Research output: Contribution to Journal/Magazine › Journal article › peer-review
<mark>Journal publication date</mark> | 1/03/2024 |
---|---|
<mark>Journal</mark> | IEEE Transactions on Intelligent Transportation Systems |
Issue number | 3 |
Volume | 25 |
Number of pages | 15 |
Pages (from-to) | 2733 - 2747 |
Publication Status | Published |
Early online date | 31/01/23 |
<mark>Original language</mark> | English |
Using the powerful communication capability of 6G, various in-vehicle services in the Internet of Vehicles (IoV) can be offered with low delay, which provide users with a high-quality driving experience. Edge computing in 6G-enabled IoV utilizes edge servers distributed at the edge of the road, enabling rapid responses to delay-sensitive tasks. However, how to execute computation offloading effectively in 6G-enabled IoV remains a challenge. In this paper, a Computation Offloading method with Demand prediction and Reinforcement learning, named CODR, is proposed. First, a prediction method based on Spatial-Temporal Graph Neural Network (STGNN) is proposed. According to the predicted demand, a caching decision method based on the simplex algorithm is designed. Then, a computation offloading method based on twin delayed deterministic policy gradient (TD3) is proposed to obtain the optimal offloading scheme. Finally, the effectiveness and superiority of CODR in reducing delay are demonstrated through a large number of simulation experiments.