Deep Learning-Based Task Offloading for Efficient and Reliable Computation in High-Mobility Vehicular Networks
Abstract
In vehicular networks, resource-constrained vehicles often face challenges in executing computationally intensive tasks due to limited local resources. Task offloading to nearby vehicles with sufficient resources provides an effective solution. However, in high-mobility scenarios, selecting the most appropriate vehicle for task offloading becomes a challenging and time-consuming process, leading to increased delays and degraded network performance. This paper proposes a novel deep learning-based task offloading technique to address this issue. The proposed approach operates in two stages. First, a deep learning model classifies nearby vehicles into eligible and unqualified nodes based on their ability to meet task requirements. Second, from the pool of eligible nodes, vehicles are ranked according to their credibility scores. Credibility scores are dynamically updated based on task completion within specified deadlines. By prioritizing vehicles with high credibility scores, the proposed technique ensures efficient and reliable task execution. Experimental results demonstrate that the proposed approach significantly improves the task execution success rate, reduces task offloading delays, and enhances the overall performance of vehicular networks.