The RCL Lab has been involved in applying advanced control methods, mainly related to numerical optimal control, to different advanced robotic structures within various grants. Solving a numerical optimization problem was traditionally a computationally expensive task: as a consequence, numerical optimal control methods such as model predictive control were typically used to control slow processes, such as chemical plants, with sampling intervals in the order of seconds or minutes; on the other hand, controlling robotic systems typically requires much shorter sampling intervals, and thus could only be done using simpler control strategies. Due to recent advances in efficient solvers, and to the availability of more powerful microprocessors, the described situation is changing. The use of numerical optimal control (especially for the simpler case of linear systems) is now been extended from the classical process control applications to applications requiring faster sampling rates in areas such as mechatronics, automotive, and power electronics.
Following this trend, in our lab we have worked on control methods based on convex optimization for spherical parallel manipulators, optimization-based motion planning (energy and time optimal) and model predictive control of variable impedance actuated robots, and neural network-based approximations of optimal control strategies (including fault-tolerant control) of mechatronic systems. This research is in collaboration with the NU ARMS Lab and the NU Alaris Lab.