Tube-based NMPC for Non-Holonomic Multi-agent System in Unknown Environments: Prelude to Modern Control
Translated title
Tube-based NMPC for Non-Holonomic Multi-agent System in Unknown Environments
Author
Hassan, Hamza Abdinassir
Term
4. semester
Education
Publication year
2024
Submitted on
2024-05-30
Pages
45
Abstract
Denne afhandling undersøger, hvordan man styrer en gruppe agenter i et fælles arbejdsområde, selv når de påvirkes af forstyrrelser og skal overholde input- og tilstandsbegrænsninger. Målet er, at hver agent følger en ønsket trajektorie, undgår kollisioner og bevarer netværksforbindelser. Vi udvikler en decentraliseret tube-baseret ikke-lineær modelprædiktiv styring (NMPC). Kort fortalt forudsiger NMPC fremtidig adfærd og opdaterer beslutninger løbende; den tube-baserede del sikrer, at den virkelige bevægelse holdes inden for et sikkert område omkring den planlagte bane for at håndtere forstyrrelser; og decentralisering betyder, at hver agent kan styre lokalt. Eksperimenter viser, at metoden følger de ønskede baner og dæmper effekten af forstyrrelser. Leder- og følgeagenter har små afstands- og orienteringsfejl, mens strategien samtidig undgår forhindringer, forhindrer kollisioner mellem agenter og overholder kommunikationsbegrænsninger.
This thesis examines how to control a group of agents sharing a workspace, even when they face disturbances and must respect input and state constraints. The goal is for each agent to follow a desired trajectory, avoid collisions, and keep network connectivity. We develop a decentralized tube-based Nonlinear Model Predictive Control (NMPC). In brief, NMPC predicts future behavior and updates decisions continuously; the tube-based component keeps the actual motion within a safe region around the planned path to handle disturbances; and decentralization means each agent can control locally. Experiments show the method tracks desired trajectories and reduces the impact of disturbances. Leader and follower agents maintain small distance and orientation errors, while the strategy also avoids obstacles, prevents inter-agent collisions, and satisfies communication constraints.
[This summary has been rewritten with the help of AI based on the project's original abstract]
Documents