I then ran each submission and for every correct answer that was obtained under this timeout, one point (out of 50) was given.
Almost everyone got more than 40 right solutions within the timeout so both the median and average grade were above 90.
|עדכון אחרון ב- 27/4/2021, 12:41:18 Last updated on 27/4/2021, 12:41:18 Последняя модификация 27/4/2021, 12:41:18 تمت الحتلنة الأخيرة ب- 27/4/2021, 12:41:18|
Student presentations are in class (!)
I am happy to announce that student presentations will be delivered in hybrid mode.
If you are presenting
(1) if for some reason choose not to present in class (highly discouraged), let me know at least three days in advance.
(2) please come to class early to set up your computer - it takes time to set things up with the hybrid format.
See you next week,
|פורסם ב- 20/4/2021, 12:00:17 Created on 20/4/2021, 12:00:17 Создано 20/4/2021, 12:00:17 تم النشر ب- 20/4/2021, 12:00:17|
Please fill in the following form regarding hybrid vs virtual presentations
|פורסם ב- 19/4/2021, 10:53:09 Created on 19/4/2021, 10:53:09 Создано 19/4/2021, 10:53:09 تم النشر ب- 19/4/2021, 10:53:09|
Mini project submission date postponed to 25/4
|פורסם ב- 12/4/2021, 11:13:41 Created on 12/4/2021, 11:13:41 Создано 12/4/2021, 11:13:41 تم النشر ب- 12/4/2021, 11:13:41|
Technion Robotic's seminar
|For anyone interested, the Technion Robotic's Seminar is back.|
The seminar will take place via zoom so there won't be in-person interaction (or Pizza) but we will have lectures given by prominent reserachers from all around the world.
For additional details, see https://robotics-seminars.net.technion.ac.il/.
|פורסם ב- 6/4/2021, 09:38:40 Created on 6/4/2021, 09:38:40 Создано 6/4/2021, 09:38:40 تم النشر ب- 6/4/2021, 09:38:40|
Lecture 1 recording
|Lecture 1 recording is available.|
Sorry for the delay and Hag Sameah,
|פורסם ב- 29/3/2021, 04:44:33 Created on 29/3/2021, 04:44:33 Создано 29/3/2021, 04:44:33 تم النشر ب- 29/3/2021, 04:44:33|
Welcome to robotics seminar & zoom link
Welcome to the robotics seminar on Multi-Agent Path Finding (MAPF).
Our first meeting will take place on Monday, 10:30 at Taub 4 and virtually using the zoom link
(also available in the course material section).
We will use the same zoom link for all hybrid / virtual meetings.
Looking forward to seeing you,
|עדכון אחרון ב- 17/3/2021, 14:01:58 Last updated on 17/3/2021, 14:01:58 Последняя модификация 17/3/2021, 14:01:58 تمت الحتلنة الأخيرة ب- 17/3/2021, 14:01:58|