Retracing steps during search.

[Angie Hawk]

Hi guys,

I'm having some problems with my AI homework. We're studying A* pathfinding and I'm trying to implement a greedy best-first search algorithm before building on it to create the A* algorithm. My algorithm should return a graph from A to B once the algorithm has done its thing.

As a parameter, I supply the search function with a grid (two-dimensional array) to find a path through. Each time the algorithm decides the next step to take, it creates a new node and adds this node to an ArrayList (of type Node). My problems start when my program finds that it's walked down a dead end due to being greedy and needs to retrace steps to try the next quickest route from where it went wrong. How would I implement this behavior? (Recording how the current position was reached and backpedalling until the point where it went bad).

Thanks in advance for your help!

[Ulrika Tingle]

Angie Hawk wrote:My problems start when my program finds that it's walked down a dead end due to being greedy and needs to retrace steps to try the next quickest route from where it went wrong.

It's hard to be specific but you need to save state in some way as you advance. One way is to push state on a stack so you can pop it back when you retreat. The simplest way usually is to use recursion. Then the state will be preserved when you make a new recursive call and it will be restored again when you return from the call.

[Paul Clapham]

Also, the usual word in the algorithms field is "backtracking" for that process, not "backpedalling".