Data prioritization aware resource allocation in internet of vehicles using multi-agent deep reinforcement learning.

Area of Science:

• Computer Science
• Electrical Engineering
• Transportation Engineering

Background:

• Intelligent Transportation Systems (ITS) and the Internet of Vehicles (IoV) face spectral resource limitations and real-time communication demands.
• Effective resource allocation in IoV is challenging, especially considering data priorities and dynamic vehicle environments.
• Optimizing transmission power and spectrum allocation is crucial for maximizing IoV performance.

Purpose of the Study:

• To design a novel time-series-based multi-agent deep reinforcement learning framework (NL-MAPPO) for resource allocation in IoV.
• To address the challenges of dynamic vehicle characteristics and diverse data priorities in IoV communication.
• To minimize transmission delays and energy consumption while maximizing vehicle-to-vehicle (V2V) link capacity.

Main Methods:

• Formulated the resource allocation problem as a multi-agent Markov decision process.
• Developed a multi-agent resource allocation algorithm utilizing a shared-critic mechanism for global channel information sharing.
• Introduced a time series-based channel information extraction mechanism to capture temporal dynamics.

Main Results:

• The proposed NL-MAPPO framework effectively optimizes spectrum allocation and transmission power.
• Demonstrated significant improvements in minimizing transmission delays and energy consumption.
• Achieved maximization of total vehicle-to-vehicle (V2V) link capacity.

Conclusions:

• The NL-MAPPO framework offers a superior solution for resource allocation in IoV compared to existing methods.
• The approach effectively balances performance metrics like delay, energy consumption, and link capacity.
• The time-series-based multi-agent deep reinforcement learning is a promising direction for future IoV research.