In traditional ‘blind’ robot systems, in order to operate a robot arm to pick up an object, fairly precise position data for the object is required to operate correctly. However, in many cases, such precision is not available. By attaching two cameras to a robot through a computer, the robot is given the ability to ‘see’ its environment. This can allow the robot to adapt to changes in its environment and be able to independently determine the position of objects to pick up within its field of view. The aim of this project is to build such a robot system with stereoscopic vision.

The system will consist of two webcams connected to a computer which will use MATLAB to convert the visual data into positions for the robot. The position data is used by LabVIEW to output digital signals to a driver unit which converts the signals into voltages to move the Mitsubishi RV-M1 robot arm. MATLAB will use a stereoscopic vision algorithm and various image processing algorithms such as edge and corner detection in order to implement the feedback control of the robot arm.

The technology can be used in industry for applications which require frequent recalibration. The system will be more expensive than a traditional ‘blind’ system but will only involve the addition of a computer and two cameras. The fixed cost, however, is relatively small if frequent recalibration is required for ‘blind’ systems.

David Lee, David Zhang, Associate Professor Dragan Nesic, John Gao, Julian Wan