Autonomous Vision-based Exploration and Mapping

This video shows the progress of a robot as it explores a real-world environment. It uses a stereo camera (and no other sensors) to detect obstacles and localize itself. It makes its own decisions about where to go, based on its expectation of the value of visiting any particular location. The environment is an indoor planar lab, about 20m on each side. More details at http://www.cs.ubc.ca/~simra

Category:

Howto & Style

Tags:

• robotics
• robot
• stereo
• vision
• slam
• mapping
• exploration

[simra73]

There is a particle filter running under the hood- from time to time an observation will cause the filter to jump from one trajectory hypothesis to another, and each trajectory has its own unique map. What you're seeing is the map associated with the new trajectory hypothesis.

[kinsel2011]

how easy is it to plot distance using stereo vision?

[dumbird2]

correct me if I'm wrong,

a trajerctory hypothesis (trajectory + map)corresponds to a sample

and the final trajecotry and map is the weighted sum of all the samples

but I don't know how a sample is created/smapled

[iwantcoolname]

something I do not get about this video, is why the map sometimes changes outside of camera cone?