Embedded Model Control for Networked Control Systems

Classical control approaches avoid the use of unreliable networks such as the Internet, as it is a source of delays, uncertainities and lost packets. This can lead to difficulties or even total loss of control over the system of interest. However, as these networks are improved, the interest in networked control systems grow, as they allow to implement controllers in different machines depending on the required computational power, adding scalability to the system. For this reason, this project aims to implement a control system using the Embedded Model Control, a model-based strategy, to overcome the difficulties associated to networks. A complete control system is implemented and deployed using a remote server at Politecnico di Torino as the control unit and a simulated plant in Simulink running in a PC in Spain, both connected using MQTT messaging protocol. It also provides a web interface to supervise the control variables, and allows different plants to share a single control server. The Embedded Model Control, as well as a variation also implemented which recalculates the control parameters on each control iteration are successfully implemented, tested and their feasability validated. Results show that they succeed in keeping errors bounded and maintain stability even inf the sampling time rises (tests are made from a nominal value of 0.5 seconds up to 10.4 seconds). It also effectively rejects disturbances (including sensor biases and noise) and adapts to neglected dynamics non considered in the modelling phase, as tested in 100 executions with random variations in model's parameters, showing the robustness of the method.