Wean Hall Slam Experiment : 08.10.2004

Experiment of August 10th, 2004 : Wean Hall 6th floor

Conducted by Deryck Morales (deryck@cmu.edu) and Abhishek Sharma (abhi@cmu.edu)

Experiment Motivation:

This experiment confirms the implemented door detection methods that will replace the pink box scheme.

Experiment Setup:

The experiment took place along the 6400 corridor of Wean Hall, using cardboard to terminate the hallway near the central elevator area. The robot would run from this artificial end of the corridor to the intersection with the so called "narrow corridor" in the direction of Doherty Hall.
We first placed pink boxes on the floor against the centers of the doors along the hallway. The robot then mapped out the edge detecting the pink boxes.
The pink boxes were removed and the robot mapped the edge detecting the doors as landmarks. As we had hoped, all of the doors were successfully detected and no false positives occurred.

Experiment Results:

The image shown is the path of the robot by encoders (black) and Kalman estimate (blue) with the landmark estimates as crosses with thier covariance as ellipses.

Discussion:

The results of this experiment are promising because we expect to have more covariance for doors than pink boxes, and now we are tuning the Kalman parameters in simulation on the gathered data to reduce the covariance as much as possible. Although we know that the covariance for the door detection should not be as good as for pink boxes, we would like to approach the accuracy we had as much as possible.

Go back to: H-SLAM Experiments Page | H-SLAM page

Conducted by Deryck Morales (deryck@cmu.edu) and Abhishek Sharma (abhi@cmu.edu)

Experiment Motivation:
This experiment confirms the implemented door detection methods that will replace the pink box scheme.

Experiment Setup:
The experiment took place along the 6400 corridor of Wean Hall, using cardboard to terminate the hallway near the central elevator area. The robot would run from this artificial end of the corridor to the intersection with the so called "narrow corridor" in the direction of Doherty Hall. We first placed pink boxes on the floor against the centers of the doors along the hallway. The robot then mapped out the edge detecting the pink boxes. The pink boxes were removed and the robot mapped the edge detecting the doors as landmarks. As we had hoped, all of the doors were successfully detected and no false positives occurred.

Experiment Results:
	The image shown is the path of the robot by encoders (black) and Kalman estimate (blue) with the landmark estimates as crosses with thier covariance as ellipses.


Discussion:
The results of this experiment are promising because we expect to have more covariance for doors than pink boxes, and now we are tuning the Kalman parameters in simulation on the gathered data to reduce the covariance as much as possible. Although we know that the covariance for the door detection should not be as good as for pink boxes, we would like to approach the accuracy we had as much as possible.