Truckbot project -- "visual memory" test

Up till now Truckbot has been a reactive robot. It used visual features extracted from the environment to "prompt" behaviour. E.g. "moving between parallel lines" or "orthogonal to crossing lines".

It now has a visual memory to record where it's been and -- in future -- be able to go back to find objects given a description of the object and/or its location.

This clip shows an old "mission" re-jiggered to dump out the visual memory at the end.

In this simple first test only the "panorama" after each major 90 turn is dumped.

The s/w pans back and forth across the panorama to show how the various images have been pasted together to form a total picture of the scene.

The pasting is very primitive at this stage, with no allowance for perspective or even camera rotations.

Category:

Science & Technology

Tags:

• computer vision
• robots
• autonomous robots
• visual memory