Abstract: In 5G communication networks with sparse beam-domain pilots, the asymmetry and time-varying characteristics of communication links have led traditional approaches to rely on centralized computing architectures, where data must be transmitted to a central server for processing to obtain delay information and generate control instructions. This approach not only causes frequent packet loss during long-distance transmission but also significantly reduces data transmission rates and increases latency, thereby hindering the timely acquisition and processing of 5G channel delays. To address these issues, this paper proposes an adaptive control method for low-latency transmission in 5G communication channels based on edge-cloud computing. The proposed method shifts the computing process to edge servers close to the data source, leveraging the rapid response capability of edge computing to shorten the data transmission distance and reduce transmission latency. Channel delay prediction is performed using a convolutional neural network, which mitigates the problems of untimely acquisition and inaccurate processing of delay information caused by long-distance data transmission in traditional centralized architectures. Furthermore, by utilizing the powerful computing, storage, and network resources of the cloud computing center, the delay difference between the current time and the predicted delay is calculated and fed into a PID control system. The controller of the 5G communication channel is adjusted through three independent control actions—proportional, integral, and derivative—and the resulting control signal is input into a delay compensator to achieve adaptive control. Experimental results demonstrate that the proposed method performs well in 5G channel delay prediction and adaptive low-latency transmission control, significantly reducing the average delay and packet loss rate while improving network performance and stability.