A Task Offloading Method Based on Deep Reinforcement Learning in Internet of Vehicles

Abstract

The invention discloses a task offloading and resource allocation method based on mobile edge computing in the Internet of Vehicles, that is, a task offloading method based on deep reinforcement learning in the Internet of Vehicles, and the task offloading method through deep reinforcement learning is applied to the processing of the Internet of Vehicles The task offloading of high-dimensional state information, the specific steps include: (1) Construct the system model framework to establish the communication scene of the Internet of Vehicles; (2) Model refinement, modeling the task processing of the user side and the server side; (3) Propose a high reliability Low latency constraints and optimization problems; (4) Transform optimization problems and introduce Markov decision process; (5) Establish optimization problem models and propose five URLLC-aware task offloading algorithms based on Deep Q‑learning Network Step composition. The present invention considers both the average measurement performance and the performance of high-order statistics, and considers the impact of extreme events on communication reliability, so that the user's vehicle can make the optimal task offloading decision through deep learning while ensuring URLLC communication requirements to meet other requirements. High-reliability and low-latency communication requirements for many applications.

Description

technical field [0001] The invention relates to the technical field of task offloading and resource allocation of mobile edge computing in the Internet of Vehicles, in particular to a task offloading method based on deep reinforcement learning. Background technique [0002] Emerging vehicular applications, such as autonomous driving, real-time traffic monitoring, and online gaming, generate a large number of computationally intensive and latency-sensitive tasks, placing stringent requirements on ultra-reliable and low-latency communications (URLLC). In the traditional vehicle-edge computing (VEC) paradigm, user vehicles (UVs) collaborate with edge servers deployed at the edge of the network, such as roadside units (RSUs), by offloading excessive tasks to edge servers for computation. However, , due to factors such as fixed location of edge servers, limited coverage, and high deployment costs, it is difficult for VEC alone to effectively meet the strict URLLC requirements. In...

AI Technical Summary

Problems solved by technology

[0004] Second, Global State Information (GSI) is agnostic

More importantly, in a complex Internet of Vehicles environment, with the increase in the number of UVs, RSUs, and VFSs, the dimension of the environmental state information that needs to be considered by the user's vehicle to make decisions increases exponentially. This dilemma is called dimensional Curse, that is, the environment in which the user's vehicle is located, the information is too complex

makes it difficult for traditional learning-based methods to solve this problem with lower complexity

Images