Oversampling-Based Control with Multi-Core and Edge Implementations

Digital control systems introduce unavoidable computational latencies. For some controllers this time delay inhibits practical use, even though they in theory could provide more efficient control. For example, solving an optimization problem each sampling period when using model predictive control. By sampling faster than the computation time and executing independent controllers on distributed hardware, e.g., a multi-core CPU or an edge/cloud server, output from the controllers can be interlaced. We call this oversampling-based control. This paper explores its implementation via simulations and experiments on a real Furuta pendulum using LQR and MPC control. We demonstrate both how control performance can be improved, without changing the control law, and practical issues that may occur.