Rebecca-E
ADAPTABLE COMPUTER VISION SYSTEM FOR AN ABSTRACT ROBOT
Project Site:
http://robotics.icstweb.org/rebecca-e/
Project Description:
http://robotics.icstweb.org/rebecca-e/about/
Usually robot control systems use a set of algorithms matched to specific task types. For example: tracking, pursuit and interception of moving objects, line following, recognition of 3D-pointing gestures for human-robot-interaction… The Robot uses concrete scheme which could be described by Finite State Machine:
For example the Microsoft Robotics Studio uses concrete VPL schemes to program robots:
The profits of such robot organization are usable in situations like car assembly factory, nuclear power station control, an airplane piloting and etc. On the other hand, Finite State Machine usage eliminates creative abilities for the robot’s AI, and by the way it eliminates chance to create more «wise» machine.
It is of interest to create an adaptable control system, able to change its structure, depending on different conditions. As a sub-task of creating such system I consider a task of creating dynamically adaptable computer vision system. In my current project I’m working on ontology-based approach. I propose to use ontology as a knowledge bank for keeping information about image processing algorithm’s.
The Aim Project focuses on developing flexible computer vision system for an abstract robot. As a base for testing the system it is proposed to use simple robot with one camera, able to perform simple operations («←», «↑», «→», «↓», «0»).
