Safe and Low-cost Robot Grasping through Impedance Control and Deep Learning

Grasping skill is significant in modern service robot that requires interaction with object under manipulation. Grasping of a new object is a trivial task for human operator while it is challenging for robot manipulators. Inspired by the grasping process of human operators, grasping control algorithms based on the integration of vision, tactile sensing. Deep learning, as an emerging technique successfully applied in many areas will be utilized to combining the advantages of both vision and tactile sensing. To guarantee the safe interaction between the robot manipulator and object, a robust interaction control scheme-impedance control will be applied. Force sensor is required in impedance control to detect the interaction force. To make the whole system affordable, the interaction force conventionally sensed by force sensor will be reconstructed by software observers.

Faculty Supervisor:

Haoxiang Lang

Student:

Yanjun Wang;Abdulrahman Al-shanoon

Partner:

Senturing Technologies Ltd

Discipline:

Engineering - mechanical

Sector:

University:

University of Ontario Institute of Technology

Program: