Deep Reinforcement Learning-Enabled Computation Offloading: A Novel Framework to Energy Optimization and...

Area of Science:

• Intelligent Transportation Systems
• Computer Science
• Network Engineering

Background:

• Vehicular Edge-Cloud Computing (VECC) faces challenges with network congestion and service delays due to uneven workload distribution.
• Ensuring data security and optimizing energy consumption are critical issues in current VECC systems.
• Conventional offloading mechanisms struggle to manage heterogeneous network parameters effectively.

Purpose of the Study:

• To develop a deep reinforcement learning-enabled computation offloading framework for multi-tier VECC networks.
• To address network congestion, enhance data security, and optimize energy usage in intelligent transportation systems.
• To improve the efficiency and scalability of vehicular edge computing.

Main Methods:

• A dynamic load-balancing algorithm analyzing RSU load, channel capacity, and latency.
• Deployment of Unmanned Aerial Vehicles (UAVs) to augment resources in high-density zones.
• Implementation of Advanced Encryption Standard (AES) with dynamic key generation for data security.
• A context-aware edge caching strategy to reduce redundant computations and energy overhead.
• Formulation of a mixed-integer optimization model and a deep learning-based algorithm for close-optimal offloading solutions.

Main Results:

• The proposed framework significantly reduces energy consumption compared to existing methods.
• Empirical evaluations demonstrate superior performance in terms of energy savings and efficiency.
• The system effectively balances workloads across Roadside Units (RSUs) and dynamically augments resources.
• Enhanced data security through AES encryption with dynamic key generation is achieved.

Conclusions:

• The deep reinforcement learning framework offers a sustainable, secure, and scalable solution for intelligent transportation systems.
• The dynamic load balancing and UAV augmentation effectively mitigate network congestion and improve resource utilization.
• The study highlights the potential of AI-driven approaches to overcome VECC limitations.