Task Offloading and Resource Allocation for ICVs in Vehicular Edge Computing Networks Based on Hybrid Hierarchical...

Area of Science:

• Intelligent Transportation Systems
• Edge Computing
• Artificial Intelligence

Background:

• Intelligent connected vehicles (ICVs) face onboard computational limitations.
• Vehicular edge computing networks (VECNs) enable task offloading to mobile edge computing (MEC) to reduce computational load.
• Dynamic transportation environments necessitate efficient task offloading and resource allocation.

Purpose of the Study:

• To develop an efficient task offloading and resource allocation strategy for vehicle-road collaborative edge computing networks.
• To minimize the combined time and energy costs associated with vehicular computations.
• To address the complexity of mixed discrete and continuous decision variables in dynamic VECNs.

Main Methods:

• Formulated the task offloading scheduling and resource allocation problem.
• Proposed a two-layer hybrid hierarchical deep reinforcement learning (HHDRL) algorithm.
• Upper layer: Enhanced Double Deep Q-Network (DDQN) with self-attention for discrete actions (communication decisions).
• Lower layer: Deep Deterministic Policy Gradient (DDPG) for continuous actions (power control, resource allocation).

Main Results:

• HHDRL demonstrated significant reduction in total computational cost compared to benchmark algorithms.
• The algorithm showed robustness and adaptability in dynamic and varying environmental conditions.
• The hybrid approach effectively decomposed complex action spaces for improved performance.

Conclusions:

• HHDRL provides an effective solution for task offloading and resource allocation in VECNs.
• The proposed algorithm enhances computational efficiency and reduces costs for ICVs.
• The study highlights the potential of AI-driven edge computing in intelligent transportation systems.