Potential error in the lecture of 'Chapter 3: How to localize the robot in the environment.'

[bluebird]

‘The weight value (wt) is defined as the absolute difference between the actual pose of the robot and the predicted pose by that specific particle.’

It is unlike to be true because the actual pose is unknown in a localization problem. As far as I know about the particle filter, the weight is calculated from the difference between the real measurement and the measurement that a particle supposes to get. But, I know nothing about the advanced AMCL. I cannot be sure.

Another thing I am worried about is the robot’s real pose. In reality, the real pose is a mystery, and we will never know. However, we are performing the simulation. Is it possible that we can receive the simulated real robot pose from Gazebo? We can compare this pose with the output of the AMCL. Then, the performance of the AMCL can be estimated.

[rzegers]

Hello @bluebird,

what it means is that the laser measurement that is taken from the actual/real robot pose and then compared with the fictive/predicted laser measurements of the particles. This doesn’t imply that the robot pose is known. So your understanding is correct.

@bluebird

Is it possible that we can receive the simulated real robot pose from Gazebo?

Yes, it is posible to get the real robot data directly from Gazebo. You could for instance get the pose of a model from a Gazebo plugin.

Cheers,

Roberto