Basic A Star pathfinding for a platformer game

Question

I'm currently writing a basic platformer game with the idea of learning the basis patterns of a 2D game development. One of them is the amazing A Star Pathfinding. After reading and study a few articles (two of them are the best: Link and http://theory.stanford.edu/~amitp/GameProgramming/Heuristics.html#S7), tutorials and source code, I did a simple A Star Pathfinding using the Manhattan method that works well for my personal learning process.

The next, and logical step, is to do the same on a platform scenario where the NPC must jump or climb ladders to reach the goal target. I've found these screencastings http://www.youtube.com/watch?v=W2xoJfT6sek and https://www.youtube.com/watch?v=Bc7tu-KwfoU that show perfectly what I want to do.

I haven't found for the moment any resource that explains the method to implement it. Can someone give me a starting point I can work with?

Thanks in advance

Answer 1

I haven't any clean A* code around fit for an example but let me give you a few pointers when implementing A*.

First off, the most essential tip that I often find neglected in tutorials is use a priority queue.

Whenever you generate possible move states, put them into the priority queue with priority being the value of your heuristic for that move state. This differentiates A* from DFS (which uses a stack and backtracking) or simple greedy path finding (which simply goes to the best state with respect to the present state).

Next, mind those pointers! I don't know what language will you implement this but even if you do it in some language that abstracts pointers (unlike C), you have to understand the concept of a deep copy versus a plain shallow copy. I remember the first time I did A* (in Python) and I did not pay attention to this. My states were repeatedly overwritten leading to unpredictable behavior. Turns out I did not make deep copies!

That said, depending on how complex the task you have in mind, all those deep copies might clutter up your memory. While good memory management will help, for big problem instances that need real-world performance, I tend to just use iterative-deepening A* so that after every iteration, I get to unclutter my memory.

Answer 2

As far as I can tell, the only difference that jumping/ladders make is that they affect possible movement. So, all your algorithm has to do is take into account where a character can and can't move to, and the costs (if you're implementing that) of moving there.