Open question on autonomous navigation/mapping.

edit

Hello everyone,

I am trying to solve a problem where i want my robot to go around our office space without hitting anything and create a map of our office which i can use it in future. One way of create a map is to teleoperate the robot and create a map, which is straight forward. But this process becomes very tedious if your office is very large. I am trying to automate the process.

The main difficulty i am facing is i don't want my robot to go into any cubical's while mapping, i just want it to stay it in the aisles. If i use LaserScan data and write a simple algorithm for obstacle avoidance, the robot goes into the cubical's and hit gets stuck or hits small objects or disturbs the people who are working, i don't want that to happen.

I thought i can use point cloud data to make my robot stay in the aisles and not go into cubical's. At present, i am able to get pointclouds from kinect and use plane model segmentation algorithm to find planes.

But i have no idea what to do next or how to use these planes to make the robot move/navigate so that they just stay in aisles.

Any ideas how i can proceed the problem?

Thanks.