I’m playing around with an RTS. I’ve got most for the basic interface coded. The program uses pro-rsoft’s PGMM to generate a heightfield mesh- uses a different image to texture the mesh, then uses a third image as a grid for obstacles on that map. The grid information is used by a python based A* script (pathfinding through obstacles/mazes).
Everything works fine- units can be selected with a mouse click or a selection box- they can be told to move to any location on the map- the A* algorithm provides a path between the current position and the destination.
The problem is speed. The python based A* script is just too slow. The image grid is 64x64- so it’s not gigantic- but selecting multiple units each having to run their own A* routine to find a path between their individual position and the destination coordinates noticably stalls the computer (Core 2 Duo) until the A* script returns the path information.
I’m aware of a many C A* scripts on the internet but I don’t know a thing about C and I don’t know how to wrap them so that my python script could make reference to the C script instead of the python A* script I’m currently using.
Another way you can use is to make the a-star algorithm non-blocking, this way the ai-controlled objects stay still for a while (during calculation) but will not block the running game.
Another way is to use a different approach to path-finding. Walk in the direction of the target until you hit a wall, then follow the wall until you have a closer position then the one on the first collision.
But if you manage to make a c extension for panda calculating a-star for a python array, i and possibly other would be happy about it
Panda3D provides everything you need to make C/C++ extensions to Panda3D.
Take PGMM, for example. In Python it wouldn’t be as fast as it is now, so I’ve implemented it into C++. You basically need to do the following steps:
[] Create a header file, source file and Inline functions file. (e.g. pgmm.h, pgmm.cxx, pgmm.I). Put in there the source you like.
[] Interrogate your program. Interrogate is a Panda3D module which generates python bindings for your program. If you want, I can give you the shell script I use to generate interrogate wrappers for PGMM.
This process basically consists of two steps: interrogate, which generates e.g. a pgmm_igate.cxx, and the interrogate_module call, which generates a pgmm_module.cxx file.
[] Nextly, you compile all your .cxx files, including the two you got from interrogate. How? Check this or take a look at PGMM’s makefile.
[] Finally, link your cxx files together to form a library. Put it somewhere in python’s site-packages directory, and off you go, you can import it from python and go use it.
I’ve heard about some sample cxx extension in the skel/ directory somewhere in the Panda3D source, you might find that useful as well.
Beware im not sure if i have implemented the a-star algorithm the correct way, but it look pretty fine.
You have to call astar2.py and it requires pygame for the output.
Unlike a “normal” a-star implementation it doesnt check the whole open-list in one turn, but has a maximum time it uses for checks (“step” function), until it returns and makes it possible to render a frame. next time the step function is called it’s continuing the calculation until it’s done, it’s finished once the result function doesnt return None…
This implementation is a bit more complicated then a usual a-star implementation, because of the non-blocking nature, and because it’s a 3d a-star version. If you run it you will see 3 images, the leftmost would be the lowest level, the middle the 2nd layer and the 3th the top layer.
im currently student. im new to panda3d and im implementing a simple RTS game for my final year project. im told to concentrate on the AI rather than on collision rays, selecting units etc.
however, i can see from your post that u already have your small RTS game which might help me a lot. can you please upload your project or link it to somewhere. i badly need help and im relying on you right now. i hope to hear from you soon
thanks in advance.
