Learn to Compress CSI and Allocate Resources in Vehicular Networks

Abstract

Resource allocation has a direct and profound impact on the performance of vehicle-to-everything (V2X) networks. In this paper, we develop a hybrid architecture consisting of centralized decision making and distributed resource sharing (the C-Decision scheme) to maximize the long-term sum rate of all vehicles. To reduce the network signaling overhead, each vehicle uses a deep neural network to compress its observed information that is thereafter fed back to the centralized decision making unit. The centralized decision unit employs a deep Q-network to allocate resources and then sends the decision results to all vehicles. We further adopt a quantization layer for each vehicle that learns to quantize the continuous feedback. In addition, we devise a mechanism to balance the transmission of vehicle-to-vehicle (V2V) links and vehicle-to-infrastructure (V2I) links. To further facilitate distributed spectrum sharing, we also propose a distributed decision making and spectrum sharing architecture (the D-Decision scheme) for each V2V link. Through extensive simulation results, we demonstrate that the proposed C-Decision and D-Decision schemes can both achieve near-optimal performance and are robust to feedback interval variations, input noise, and feedback noise. Index TermsVehicular networks, deep reinforcement learning, spectrum sharing, binary feedback.

Deep-Learning-Based Wireless Resource Allocation With Application to Vehicular Networks

“…In contrast, the random baseline does not enjoy such intelligence and the more vulnerable Link 3 fails the task finally. We carry the multi-agent RL idea further in [90], where we restructure our network architecture substantially to enable a centralized decision making yet with very low signaling overhead. In particular, each V2V transmitter constructs a DNN that learns to compress its local observation (measurement), which is then fed back to the central base station to serve as the input of the stored DQN.…”

Network Slicing with MEC and Deep Reinforcement Learning for the Internet of Vehicles

“…The number of source vehicles 𝑚 and destination vehicles 𝑛 is randomly chosen. The important simulation parameters are given as follows [22,23]. The carrier frequency is 2 GHz, the per-RB bandwidth is 1 MHz, the vehicle antenna height is 1.5 m, the vehicle antenna gain is 3 dBi, the vehicle receiver noise figure is 9 dB, the shadowing distribution is log-normal, the fast fading is Rayleigh, the pathloss model is LOS in WINNER + B1, the shadowing standard deviation is 3 dB, and the noise power 𝑁 0 is −114 dBm.…”

Resource Allocation Strategy Based On Tripartite Graph in Vehicular Social Networks

“…Based on the channel state information, such as interference and physical distance, the effect function is developed to game the content distribution vehicles with different connection times, thereby minimizing the average network delay. However, most of these algorithms are based on channel state information, and using methods such as deep learning and game theory [17][18][19], they cannot effectively solve the problems of increasing service interruption probability and low efficiency of content delivery that is caused by unstable communication links and short connection time between vehicles or between vehicles and RSUs. Therefore, resource allocation algorithms based on social attributes have been proposed and widely used in mobile networks, such as vehicular networks.…”