Login | Register

Connectivity Preservation in Multi-Agent Systems using Model Predictive Control


Connectivity Preservation in Multi-Agent Systems using Model Predictive Control

El-Hamamsy, Ahmed (2019) Connectivity Preservation in Multi-Agent Systems using Model Predictive Control. Masters thesis, Concordia University.

[thumbnail of El-Hamamsy_MASc_S2020.pdf]
Text (application/pdf)
El-Hamamsy_MASc_S2020.pdf - Accepted Version


Flocking of multiagent systems is one of the basic behaviors in the field of control of multiagent systems and it is an essential element of many real-life applications. Such systems under various network structures and environment modes have been extensively studied in the past decades. Navigation of agents in a leader-follower structure while operating in environments with obstacles is particularly challenging. One of the main challenges in flocking of multiagent systems is to preserve connectivity. Gradient descent method is widely utilized to achieve this goal. But the main shortcoming of applying this method for the leader-follower structure is the need for continuous data transmission between agents and/or the preservation of a fixed connection topology. In this research, we propose an innovative model predictive controller based on a potential field that maintains the connectivity of a flock of agents in a leader-follower structure with dynamic topology. The agents navigate through an environment with obstacles that form a path leading to a certain target. Such a control technique avoids collisions of followers with each other without using any communication links while following their leader which navigates in the environment through potential functions for modelling the neighbors and obstacles. The potential field is dynamically updated by introducing weight variables in order to preserve connectivity among the followers as we assume only the leader knows the target position. The values of these weights are changed in real-time according to trajectories of the agents when the critical neighbors of each agent is determined. We compare the performance of our predictive-control based algorithm with other approaches. The results show that our algorithm causes the agents to reach the target in less time. However, our algorithm faces more deadlock cases when the agents go through relatively narrow paths. Due to the consideration of the input costs in our controller, the group of agents reaching the target faster does not necessarily result in the followers consuming more energy than the leader.

Divisions:Concordia University > Gina Cody School of Engineering and Computer Science > Electrical and Computer Engineering
Item Type:Thesis (Masters)
Authors:El-Hamamsy, Ahmed
Institution:Concordia University
Degree Name:M.A. Sc.
Program:Electrical and Computer Engineering
Date:17 November 2019
Thesis Supervisor(s):Aghdam, Amir and Aghili, Farhad
ID Code:986216
Deposited By: Ahmed Elhamamsy
Deposited On:26 Jun 2020 13:09
Last Modified:26 Jun 2020 13:09
All items in Spectrum are protected by copyright, with all rights reserved. The use of items is governed by Spectrum's terms of access.

Repository Staff Only: item control page

Downloads per month over past year

Research related to the current document (at the CORE website)
- Research related to the current document (at the CORE website)
Back to top Back to top