A Robust Multiobjective Control Framework for Human–Machine Collaborative IoT Systems via Reinforcement Learning and Multimodal State Fusion

ABSTRACT Human–machine collaborative IoT systems are a core component of Industry 5.0 intelligent manufacturing, yet their control performance is often degraded by environmental uncertainty, heterogeneous device dynamics, and dynamic task workloads. Existing control strategies struggle to jointly achieve robustness, adaptability, and real‐time scalability in multidevice scenarios. This paper proposes a robust multiobjective control framework that integrates reinforcement learning‐based task scheduling, model predictive control for device actuation, and multimodal state fusion for reliable system perception. A composite optimization objective is formulated to minimize task completion time, energy consumption, and scheduling delay under practical industrial constraints. Multimodal sensor data are fused using Kalman filtering to provide noise‐resilient state estimation and stable closed‐loop control. Experiments on an Industry 5.0 IoT dataset show that the proposed method achieves a 12.4% reduction in task completion time and a 10.9% improvement in energy efficiency compared with state‐of‐the‐art baselines, while maintaining superior robustness under dynamic disturbances.