In smart factories, Multi-access Edge Computing (MEC) enhances Wireless Networked Control Systems (WNCS), strengthening wireless collaboration between edge devices and servers. However, limited communication resources pose challenges to the real-time performance and control accuracy of edge devices, while underutilized local computing resources leads to waste and increased communication overhead. In this paper, we first construct an Integrated Communication, Computing and Control (ICCC) system model to accurately characterize the closed-loop control process of Automated Guided Vehicles (AGVs) in material handling applications. Then, a form of nonlinear Proportional-Integral (PI) controller is deployed on the AGV, while Model Predictive Control (MPC) is employed on the MEC server to optimize control performance. Furthermore, given constrained communication resources, this paper models the joint optimization of task offloading and multi-channel access as a Decentralized Partially Observable Markov Decision Process (Dec-POMDP). To solve this problem, we propose a reinforcement learning-based Dynamic Decision-making Closed-loop Control (DDCC) algorithm. Leveraging local AGV observations, the algorithm dynamically optimizes task offloading, synergizing AGVs and edge computing resources to enhance control precision. Simulation results demonstrate that the proposed algorithm outperforms baseline methods in terms of the cumulative reward and the trajectory tracking error of AGVs.