I think there is a typo. Otherwise, the argument on the magnitude of v becomes meaningless. When v is too big, the left side becomes negative since the condition is ≥ 0, it's violated.
Hello Jason! It was really great lecture, and it helped me a lot to understand the concept in the Safe RL field. I am studying self-driving tech in Tsinghua University, and I want ask you a small favor. I want to post this lecture by translating Korean into my blog. Could I post your lecture in Korean language? I want many Korean people to be able to understand this SafeRL field, and I think your lecture and the ppt slides are really good to be understood, as well as I also review the korean translated version of this lecture whenever I confuse concepts.
May I ask if it is a high-order system, such as a fourth-order system x=[x1,x2,x3,x4]. dot_x1=x2, dot_x2=x3, dot_x3=x4, dot_x4=f(x)+g(x)*u. I want x1 to be within a fixed boundary, such as (-boundary, boundary). Use CBF to solve the problem, select B(x)=boundary*boundary-x1*x1. Then when solving, I found that LgB(x)u=0, which is completely impossible to solve. How to solve this problem?
Does anyone have any resources (e.g books, papers) that expand on the idea that given F that is Lipschitz in x, continuous in u and u piece-wise continuous in t, THEN given an initial state vector, a state trajectory exists and is unique? I have been interested in this question for a while!
Thanks a lot. You made it easy to understand.
This was exactly what I was looking for, excellent! Thank you very much!
Nice presentation about Lyapunov Function and barrier function! That description rings bells with me about convex optimization and stability.,
Thank you very much for the detailed introductions about CBF. It is very interesting!
This is great. I finally understand CBFs
thank you for your excellent presentation
Very well explained! This can be useful for many people.
Very good video, thank you. I just have one question. How did you go from ≥ at 34:24 to ≤ at 54:55? When did that condittion flip?
I think there is a typo. Otherwise, the argument on the magnitude of v becomes meaningless. When v is too big, the left side becomes negative since the condition is ≥ 0, it's violated.
Exactly what I was looking for. Thanks for sharing.
Great Job
Thank you for sharing this enlightening lecture. I'm wondering if will could apply this to MIMO non affine problem in process control.
Can we use average dwell time for relevant results? What are the advantages of common Lyapunov function over average dwell time?
Hello Jason! It was really great lecture, and it helped me a lot to understand the concept in the Safe RL field. I am studying self-driving tech in Tsinghua University, and I want ask you a small favor. I want to post this lecture by translating Korean into my blog. Could I post your lecture in Korean language? I want many Korean people to be able to understand this SafeRL field, and I think your lecture and the ppt slides are really good to be understood, as well as I also review the korean translated version of this lecture whenever I confuse concepts.
안녕하세요 ~ 혹시 포스팅 하셨나요~? 궁금합니다!
May I ask if it is a high-order system, such as a fourth-order system x=[x1,x2,x3,x4]. dot_x1=x2, dot_x2=x3, dot_x3=x4, dot_x4=f(x)+g(x)*u. I want x1 to be within a fixed boundary, such as (-boundary, boundary). Use CBF to solve the problem, select B(x)=boundary*boundary-x1*x1. Then when solving, I found that LgB(x)u=0, which is completely impossible to solve. How to solve this problem?
Does anyone have any resources (e.g books, papers) that expand on the idea that given F that is Lipschitz in x, continuous in u and u piece-wise continuous in t, THEN given an initial state vector, a state trajectory exists and is unique? I have been interested in this question for a while!
You can check Nonlinear systems, Khalil, 2nd edition for an answer to this question.
@@farshadaali Thank you so much!
This is Picard-Lindelöf theorem.
16:51, s should be x
I believe the first CBF constraint in the toy example th-cam.com/video/_Tkn_Hzo4AA/w-d-xo.html should be >= 0 instead of
Pedestrian at 2:10 should have to invest in the R&D's company
Cool talk
6:44
Invariant?
17:34
22:55 graph