A computation offloading strategy for multi-access edge computing based on DQUIC protocol

Computation offloading can efficiently expand edge resources and is widely used to perform computing-intensive and delay-sensitive tasks. The inability of existing offloading strategies to pay attention to both packet loss problem and performance problems caused by channel noise usually lead to serious encoding costs and retransmission costs in offloading by traditional communication protocols. To address these issues, we propose a dynamic analog-digital coding QUIC (DQUIC) protocol to ensure the efficiency and reliability of edge computing data transmission. The DQUIC protocol uses a dynamic encoding method based on continuous slot communication state to handle sudden errors with a small encoding cost. Moreover, we design a dynamic multi-access edge computing (MEC) model using the DQUIC protocol for communication, which considers the impact of channel noise on communication rate and channel packet loss rate. In the dynamic MEC environment, the double deep Q-learning (DDQN) algorithm is used to solve the offloading decision problem and find the optimal offloading strategy. The experimental results demonstrate that our computation strategy, which leverages DQUIC, surpasses those strategies grounded in the DQUIC protocol and Coco protocol within a dynamic MEC environment.