Computation Offloading in a Cognitive Vehicular Networks with Vehicular Cloud Computing and Remote Cloud Computing.

Area of Science:

• Computer Science
• Networking
• Artificial Intelligence

Background:

• Vehicular applications face increasing computational demands.
• Existing research primarily focuses on remote cloud computing (RCC) for offloading, with less attention to vehicular cloud computing (VCC).
• VCC-based offloading is challenging due to dynamic on-board resources and communication uncertainties in vehicular environments.

Purpose of the Study:

• To investigate and address the computation offloading problem in cognitive vehicular networks (CVN).
• To jointly consider vehicular cloud computing (VCC) and remote cloud computing (RCC) for computation offloading.
• To propose a novel perception-exploitation approach for leveraging VCC resources.

Main Methods:

• A Long Short-Term Memory (LSTM) model is used to predict on-board resource utilization.
• A decentralized multi-agent Deep Reinforcement Learning (DRL) algorithm is developed for collaborative offloading.
• The approach involves two stages: resource discovery and computation offloading.

Main Results:

• The proposed LSTM model effectively predicts on-board resource utilization.
• The DRL algorithm successfully manages collaborative computation offloading between VCC and RCC.
• Numerical simulations validate the effectiveness of the proposed algorithms across various scenarios.

Conclusions:

• The developed perception-exploitation strategy enhances computation offloading in CVNs.
• Joint VCC and RCC offloading, enabled by predictive resource discovery and DRL, is a viable solution for demanding vehicular applications.
• The study provides a robust framework for efficient resource management in dynamic vehicular networks.