Hi Eitan,
In response to your question of whether I was publishing /initialpose, I am.
It looks like this:
header:
seq: 49665
stamp:
secs: 1291167633
nsecs: 17403000
frame_id: /map
pose:
pose:
position:
x: 0.0
y: 0.0
z: 0.0
orientation:
x: 0.0
y: 0.0
z: 0.0
w: 1.0
covariance: [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
Just publishing an initialpose message isn't enough for amcl to figure out where the robot is. You need to try and give it a reasonable close guess as to the robot's actual position to start with. Technically, there's a way to configure amcl to use a lot of particles that would allow amcl to solve this "kidnapped robot" problem for you, but its computationally intensive, takes a fair amount of time, and its doubtful that you have amcl configured to do this. The message that you're publishing will just place the center of the robot in the lower left corner of the map, meaning much of the robot will actually be off the map.
Are you sure that laser scans, map, and odometry information are all going into amcl correctly. You'll need all three for it to do anything.