Optimal Control for Aerial Manipulation
Optimal control is a powerful tool to synthesize motions for complex robotic systems. This has become the state-of-the-art technique for planning motions and controlling legged robots. This technique, however, is not being widely applied within the field of aerial manipulation. Instead, this community has typically leaned to consider methods that require the manipulation of the analytical expressions. This is not always possible for a complex system and therefore, simplifications are required either limiting the movement of the robot or simplifying the geometry (sometimes both).
Inspired by the work done in the legged robotics community, my thesis aims to apply optimal control techniques to both plan motions and control aerial manipulators. This thesis is specially focused on aggressive maneuvers including those involving contacts in favor of the motion. Optimal control allows us to achieve that, but we have to solve some challenges in exchange. One of the biggest ones is being able to solve a fairly big optimization problem in a fairly short amount of time. This explains the interest growth towards Differential Dynamic Programming (DDP) solvers since they have a low computational cost. As always, this comes at the price of some drawbacks: poor globalization capability and the inability to consider constraints. These are both hot topic research within the robotics community and, in fact, so they are of my thesis.
During my talk, I will give an overview of the work done so far in my thesis regarding the application of optimal control techniques in aerial manipulation, with special attention to the addition of constraints into DDP-like solvers.