DRL-Assisted Resource Allocation for NOMA-MEC Offloading with Hybrid SIC

Area of Science:

• Wireless communication systems
• Mobile computing
• Network optimization

Background:

• Multi-access Edge Computing (MEC) and Non-Orthogonal Multiple Access (NOMA) are key technologies for enhancing 6G mobile systems.
• Existing NOMA systems often use fixed Successive Interference Cancellation (SIC) decoding orders, which can be suboptimal.
• Efficient user grouping and resource allocation are critical challenges in hybrid NOMA-MEC systems.

Purpose of the Study:

• To develop and evaluate a hybrid NOMA-MEC system for 6G.
• To propose a Deep Reinforcement Learning (DRL) algorithm for optimal user grouping.
• To minimize energy consumption in the NOMA-MEC system through resource allocation.

Main Methods:

• Implementation of a hybrid NOMA-MEC system combining NOMA for simultaneous task offloading and Orthogonal Multiple Access (OMA) for delay-tolerant users.
• Integration of a hybrid SIC scheme with dynamic decoding order adaptation.
• Development of a DRL-based algorithm for user grouping and a closed-form solution for resource allocation.

Main Results:

• The proposed DRL algorithm achieved a near-optimal user grouping policy.
• The system successfully minimized offloading energy consumption.
• Simulation results demonstrated fast convergence of the DRL algorithm.
• The hybrid NOMA-MEC scheme significantly outperformed the conventional OMA scheme.

Conclusions:

• The proposed hybrid NOMA-MEC system with DRL-based optimization offers a superior solution for 6G mobile systems.
• Dynamic SIC adaptation and intelligent resource allocation are crucial for efficient MEC and NOMA integration.
• This approach effectively balances computation capability, offloading efficiency, and energy consumption.