<![CDATA[This review examines decision intelligence systems for autonomous production planning and control in Industry 4.0 manufacturing. It addresses the limits of conventional production planning and control systems, which depend on deterministic models, historical data, human expertise, and a two-tier structure of schedule generation and execution control. Such systems perform poorly when demand variability, resource constraints, machine workload changes, material shortages, and quality disruptions make released schedules suboptimal. The review synthesizes evidence from multi-agent systems, holonic manufacturing, cyber-physical production systems, digital twins, machine learning, reinforcement learning, data-driven scheduling, and prescriptive analytics. It shows how decision intelligence links data acquisition, analytics, optimization, and execution through a closed-loop sense-decide-act structure. The framework rests on four core components: data acquisition and management, analytical models, decision algorithms, and execution mechanisms. It evaluates five operational domains: dynamic scheduling, inventory and supply chain coordination, resource allocation, predictive maintenance, and resilient production control. Numerical evidence strengthens the review. Digital twin literature shows 38% use simulation models to represent physical systems and predict future states, while 29% integrate optimization into simulation models. A validated multi-agent logistics case reported savings of 2.6 million pallet-days per year through better use of terminal free time. The review also identifies six unresolved barriers: data quality and integration, model interpretability, scalability, real-time implementation, human-system interaction, and lack of standard frameworks. Future systems should combine explainable AI, edge-cloud computing, digital twin-based optimization, hybrid control, and standardized evaluation protocols to achieve adaptive, scalable, and trusted autonomous production control.]]>
