A Vision-Guided Multi-Robot Cooperation Framework for Learning-by-Demonstration and Task Reproduction
What is the paper contribution/novelty?
-
Doing LfDs with humans performing the demonstrations with both hands.
-
Using robust tracking systems (e.g., a kalman filter to track the data) for better visual servoing.
It's still not clear to me, though ... however, the combination of LfD and visual servoing seems fairly common, I remember Sanjay's done that with the dVRK. They use a Kalman filter for this, but again that's very common so where is the novelty?
Maybe their main novelty is the "bimanual" task where we have two robot arms.
They also focus on using tools, so move the tool's tip by moving the end-effector's top, under the assumption that the tool is held rigidly. That's another important thing I can do.
Their main experimental/task contribution is a bimanual sewing task.
Why does their method work?
- Unfortunately I'm not totally sure ... it's a bit confusing but I'll re-read the paper.