Application of Event-based Cloud NMPC with Time Delay Compensation
Alvin Immanuel Surjana, Elmar Ahle, Dirk Söffker · 2024
The increasing availability of cloud computing services has revolutionized industrial productivity by providing essential tools for data storage, data transfer, and computation. In a cloud-based control system (CCS), the controller is located in the cloud, thereby enabling complex computation processes such as nonlinear model predictive control (NMPC). Using cloud computing, expensive control algorithms can be deployed without the upfront cost of a local computing unit. The computation time can be drastically reduced due to the improved computing power, which is important in the case of systems with fast dynamics in combination with NMPC-based control to maintain the existing time step size with regard to the calculation time. Placing the controller in the cloud also introduces communication latency in the control loop. It has been highlighted in former studies that computation time and communication latency can reduce the reliability and stability of a CCS. This paper proposes a cloud-based model predictive control strategy that extends simple time delay compensation with event-based control to reduce the effect of computation and communication delays. The optimal input variable sequence is used until a new sequence is available. This event-based method reduces the amount of communication and redundant calculations. The developed CCS algorithm is implemented in the real-time control of an experimental pendulum crane system. The effects of the proposed algorithm in mitigating computation and communication delay are compared to another CCS strategy that only uses the first element of the sequence.