Design of networked control system for stabilizing ball plate system using Smith predictor integrated with GRU-type neural network.

Time-varying network communication delays cause significant challenges, such as reduced control performance and destabilization of closed-loop systems in networked control system. Addressing these issues is critical for achieving stable and efficient operation. A three-dimensional ball-on-plate system was developed using a six-degree-of-freedom industrial manipulator in this study . This system faces two main sources of delay: the round-trip time of network packets and the latency caused by observing the ball's position with a camera. These delays are time-varying and difficult to predict, potentially impairing the system's ability to maintain stability. a control approach was proposed to mitigate these effects. The time-varying delays were predicted using a Gated Recurrent Unit (GRU), a type of recurrent neural network known for its capability to model sequential data. The predicted delays were incorporated into a Smith predictor-based controller, which is well-suited for systems with known delays. This integrated control strategy was implemented to stabilize the ball-on-plate system. Experimental results demonstrated the feasibility and effectiveness of the proposed method compared to feedback control without delay compensation.