Collaborative robots, or cobots, are entering the manufacturing scene as the next evolution in automated aids. One of the most notable difference between the classical industrial vs. collaborative robots is the lack of physical barriers separating the human from the robot, which implies that the robot’s environment is now dynamic, requiring active sensing and acceptable physical interactions. The work done at BARC attempts to address what the industry deems acceptable for these new types of collaborative robots.
Projects
- Physical Human-Robot Interaction
- Programming by Demonstration (PbD)
- Shared Control
Physical Human-Robot Interaction
Respective Student(s): Tony Piaskowy
Robots have traditionally been used for swift, highly accurate, and repeatable movements, however this inherently requires lots of preprogrammed information, a stiff structure, and powerful motors. This situation/ has restricted human-robot proximity for obvious safety reasons. However, if there is a way to bring humans and robots together, integration of a new task or dealing with unexpected variations can become much easier with a human element present. The process could merge the flexibility, wealth of sensor information, and trained expertise of the operator along with the strength and accuracy of a robot. Research at the BARC is just starting to look into this problem and currently focuses on efforts including: impedance control, torque/force sensing, and disturbance estimation.
Impedance control can allow this type of experience to occur.
(http://i585.photobucket.com/albums/ss297/stcasey34/kid-lifting-car.gif)
Programming by Demonstration (PbD)
Respective Student(s): Tony Piaskowy
Robot movement is not a trivial task in automation. Traditional approaches require programming explicitly the 6 degrees of freedom (DOF), 3 for position and 3 for orientation, for each waypoint along a robot’s path. This strategy can be awkward and unintuitive to a non-expert robot programmer. Programming by Demonstration (PbD) attempts to remedy this discomfort by using intuitive user performed task trajectories as implicit 6-DOF inputs. The robot can then ‘replay’ these trajectories to mimic its human counterpart, avoiding the tedious traditional definition of robot programming.
Programming by Demonstration (PbD) allows human operators to program intuitive movements without using a keyboard.
Example of a Gaussian Process Regression for taking noisy data and processing the underlying shape.
Shared Control
Respective Student(s): Parker Owan
Augmenting human performance with the machine precision has enabled modern engineering. From the industrial revolution to space age computing, machines have aveliated the burden of human error and fatigue limits. Contemporary manufacturing has utlized robotics to meet the requirements of highly repetitive and highly accurate tasks, however often these tasks are fixed preprogrammed routines that are unable to handle large process variation. The work done in the BARC has focused on sharing tool control between the robot and human counterpart, relying on the robot for well informed movements, but enable the robot to request human intervention in areas of uncertainty.
Example of a robot path showing periods of uncertainty from lack of enviroment/task information (red), where the machine control would stop and the human operator control would be requested.
Related Publications
P. Owan, J. Garbini, and S. Devasia, International Conference on Advanced Intelligent Mechatronics (AIM), Munich, Germany, Addressing Agent Disagreement in Mixed-Initiative Traded Control for Confined-Space Manufacturing, July 3-17, 2017.
https://ieeexplore.ieee.org/document/8014022
P. Owan, J. Garbini, and S. Devasia, IEEE/RAS International Conference on Intelligent Robots and Systems (IROS 2018), Madrid, Spain Managing Off-Nominal Events in Shared Teleoperation with Learned Task Compliance, Oct. 1-5, 2018.
https://ieeexplore.ieee.org/document/8594195
P. Owan, J. Garbini, and S. Devasia, IEEE Robotics and Automation Letters Faster Confined Space Manufacturing Teleoperation through Dynamic Autonomy with Task Dynamics Imitation Learning, Vol. 4(4), pp. 3844-3851, October 2019, and presented at the IIEEE International Conference on Robotics and Automation ICRA 2020.
https://doi.org/10.1109/LRA.2020.2970653