What type of study is this?

September 12, 2025Open Access

A Meta Multi-Objective Reinforcement Learning Framework For Non-Orthogonal, Age-Optimized Information Dissemination in Vehicular ‎Networks

Key Points

The proposed method reduces delay in information updates and minimizes power needed for transmissions.
Simulation results show significant improvements in efficiency compared to older approaches, enhancing data sharing.
Reinforcement learning models, including deep Q-networks and DDPG, are utilized to tackle complex multi-objective problems.
Meta-learning enables quick adaptation of solutions without retraining, saving time and computational resources.

Abstract

This paper studies how to send fresh and efficient information in vehicular networks. A roadside unit ‎‎(RSU) updates vehicles about different events. The goal is to reduce the delay in receiving information ‎while also saving power during transmission. The system uses an innovative way of transmitting data. ‎It sends multiple messages at the same time using superposition. Vehicles cancel unwanted signals with ‎a technique called successive interference cancellation (SIC). This method helps to improve the efficiency of communication. The problem is complex because two objectives must be optimized. One is to ‎minimize the delay in information updates, and the other is to minimize the power needed to send ‎updates. This is a multi-objective problem that is difficult to solve using traditional methods. To ‎address this challenge, the paper uses reinforcement learning (RL). A deep Q-network (DQN) decides the ‎best way to decode messages, while a deep deterministic policy gradient (DDPG) model determines ‎the optimal power allocation. Each learning model trains separately for different cases, which increases ‎computational time and effort. Instead of training models separately, the paper proposes a meta-‎learning approach. This helps estimate good solutions quickly without the need for retraining every ‎time. The meta-model adapts with small updates, saving significant time and computational resources. ‎Simulation results demonstrate that the proposed method outperforms older approaches. It reduces ‎training time while still achieving high efficiency. Moreover, it provides a better balance between ‎formation freshness and power consumption. These improvements make it highly suitable for real-time ‎data sharing in vehicular networks. This research has practical implications for enhancing road safety ‎and smart transportation systems. By optimizing data dissemination it contributes to the development ‎of more reliable and efficient vehicular communication networks‎.

Read Full Paperexternally

Ask AI

Helpful

Bookmark

View Full Paper