Closed DanielePucci closed 6 years ago
Possible discussion points.
[ ] Is pairing with learning enough to address the existing barrier for online MPC in aerial or terrestrial locomotion? Is the barrier even there?
[ ] What are the prospects for transporting the more formal treatment of control we observed in aerial manipulation talks to loco-manipulation?
[ ] To what degree might whole-body control strategies benefit aerial manipulation? What modifications might be necessary to address underactuations present?
[ ] We saw that mixed-integer convex optimization could be used for provably optimal trajectory optimization for low-dimensional models. What are the practical limits of this approach scaling more broadly?
Are linear models of walking robots a good idea? That is, are we gaining enough from how easy it is to work with them to justify how much they might deviate from the true dynamics?
Do not hesitate to comment this issue for adding questions you would like to be discussed during the
Panel discussion
at 15.15