Embedded Model Predictive Vibration Control On Low-End 8-Bit Microcontrollers Via Automatic Code Generation

Model predictive control is a promising approach for vibration attenuation, as it allows the direct consideration of constraints and the use of system models for predicting future behavior. However, the straight application of model predictive control is in general challenging, as the efficient implementation requires expert insight, especially for fast systems on low-end embedded platforms. This paper investigates the practical feasibility and deployment of model predictive control for the active vibration attenuation of a cantilever beam using a low-end, low-cost 8-bit microcontroller. The control aim is to suppress vibrations while respecting the input constraints due to the depolarization limits of the piezoceramic actuators. The embedded model predictive control code is automatically generated using the AO-MPC code generation tool. The resulting implementation is real-time feasible even on the utilized low-cost, low-end Arduino UNO electronics prototyping platform, which is based on the widely used Atmel ATmega328, a 8-bit AVR architecture reduced instruction set processor. The memory demand, computation time and performance are examined under various settings. As shown, real-time execution of the AO-MPC generated code for horizon lengths of up to 7 steps for a second order model and a 20 ms sample time are computationally feasible even on the low cost hardware used here.