<p>Cloud computing implementations are economical, scalable, and manageable for industrial applications. The abundance of available services renders the cloud paradigm one of the most favored industries in contemporary society. It cannot accommodate latency-sensitive Industrial Internet of Things (IIoT) applications, including intelligent communications, Industry 5.0, robotics, oil and gas, and automotive sectors. Consequently, fog computing has arisen as a viable alternative for IIoT applications that are latency-sensitive. Consequently, fog computing can provide enhanced quality of service and diminished latency. Cloud computing is superior for applications that operate in real-time. Despite its theoretical foundations, the challenge of accurately placement industrial IoT services to fog nodes persists and has attracted considerable academic attention. This paper proposes a conceptual computing paradigm for optimizing service location in the IIoT, founded on middleware for cloud-based fog control. This issue is given as an independent scheduling model for handling service demands within defined constraints, considering the diverse schedules and resources. To optimize the use of fog resources and improve service quality, an autonomous evolutionary approach based on reinforcement learning techniques has been proposed to tackle the problem of IIoT service location. The advanced reinforcement learning approach utilized is the heterogeneous benefit operator-critic method, designed to optimize long-term cumulative rewards. Experimental research has been undertaken in a simulated artificial environment based on many metrics, including fog use, finished services, response time, and service latency. The results of the comparisons indicated that the proposed reinforcement learning-based framework surpasses the advanced methodologies documented in the literature.</p>