On May 10, we organized a scientific seminar on “Revisiting Reachability-Driven Explicit MPC for Embedded Control“. Juraj Holaza, a lecturer at Slovak University of Technology in Bratislava, led the seminar, which was organized in the framework of the FrontSeat project as part of the seminar series on “Research Seminar on Smart Cybernetics.”
Abstract:
Real-time implementation of explicit Model Predictive Control (MPC) is hindered by the evaluation of potentially large lookup tables. This presentation introduces an efficient method for explicit MPC, driven by reachability sets, which addresses this issue by dividing the lookup table into “relevant” subsets. This significantly reduces real-time complexity through online pruning and traversal of the sorted lookup table associated with optimal control law evaluation. Technically, we accelerate the point location problem by decreasing the number of explored critical regions by feeding the sequential search algorithm only by those regions that are “reachable” from the previous region. As a consequence, computational effort (i.e., the energy required) during the online evaluation of explicit MPC policies is minimized. It’s important to note that while the optimality of the control law is preserved, the cost for the accelerated point location problem lies in additional offline computation effort and a minor increase in memory requirements for the controller. The benefits of this method will be demonstrated through an illustrative case study.
This project has received funding from the European Union’s Horizon under grant no. 101079342 (Fostering Opportunities Towards Slovak Excellence in Advanced Control for Smart Industries).
0 Comments