CEA-List is coordinating an EU project called MERGING (Manipulation Enhancement through Robotic Guidance and Intelligent Novel Grippers) to develop tools to make universally dextrous systems for robots that can be quickly and easily adapted to a wide variety of tasks, including handling large, flexible, or fragile objects.
Soft materials do not behave predictably in the “hands” of robots and are especially prone to damage during robotic handling. Two of the use cases the project is focusing on involve textile and food packaging materials. These items have to be grasped and handled gently—something that today’s robots are not very good at. The development of a universally dextrous system will depend on fine, adaptive control of both force and movement, beyond what is possible with today’s robotics technologies..
CEA-List’s solution leverages two enabling technologies—skills-based programming and motion-capture programming-by-demonstration—to make the kind of complex tasks mentioned above fast and intuitive to program and deploy.
Custom software and an associated catalog of skills (high-level actions that are easy for operators to understand and that can be deployed to complete a variety of tasks) were developed for the use cases addressed by the MERGING project.
The software’s user-friendly interface allows operators to choose skills like “grasp the bag,” “insert the bag in the track,” and “pull on the bag to check that it is inserted correctly.” The skills can be combined and adapted to different kinds of robots. The more high-level skills the robot already masters, the less the operator needs to program—a huge time saver.
The other solution developed for the MERGING project is a motion-capture-based programming by demonstration (PbD) system where the operator uses an off-the-shelf motion capture system and remote control to “show” the robot the desired trajectories and end points. Robots can be taught new skills directly by the operator offline or, using the remote control, online.
The MERGING project deliverables will be tested by partner companies Thimonnier (packaging solutions for liquids), Selmark (textiles), and VDL Fibertech Industries (fiber-reinforced composite parts). Meanwhile, additional headway on the software and skills library is being made through a CEA moonshot R&D program on self-learning robots. A more advanced capability—programming not just individual skills, but sequences of skills by demonstration—is being developed for TraceBot, another EU project.
For human operators, moving heavy or bulky objects is not only hard, it is also a potential source of accidents and injuries. That’s why collaborative robots—or cobots—are garnering increasing interest across virtually all industries. But getting robots to cooperate with human operators requires special control laws. CEA-List developed two such laws so that operators can easily program robots to help move and position large, fragile parts.
The first new control law improves collaboration between the human operator and robot for moving large parts. Low-level instructions are sent to each of the robot’s motors to ensure that the robot always adapts its movements to those of the human operator. A force control feature lets the robot detect any increase in resistance—which indicates that the human operator is changing direction—and adapt parameters like gripping force, speed, and trajectory to effectively work “with” the human operator.
The second new control law utilizes a hybrid force/speed controller that allows the operator to remotely control the cobot to position an object with a high degree of precision. When the operator is holding the part, the force exerted is automatically detected and used to control the robot. When the robot is holding the part, the operator controls the speed and position of the part with a hand-held remote.
Together, these two control laws deliver beyond-state-of-the-art performance. The integrity of the part being handled is protected through controlled torsional and other forces, for example, and deployment of new tasks is faster and requires no specific programming skills. The control laws came through validation testing in the lab on a VDL Fibertech Industries use case (handling large, fragile parts) with flying colors.
This work will soon be transferred to integrators who will apply it to various industrial cases. To be continued…