Question

I started to refresh my ai knowledge so i implemented some pathfind algorithms to solve 8-Puzzle.

• A*
• IDA*
• BFS
• IDFS

I was wondering why my implementation of IDA* has a longer path. It should be optimal like A*.

% python puzzle8.py -a idastar -d hard
IDASTAR - RESULT in 161.6099:
1 | 2 | 3
4 | 5 | 6
7 | 8 | N

cost: 0 total_cost: 121
...
nodes 28

% python puzzle8.py -a astar -d hard
Max nodes 665 loops 1085
ASTAR - RESULT in 0.3148:
1 | 2 | 3
4 | 5 | 6
7 | 8 | N

cost: 0 total_cost: 115
...
nodes 24

Code is on gist https://gist.github.com/1629405

Update:

Code is now pointing to a working version.

% python puzzle8.py -a idastar -d hard
IDASTAR - RESULT in 234.4490: 

1 | 2 | 3
4 | 5 | 6
7 | 8 | N
...
nodes 24

But i’m still wondering why IDA* takes so much longer under python than A*.

Update 2:

Code is changed prints now visited nodes.

IDASTAR creates 4184368 ASTAR 1748 nodes.

Answer 1

Because your implementation of IDASTAR increments the limit by 10 with each iteration, which only guarantees that your solution will be no more than 9 more than optimal. Change the increment to 1, and you should get an optimal result (but take longer to do so).