HW3 Graded | |
Hi guys, HW 3 has been graded - thanks for the hard work. While we didn'r explicitlt ask for super-fast code, we do expect to you use the algorithmic inruition we developed throughout the course to make sure your planners are efficient. I used best planning times as a reference and then created a threshold (e.g., best average inspection planning times were around 1-3 seconds so I created a threshold of 60 seconds). Anyone above this threshold lost 5 points (5 for motion planning, 5 for inspection planning to a 50% coverage and 5 for inspection planning to a 75% coverag). In addition if the explanation of you improvements was unclear / VERY informal, I deduced 5 extra points. This means that (assuming you submitted all the results), your grade should be either 80, 85, 95, 95 or 100. If you would like additional feedback, feel free to reach out to me. Hag Sameah, Oren |
פורסם ב-16/4/2024, 08:20:37 Created on 16/4/2024, 08:20:37 Создано16/4/2024, 08:20:37 تم النشر ب-16/4/2024, 08:20:37 |
הקלות למילואימניקים בבחינה | |
כל מי ששירת מעל 21 יום מילואים במהלך הסמסטר רשאי להיבחן עם חומר פתוח. ע"מ למנוע עוגמת נפש, אם אתם לא בטוחים אם אתם עומדים בקריטריון, אנא פנו אלי ישיקות. בברכת חג שמח, תקופת מבחנים קלה והכי חשוב - שנדע ימים טובים יותר אורן |
פורסם ב-15/4/2024, 15:13:39 Created on 15/4/2024, 15:13:39 Создано15/4/2024, 15:13:39 تم النشر ب-15/4/2024, 15:13:39 |
[Lab pilot] - Minor correction | |
It appears that there is an error in one of the configurations provided in Lab3. Please replace the following value in task3_run.py: cube4_approach = np.deg2rad([97.6, -38.3, 52.1, -100.8, -90.1, 8.5]) Apologies for the inconvinience, Oren |
פורסם ב-7/4/2024, 11:18:50 Created on 7/4/2024, 11:18:50 Создано7/4/2024, 11:18:50 تم النشر ب-7/4/2024, 11:18:50 |
[Lab pilot] - Meeting for third task (reminder) | |
Hi all, This is a reminder that we will meet on Wednesday April 10th at 12:30 at the CRL for running your final paths in the lab. Oren |
פורסם ב-7/4/2024, 08:55:15 Created on 7/4/2024, 08:55:15 Создано7/4/2024, 08:55:15 تم النشر ب-7/4/2024, 08:55:15 |
[Lab pilot] - Registration | |
Hi guys, It came to my attention that some of you may that are participating in the pilot did not register to 236006. On 5.2.2024 I sent a message asking anyone not registered yet to register directly and asking you guys to tell me if there is any issues. It is VERY late to handle registrations at this point but we will sort things out. If you are not registered, send me an email ASAP so I can see how many students are we talking about and how to handle this confusion. Best, Oren |
פורסם ב-7/4/2024, 08:54:34 Created on 7/4/2024, 08:54:34 Создано7/4/2024, 08:54:34 تم النشر ب-7/4/2024, 08:54:34 |
[Lab pilot] - Task 2 graded | |
Hi guys, Task 2 was graded. On the one hand, clearly everyone implemented the task and did a good job. On the other hand results and plot were all over the place. I take a big part of the responsibility but when one writes plot a as a function of b then a is the y-axis and b is the x-axis. In our setting, the idea was to see how the quality of the path of RRT* changes over time and to see if some parameters allow for an initial path fast but fail to improve path quality. Since no one did this, I guess that the instructions were not clear enough (that is why we offered this as a pilot). In terms of grades (1) Everyone who did the full task and gave a path that can be executed on the robot got 100 for this task. (2) Grading for the final project will be follow an EXTREMELY high bar. Most of you will likely get close to 100 for the lab pilot. This means that a grade of 80 in the project will yield a final grade of 90 (which should be the average grade for the project). So do not feel frustrated if you get a grade of 85 or 90 after working very hard. Final grades will be high... Oren |
פורסם ב-2/4/2024, 18:45:10 Created on 2/4/2024, 18:45:10 Создано2/4/2024, 18:45:10 تم النشر ب-2/4/2024, 18:45:10 |
Sol to Exam 1 | |
Hi guys, Thanks again for making the course a great experience for me. I uploaded a solution to Exam 1 which may contain slight errors. It was written for my TA and NOT to be published. Having said that, I wanted to give you some solutions to work with. Since this is a relatively new course and the exam was not given many times, there is a small pool of previous exams that I can share and I apologize for that. Good luck, Oren |
פורסם ב-2/4/2024, 11:40:40 Created on 2/4/2024, 11:40:40 Создано2/4/2024, 11:40:40 تم النشر ب-2/4/2024, 11:40:40 |
Exams | |
Under course material, I added two exams from previous years. Before the class tomorrow, please go over them. Oren |
פורסם ב-31/3/2024, 09:36:36 Created on 31/3/2024, 09:36:36 Создано31/3/2024, 09:36:36 تم النشر ب-31/3/2024, 09:36:36 |
[Lab pilot] - Meeting for third task | |
Hi all, We will meet on Wednesday April 10th at 12:30 at the CRL for running your final paths in the lab. I will extend the submission deadline for the 11th so you can submit the final video together with your writeup. Oren |
פורסם ב-28/3/2024, 14:25:56 Created on 28/3/2024, 14:25:56 Создано28/3/2024, 14:25:56 تم النشر ب-28/3/2024, 14:25:56 |
HW2 Graded | |
Hi guys, I finally finished checking the HW. Again sorry for the delay. In general, the submissions were top quality and I enjoyed checking them. There were several recurring mistakes, I list them below with the points deducted. If you got points deducted due to one of these issues, I did not return your HW. If you got points deducted due other reasons, I returned your HW together with feedback. In addition, under course material I added a complete solution to the problem of planing for a tethered robot. Oren ************************ Distribution of points in high-dimensional spaces (*) Some of you plotted \eta as a function of \epsilon for different values of d (and not \eta as a function of d for different values of \epsilon). 3 points were deducted in this case. (*) In the discussion, you were expected to pinpoint that the constants used when connecting points in a sampling-based algorithm really matter. If you can shave even a bit in r (the connection radius), it matters a lot when d is large. 2 points were deducted in this case. Tethered robots (*) You were expected to *explicitly* mention that the path of the robot and the tether lie on the visibility graph. 1 point was deducted in this case. (*) You were expected to *explicitly* mention that the tether encoding can done be done using its homotopy class. 1 point was deducted in this case. (*) Some of you ran the Dijkstra-like search from p_s and not p_b In the pre-processing step you don't have p_s or the tether's initial homotopy class. 3 points were deducted in this case. (*) You were expected to *explicitly* state that the connecting for p_s will account for the homotopy path (as we know the initial tether signature) while for p_t we are only given a goal point and not the signature so we need to connect to neighboring vertices regardless of the signature (assuming the tether length is long enough). 3 points were deducted in this case. (*) Some of you gave insufficient detail / informal high-level description in 3.4 3 points were deducted in this case. Motion Planning: Search and Sampling (*) You were asked to plot (1) the success rate to find a solution as a function of time and (2) the quality of the solution obtained as a function of time for one representative run If you design the experiment correctly, both the success rate and the solution quality should monotonically increase (or the cost should monotonically decrease). Giving plots where this does not happen means that your experiment was off (e.g., instead of sampling solution quality at different times, you re-ran the algorithm from scratch using different seeds). You were expected to note such flaws in your experiment. 4 points were deducted if you presented a plot with decreasing success rate. 4 points were deducted if you presented a plot with increasing path length. 3 points were deduced in this case. |
פורסם ב-19/3/2024, 15:14:15 Created on 19/3/2024, 15:14:15 Создано19/3/2024, 15:14:15 تم النشر ب-19/3/2024, 15:14:15 |
HW3 published | |
HW3 published and is due 7.4. For any technical assistance (such as Python-related questions), please contact our TA Dean Zadok (udeanzadokatcampus.technion.ac.il) Good luck! Oren |
פורסם ב-17/3/2024, 13:32:16 Created on 17/3/2024, 13:32:16 Создано17/3/2024, 13:32:16 تم النشر ب-17/3/2024, 13:32:16 |
[Lab pilot] - Third task published | |
Third task is published - I added a task to the HW section under 236901. For any support, please contact Ori Manasherov (uori.manatef.technion.ac.il) and cc me. Good luck, Oren |
פורסם ב-17/3/2024, 13:04:33 Created on 17/3/2024, 13:04:33 Создано17/3/2024, 13:04:33 تم النشر ب-17/3/2024, 13:04:33 |
[Lab pilot] - Meeting | |
Hi guys, If you remember, I mentioned that we will meet this Wednesday at 12:30 in the lab to run your pre-computed paths on the robot. I prefer to stick with this schedule but if you don't think that you can make the deadline, please let me know via email. Best, Oren |
פורסם ב-10/3/2024, 15:00:02 Created on 10/3/2024, 15:00:02 Создано10/3/2024, 15:00:02 تم النشر ب-10/3/2024, 15:00:02 |
[Lab pilot] - Feedback | |
Hi guys, I checked the first submission of the lab pilot. In general, you all did great with some VERY minor hickups. Unless you got feedback, you received a grade of 100. If you have some costructive feedback on this part, please send it over and cc Ori. Best, Oren |
פורסם ב-10/3/2024, 14:57:27 Created on 10/3/2024, 14:57:27 Создано10/3/2024, 14:57:27 تم النشر ب-10/3/2024, 14:57:27 |
HW2 extra extension (1 day) | |
Following some requests I added 1 day to the deadline for HW2. Oren |
פורסם ב-5/3/2024, 20:24:43 Created on 5/3/2024, 20:24:43 Создано5/3/2024, 20:24:43 تم النشر ب-5/3/2024, 20:24:43 |
HW2 - Clarifications | |
Hi everyone, Following a few emails we received in the last couple of days, we wanted to clarify a few things: 1. Regarding RRT* with k = log(n): Note that here k changes as the number of nodes in the tree increases, so make sure to adapt your code, such that when no k is provided in the RRT*, compute log(n) and feed it to get_k_nearest_neighbors. 2. Regarding the termination criteria for the two planners: (a) When plotting the success rate to find the solution in real time, you can terminate the algorithm once a solution is found. (b) When plotting the quality of the solution as a function of time, change the termination criteria to be some large time limit and obtain intermediate values of (time, quality of solution). 3. Submission: You are required to submit a PDF document with the results, and your code in one ZIP file. Good luck! Course Staff |
פורסם ב-3/3/2024, 12:27:39 Created on 3/3/2024, 12:27:39 Создано3/3/2024, 12:27:39 تم النشر ب-3/3/2024, 12:27:39 |
HW2 Extensiom | |
Hi guys, I broke my finger Thursday and was totally unresponsive until today. Some of you sent questions about the HW and got no reply so I would like to give you a few extra day (until Wed.6.3). I'll try to respond to questions ASAP. Oren |
פורסם ב-3/3/2024, 09:08:55 Created on 3/3/2024, 09:08:55 Создано3/3/2024, 09:08:55 تم النشر ب-3/3/2024, 09:08:55 |
HW1 Graded | |
Hi guys, I finally got around to grading HW1, sorry for the delay. In general, great job! Here is some feedback. If you want to see exact feedback on your HW, just shoot me an email. (1) Informal response to Q1. E.g., what is the MK of a line and a point? The answer "a shifted line" does not suffice. Explain what the shift is. Similarly, what is the MK of two disks? The answer "a disk" does not suffice. Explain that this disk has a center that is the MK of the centers of the original disks and with radius eqyal to the sum of the two disks radii. (2) Missing additional scenarios for the second exercise. You were explicitly asked: "... generate one more instance. Show results for both settings (the one provided and the one you generated)." This caused a deduction of 5 points per section (up to 15 points in total). (3) Missing / wrong complexity analysis -5 pts. (4) No visualization provided: -4 pts per section (up to 12 points in total). (5) Missing complete answer to non-convex setting: How can non-convex obstacles affect the results? Why? When? Support your claim with examples. Some students did not support their claims with specific examples: -4pts |
פורסם ב-25/2/2024, 15:26:33 Created on 25/2/2024, 15:26:33 Создано25/2/2024, 15:26:33 تم النشر ب-25/2/2024, 15:26:33 |
[Lab pilot] - Extension | |
Hi guys, Following a few requests, I extended the submission date for the first exercise in the lab pilot until Thursday. As I mentioned earlier, please send all technical questions to Ori (uori.manattechnion.ac.il). Best, Oren |
פורסם ב-25/2/2024, 09:18:35 Created on 25/2/2024, 09:18:35 Создано25/2/2024, 09:18:35 تم النشر ب-25/2/2024, 09:18:35 |
[Lab pilot] - Clarifications | |
Hi guys We would like to provide some insights into the local planner's functionality, specifically regarding the determination of the validity of transitions between configurations and collision checking. I know that it was unclear and this may have delayed your progress (we mentioned in advance that this is a pilot and there may be hickups). If you need a few extra days, let me know. The self.resolution parameter plays a key role in this process by influencing the number of configurations checked for collision. We can set the number of intermediate configurations as a fixed number. for example 2, and then you check for collision only the configurations themselves without intermediate configurations. If the fixed number is 3, we check for collision for the configurations themselves and one intermediate configuration. To enhance the efficiency of collision checking, instead of using a fixed number of intermediate configurations, we have implemented a more dynamic approach where the number of configurations to check is a proportional number to the difference between the two configurations. Here's a brief overview: 1. Determining the Number of Configurations to Check: - Calculate the number_of_configurations_to_check = (maximum angle difference between two configurations) / self.resolution. - Ensure that the number_of_configurations_to_check is at least 3; if not, set it to 3. 2. Generating Configurations for Collision Checking: - Utilize the numpy.linspace function to obtain the configurations to be checked. set endpoint parameter to True. 3. Collision Checking: - Verify for collision-free at each configuration. 4. Local Planner Decision: - If no collision is detected for all the configurations, the local planner returns True. This approach allows for a more flexible and context-aware determination of the number of configurations to check, leveraging the self.resolution parameter. It ensures a balance between computational efficiency and accuracy in collision detection. |
עדכון אחרון ב-22/2/2024, 10:14:56 Last updated on 22/2/2024, 10:14:56 Последняя модификация22/2/2024, 10:14:56 تمت الحتلنة الأخيرة ب-22/2/2024, 10:14:56 |
[Lab pilot] - First task extension and update | |
Hi guys, There have been some questions regarding the resolution to use. I would like Ori who takes care of all the implementation details to answer these questions but he will only be availanle starting Wednesday. Thus (i) we will publish official clarifications on Wednesday and (ii) I extened the submission date to Sunday 25.2. Thanks for understanding, Oren |
פורסם ב-19/2/2024, 21:52:51 Created on 19/2/2024, 21:52:51 Создано19/2/2024, 21:52:51 تم النشر ب-19/2/2024, 21:52:51 |
HW2 published | |
HW1 published and is due 3.3. For any technical assistance (such as Python-related questions), please contact our TA Dean Zadok (udeanzadokatcampus.technion.ac.il) Note that we did not cover all of the material yet (especially second dry question) but you can already start working on the wet exercise. Good luck! Oren |
פורסם ב-18/2/2024, 15:07:00 Created on 18/2/2024, 15:07:00 Создано18/2/2024, 15:07:00 تم النشر ب-18/2/2024, 15:07:00 |
[Lab pilot] - First task published | |
*** Note - this message is only relevant to studennts enroled to the lab pilot *** Hi guys, (1) Registration - I have all of you recorded in my file. To formally register, Dr. Sarah Keren opened the registration to all and you can add yourself to the course (remember, the course number is 236006). If you have any problem, just reach out to me. (2) First task is published - I added a task to the HW section under 236901. For any support, please contact Ori Manasherov (uori.manatef.technion.ac.il) and cc me. Good luck, Oren |
פורסם ב-5/2/2024, 07:15:16 Created on 5/2/2024, 07:15:16 Создано5/2/2024, 07:15:16 تم النشر ب-5/2/2024, 07:15:16 |
HW1 update: correction for obstacles position | |
Dear students, There was a mistake in the order of the points of the third obstacle in the file 'obstacles'. The order should have been counterclockwise and not clockwise. The file Wet.zip was updated in the assignment section. Those who fixed it independently and/or adapted the code to work with both options may continue using their version. Thanks to the students who found it and good luck! Course Staff |
פורסם ב-3/2/2024, 18:53:44 Created on 3/2/2024, 18:53:44 Создано3/2/2024, 18:53:44 تم النشر ب-3/2/2024, 18:53:44 |
HW1 published | |
HW1 published and is due 15.02. For any technical assistance (such as Python-related questions), please contact our TA Dean Zadok (udeanzadokatcampus.technion.ac.il) Good luck! Oren |
פורסם ב-30/1/2024, 14:01:19 Created on 30/1/2024, 14:01:19 Создано30/1/2024, 14:01:19 تم النشر ب-30/1/2024, 14:01:19 |
Registration to pilot project | |
Hello, Thanks for whoever sent an email regarding to the pilot project. I did not respond personally to each email but have noted all emails. If you want to participate in the pilot project, send me an email by Sunday 21.1, end of day. Thanks, Oren |
פורסם ב-18/1/2024, 14:31:47 Created on 18/1/2024, 14:31:47 Создано18/1/2024, 14:31:47 تم النشر ب-18/1/2024, 14:31:47 |
(one time) change in class | |
Tomorrow (only), our class will be held in Ulman 309. Oren |
פורסם ב-14/1/2024, 12:35:21 Created on 14/1/2024, 12:35:21 Создано14/1/2024, 12:35:21 تم النشر ب-14/1/2024, 12:35:21 |
Accompanying project | |
Hi guys, The course is currently 2 academic points and we are in the process of extending it to 3 academic points via a workshop that includes extra hansd-on experience with a robotic manipulator. You will not enjoy this extra academic credit but you may participate in a pilot as part of a project (under course number 236006). If you are interested (this is completely up to you), you can register to the project (the project gives 3 academic points). The project includes two parts (submission can be done in pairs): Part 1) This will be the workshop on the the manipulator where you will have to complete three exercises in Python and then run your planner on the UR5 manipulator we have in the lab. This part will be identical to every student registered to the project Part 2) This will be a small project where you will have to extend the basic parts implemented in the first part. This will be different for each group. I will explain this in the first lecture as well but wanted to give you a heads up so you can plan your semester accordingly. Oren |
פורסם ב-3/1/2024, 17:38:24 Created on 3/1/2024, 17:38:24 Создано3/1/2024, 17:38:24 تم النشر ب-3/1/2024, 17:38:24 |
Welcome to Alg Motion Planning (236901) | |
Dear students, We are starting a new semester in one of the hardest times we have all faced as students, as adults and as a society. My name is Oren and I will be your lecturer this semester where, together with the amazing staff we have, we will navigate these hard times together. First and foremost, I would like to let you know that we are here for you. We don't have all the answers upfront and we will have to be adaptive. I urge you to reach out as early as possible with any concern or problem that you have. I uploaded all the lectures (slides and videos) from previous semesters to the website to allow you to prepare in advance or to make up for any material missed. However, these are subject to change (no material will be added but we will remove some material - we typically have a guest lecture and I give a LONG intro which will be reduced). I urge you to make every effort to come to class in person - the added value of in-person classes is invaluable. Both from an academic perspective and from a social perspective, bringing normality back to our lives in these abnormal times. This semester we are starting a pilot on hands-on experience with real manipulators. Details are given in separate message May we see peaceful times as soon as possible, Oren |
פורסם ב-3/1/2024, 17:36:58 Created on 3/1/2024, 17:36:58 Создано3/1/2024, 17:36:58 تم النشر ب-3/1/2024, 17:36:58 |