I got comments back on my worksheet 4 (shown below):
-
Sounds good to me. The biggest issue I would see with this is actually nothing to do with how older games approximate physics, as it's more an issue of scale. A simple geometric shape like a cone, sphere, or box is simple as you can simply check against the parametric surface and ignore the mesh itself, but as you start trying to do the same physics calculations upon meshes made up of multiple convex and concave shapes it's really only feasible to do “perfect” simulations when there is very little to nothing else being done by the CPU and even the most high-end gaming PC can probably only handle a small amount of low-poly objects. With the point mass and sphere you will most likely find it to be quite simple to do but the real trick will be to optimize your calculations to handle hundreds or thousands of point masses or spheres at a time, preferably without causing a bottleneck on the CPU which prevents your application from doing anything else. So, basically, what I'm saying is there is no question of whether or not you can perform “perfect” simulations, but whether or not you can perform those simulations in a game on objects which actually matter to the game.
-
Also in the email was this comment:
-
I hope I don't sound like I am being too critical of your project. It can be a really good project but from what you've written in the worksheet it sounds like either you expect problems and so don't wish to say your going to do too much for the project (which isn't really bad), or you don't know much about programming physics simulations and are drastically underestimating the computing power you'll be working with.
-
--------------------------------------------------------------------------
-
As I mentioned before, I have had worries and doubts about the project and these comments sort of made me feel it more. I am actually glad that someone else could see it too and so I wanted to find out more. I emailed him saying that I was having these feelings towards the project and that I understood where he was coming from but I wondered if he had any advice on what to do about it. He emailed back with:
-
Well if you still want to do the same project but put a little more into it, there are quite a number of methods for either parametrization of mesh surfaces to better fit their usage in complex calculations that you could investigate. Simply changing your question to involve improving physics calculations efficiency as opposed to asking if it's possible would easily be good enough though. For instance, instead of your question of "Can the movements and collisions of an object on a curved surface be realistically simulated in real-time?", you could change it to something along the lines of 'How can physics be simulated realistically without reducing frame-rate to less than "real-time" rates?' With a question similar to that you could easily do exactly as you mentioned in the "Addressing the Question" section of your worksheet #4 and perhaps focus more on what can be calculated exactly and what can be approximated without losing realism and still make it run fast. A single point-mass or sphere interacting with a bezier surface would probably be difficult to use to benchmark any calculations but you could probably have lots of objects interacting with the same surface without too much problem, or even those objects and the surface interacting with everything in the scene(probably much harder).I hope this helped. You really don't need to change your project or anything, I just thought that perhaps it would be difficult to write a good dissertation on something which could probably be answered as a yes or a no.
-
This seemed a little better to me and got me thinking about changing the question to the above mentioned, as I could see from that version that there was a way of obtaining results if you like and drawing a conclusion. I did decide to tweak the question however, it would not be completely final until I had a work with my supervisor. I am due to see Dr Lucas on Friday to show him my draft presentation and so this would give me the perfect chance (and my final chance!) to speak to him about my worries over the project, ahead of the presentations on Monday.
-
-----------------------------------------------------------------------
-
Just had my meeting with Dr Lucas. He was a little confused by the comments I was given and where I was thinking of changing the project. Eventually, he showed me that it was perfect the way it was. We discussed my doubts on how I would be able to answer the question. He said that I did not really need to look into frame rates as such and just the fact that the ball bounced on the curved surface, showed that the question was answered. It really went back to something I already read and knew. When it comes to physics, 'if it looks right then it is right'. Therefore, final implementation will show that realistic movement can be implemented and I can take into account the limitations in the dissertation. Dr Lucas had a look at my draft presentation too and was fairly happy with it however, he did add a couple of things (such as issues that may come up) to it that need to be put into another slide perhaps. Overall, I am much happier. Even though in the end, the question was not changed, I felt much better about how my project would turn out and how I will gather the information I will need, form a simulation and write up my findings in a concise and structured manner. Now to do the presentation!!
No comments:
Post a Comment