Dynamically Allocating Network Resources for Teleoperation Applications using Intent-based Inference
Loading...
URL
Journal Title
Journal ISSN
Volume Title
Perustieteiden korkeakoulu |
Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.
Authors
Date
2022-12-12
Department
Major/Subject
Data Science
Mcode
SCI3115
Degree programme
Master's Programme in ICT Innovation
Language
en
Pages
39
Series
Abstract
Due to the limitation of network resource capacity, maintaining high-quality communication for large-scale teleoperation applications, in terms of ultra-low latency and ultra-reliability requirements, is challenging. Most network infrastructures are not designed for such extreme demands compared to other applications. However, if the future behavior of the teleoperation application is captured, we can allocate network resources more efficiently. Using the intent-based inference mechanism to predict future behaviors, the goal of this research is to deliver a low latency environment while guaranteeing high-reliability requirements. Considering a group of remote-controlled robots working in an industrial environment, the intent is interpreted as the target destination to which the robot is heading. This research proposes a Recurrent Neural Network (RNN) solution for network systems, which predicts the target destination and allocates network resources based on the predicted result. The objective is to minimize the unproductive time of robots, which is the duration robots stay idle. To quantify the performance, the proposed solution is tested under a comprehensive simulator that allows multiple actors and data streams to work simultaneously. Results show that the RNN integration for robotic applications can significantly improve network performance under different scenarios. However, care should be taken when applying since not all network systems benefit positively and equally from proposed solutions.Description
Supervisor
Le, Viet DucThesis advisor
Elbamby, MohammedKeywords
robot destination prediction, recurrent neural network, scheduling and matching, network simulation