Base station selection method based on deep reinforcement learning in LTE-V

Abstract

The present invention relates to a base station selection method based on deep reinforcement learning in an LTE-V. The method comprises the following steps of: 1) constructing a Q function according to LTE-V (Long Term Evolution-Vehicle) network communication features and base station selection performance indexes; 2) obtaining state information of a vehicle in a network through a mobile management unit, constructing a state matrix, and storing the state matrix into a playback pool; 3) taking an experience playback pool as a sample, and based on the constructed Q function, employing a dueling-double training mode to obtain a main DQN (Dueling-Double Deep Q Network) used to select an optimal access base station; and 4) performing processing of the input information by employing the main DQNobtained through training, and outputting a selected access base station. Compared to the prior art, the base station selection method based on deep reinforcement learning can consider the delay performance and the load balance performance of communication at the same time to allow the vehicle to timely and reliably perform communication, and is high in base station selection efficiency and highin accuracy.

Description

technical field [0001] The invention relates to LTE-V communication technology and DRL technology, in particular to a base station selection method based on neural network continuous decision-making, which is used to reduce the congestion rate of LTE-V network. Background technique [0002] LTE-V (Long Term Evolution-Vehicle) is a V2X technology with independent intellectual property rights in my country. It is an ITS system solution based on Time Division-Long Term Evolution (TD-LTE). An important application branch of LTE subsequent evolution technology. In February 2015, the LTE-V standardization research work of the 3GPP working group was officially launched, and the proposal of Release 14 marked the official start of the LTE-V technical standard formulation work in the 3GPP working group plan, and it will also be compatible and performance in 5G a substantial increase. LTE V2V Core part was completed at the end of 2016, and LTE V2X Core part was completed at the beginn...

AI Technical Summary

Problems solved by technology

Without a reasonable congestion control scheme, the load generated by these messages will cause serious information delays and pose a severe test to the LTE network capacity

In addition, vehicles choose the base station with the best channel conditions through random competition, which can easily lead to network congestion when there is a large traffic flow

Images