Everyone trying to learn how to make their first physics engine is searching google madly to find all the confusing incomplete articles and stackoverflow posts instead of just finding this excellent video.
Great job on this video. I have tried to understand this theorem so many times, but you finally got me to actually understand it. I am following your tutorial series in java/Processing just because I wanna learn the physics-engine basics before I implement it into my render-engine. Keep it up man!
16:37 What is the reason for looping through verticesB? Isn't the first loop enough to determine if there is overlap? I tried without the second loop and collision detection still seems to work (borders turn red). Edit: OK, the answer was in the next video, it's for calculating the depth and the normal.
I would say that there are probably better algorithms for 3D. In 3D I would probably look at "GJK": en.wikipedia.org/wiki/Gilbert%E2%80%93Johnson%E2%80%93Keerthi_distance_algorithm
Does the direction of the normal matter for this test? I am thinking that an axis doesn't really have a direction; it is just sort of a definition for an infinite line in space We are projecting each vertex onto that line, and I am thinking that it doesn't matter which way that line is pointing per say
Yes! You are correct! We use the normal direction in later videos to resolve the collision in the correct direction. Pushing the shapes apart according to the direction of the normal or "axis." Thanks for the clarification!
This one is specifically 2d and yes it is easier in 2d. You can look up separating hyperplane theorem for possible use in 3d. But in 3d, I would probably use GJK.
![Polygon Collision Resolution - Let's Make a Physics Engine [06]](http://i.ytimg.com/vi/SUyG3aV_vpM/mqdefault.jpg)
![Polygon Collision Resolution - Let's Make a Physics Engine [06]](/img/tr.png)
![Circle Polygon Collisions Using Separating Axis Theorem - Let's Make a Physics Engine [07]](http://i.ytimg.com/vi/vWs33LVrs74/mqdefault.jpg)
![Friction Physics - Let's Make a Physics Engine [24]](/img/n.gif)
Everyone trying to learn how to make their first physics engine is searching google madly to find all the confusing incomplete articles and stackoverflow posts instead of just finding this excellent video.
This series is the absolute best resource for learning 2d physics systems. Thanks for making it, it really helped me alot!
Your welcome! Thanks for watching!
Wow, I’m impressed! Even though english isn’t my first language, you make it so easy to understand.
I'm so glad I've found your channel, I finally found someone that can explain this theorem in a way that I could understand, thank you very much!
Am currently working on a project that uses SAT, was able to understand you and translate the code into python. Subbed
Thank you so much for the series! You helped me tremendously with my school project.
Awesome! Glad to hear that
the best resources for learning sat i i have found, thanks!!!
Great job on this video.
I have tried to understand this theorem so many times, but you finally got me to actually understand it.
I am following your tutorial series in java/Processing just because I wanna learn the physics-engine basics before I implement it into my render-engine.
Keep it up man!
been following around, this is pretty neat. thank you for making these videos!
Glad you like them!
Great serie. Thank you. Greetings from Popayan, Colombia.
Pure gold. So many thanks for these.
You're welcome! Thanks for watching!
16:37 What is the reason for looping through verticesB? Isn't the first loop enough to determine if there is overlap? I tried without the second loop and collision detection still seems to work (borders turn red).
Edit: OK, the answer was in the next video, it's for calculating the depth and the normal.
Very clear!
Thanks for watching!
First of all great video, but I have a question, do you think SAT can be good to 3d collision? thanks anyway and great video!
I would say that there are probably better algorithms for 3D. In 3D I would probably look at "GJK": en.wikipedia.org/wiki/Gilbert%E2%80%93Johnson%E2%80%93Keerthi_distance_algorithm
@@two-bitcoding8018 Thanks!!!
Does the direction of the normal matter for this test?
I am thinking that an axis doesn't really have a direction; it is just sort of a definition for an infinite line in space
We are projecting each vertex onto that line, and I am thinking that it doesn't matter which way that line is pointing per say
Yes! You are correct! We use the normal direction in later videos to resolve the collision in the correct direction. Pushing the shapes apart according to the direction of the normal or "axis."
Thanks for the clarification!
Could you explain why this is better than checking intersecting lines + point in polygon?
I think the main advantage comes in the next step where we are able to easily obtain the normal of the collision from the separating axis.
@@two-bitcoding8018 Could line v line collisions potentially be faster than SAT collisions if there's a fast way to get the normal of collisions
Does it's work in 3 dimensional?
In 2D it's easier to find overlaping vertices
This one is specifically 2d and yes it is easier in 2d. You can look up separating hyperplane theorem for possible use in 3d. But in 3d, I would probably use GJK.
@@two-bitcoding8018 just read and watch what is GJK, I still don't understand how to determinate direction
@@bamsgian9759 Here's a good resource: ua-cam.com/video/DGVZYdlw_uo/v-deo.html
How do i know the dot product?
Check out this video on my channel for dot product and other math for physics:
ua-cam.com/video/TJVA2gKFxH0/v-deo.html
@@two-bitcoding8018 Thanks!
C sharp lol shit ahh well 😅