Safety Improvement and Control Strategies in Human-Robot Collaboration Using Non-Rectified Multicamera Images

forthcoming

Human-robot collaboration (HRC) is an important and established subject of study and research in the field of robotics (as part of Industry 4.0) which currently pervades many aspects of human endeavour. Advances in HRC have diversified the application of robots from the traditional production line to intelligent manufacturing, homecare, and healthcare. In such applications, human-robot collisions are bound to occur, with adverse consequences. To overcome the challenge of collision and improve safety in HRC in a shared workspace, several techniques (which employ vision, acoustics or haptics technology) have been developed. The techniques include external positioning of cameras in the workspace, use of robot skin embedded with sensors, and image-based feedback control. These three methods typically apply vision-based technology. This proposal aims to employ a novel and validated convergent stereo camera model (that does not require image rectification) in robots with a view to improving safety in HRC. The model is characterized by less computational complexity and short execution time relative to existing reconstruction models in stereovision literature. Subsequently, the project will introduce an improvement for robot skin to also have visual information from high-resolution cameras by extending the geometric analysis and mathematical modeling of non-rectification stereo imaging to multicamera imaging in the light of safety improvement in HRC using robot skin. This will complement the tactile sensation capabilities of robot skin with enhanced visual features that incorporate elements of the validated stereo camera model. Finally, a double-view image-based visual servoing control strategy will be developed for the improvement of safety in HRC using the validated model. While contributing to robotics literature and presenting new approaches to collision avoidance in HRC, this project will ultimately deepen the researcher’s understanding of robotic vision and control.