<p>Model predictive control (MPC) has become vital for managing multivariable industrial control systems with constraints, providing optimal control performance. However, traditional single MPC implementations are often limited by their vulnerability to faults, computational failures, and lack of redundancy, especially in time-critical applications. This work presents a redundant MPC over Edge (RMPCoE) architecture that leverages edge computing to maintain uninterrupted control even when local hardware fails. Unlike traditional approaches that rely on On–Off switching, PID controllers, or centralized MPC alone, RMPCoE maintains synchronized controllers on both the local programmable logic controller (PLC) and a separate edge device. The simulation of a first-order thermal system shows that RMPCoE achieves 99.3% improvement in post-fault RMSE and 1.79% false alarm rate. Universal Error Function (UEF)-based detection optimized for thermal control applications, achieves approximately 570 ms detection latency with selected speed-priority weight configuration. RMPCoE demonstrates competitive performance while offering advanced predictive capabilities and constraint handling. The proposed architecture is beneficial for Industry 4.0 smart factories.</p>