Optimized multi-tier task offloading strategy for sustainable IoV systems in 6G networks.

Area of Science:

• * Vehicular communication and networking
• * Edge computing and mobile cloud computing
• * Artificial intelligence and machine learning for network optimization

Background:

• * Sixth-generation (6G) networks enable advanced Internet of Vehicles (IoV) applications with strict latency and reliability requirements.
• * Efficient task offloading in IoV is challenging due to vehicle mobility, dynamic channels, resource constraints, and energy demands.
• * Existing solutions struggle to balance latency, energy, and task success rates in complex IoV environments.

Purpose of the Study:

• * To propose an Optimized Multi-Tier Task Offloading Strategy (OMTOS) for sustainable IoV systems.
• * To formulate a generalized latency-energy optimization problem for task allocation across a four-tier architecture.
• * To develop an adaptive and efficient task offloading solution for 6G-enabled IoV.

Main Methods:

• * Implementation of a four-tier computing architecture: vehicles, roadside units (RSUs), mobile edge computing (MEC) servers, and cloud.
• * Formulation of a latency-energy optimization problem considering task deadlines, resource capacity, and communication/computation delays.
• * Application of a centralized training with decentralized execution (CTDE) based multi-agent Soft Actor-Critic (SAC) method for dynamic offloading decisions.

Main Results:

• * OMTOS significantly reduces average delay and energy consumption compared to baseline methods (LE, EO, RO, GO, DQN, DDPG, SAC, MADDPG, MAPPO).
• * The strategy achieves a high task success rate and demonstrates superior convergence.
• * Sensitivity analysis confirms OMTOS's adaptability to diverse IoV service requirements, balancing delay-critical and energy-conscious needs.

Conclusions:

• * OMTOS provides an effective and sustainable solution for task offloading in 6G IoV environments.
• * The multi-agent RL approach enables intelligent, decentralized decision-making for complex vehicular networks.
• * The framework's adaptability makes it suitable for a wide range of current and future IoV applications.