Application of Efficient Mobile Robot Navigation through Machine Learning Technique

dc.contributor.authorZabar, Hiba Fouad
dc.contributor.authorAlaiwi, Yaser
dc.date.accessioned2025-02-06T18:01:20Z
dc.date.available2025-02-06T18:01:20Z
dc.date.issued2024
dc.departmentAltınbaş Üniversitesien_US
dc.description2nd International Conference on Engineering and Science to Achieve the Sustainable Development Goals, ICASDG 2023 -- 9 July 2023 through 10 July 2023 -- Hybrid, Tabriz -- 197984en_US
dc.description.abstractMapping and exploration of previously unexplored regions are critical competencies for autonomous mobile robots. To steer robots to the unexplored portions of their surroundings that will give the most new information to an occupancy map, information-theoretic exploration methods were created. In the interim, concurrent confinement and planning (hammer) is a technique for surveying map exactness and overseeing restriction mistakes when there is loud relative information. Thus, Hammer-based investigation, or dynamic Hammer, has been effectively used to independently plan an obscure climate while managing robot milestones and stance equivocality. Be that as it may, the dynamic part of dynamic Hammer investigation methods is tedious because of the prerequisite for forward reproduction of future robot estimations and the expectation of the resulting guide and stance vulnerability. Due to the high time complexity of such methods, this strategy will eventually fail for real-time decision-making with the rising dimensionality of the state space and the action space. In this proposed work, we introduce learning-based exploration algorithms to provide reduced computation time and near-optimal exploration techniques under uncertainty. First, we offer a method for solving autonomous mobile robot exploration problems using a robot's local map and deep reinforcement learning (DRL) without taking localization uncertainty into account. During the online testing phase, the DRL controller provides robot sensing actions that are almost as informative and efficient as those of a normal mutual information-maximizing controller while requiring significantly less computational work. © 2024 American Institute of Physics Inc.. All rights reserved.en_US
dc.identifier.doi10.1063/5.0199704
dc.identifier.issn0094-243X
dc.identifier.issue1en_US
dc.identifier.scopus2-s2.0-85188323338
dc.identifier.scopusqualityQ4en_US
dc.identifier.urihttps://doi.org/10.1063/5.0199704
dc.identifier.urihttps://hdl.handle.net/20.500.12939/5335
dc.identifier.volume3092en_US
dc.indekslendigikaynakScopus
dc.language.isoenen_US
dc.publisherAmerican Institute of Physicsen_US
dc.relation.ispartofAIP Conference Proceedingsen_US
dc.relation.publicationcategoryKonferans Öğesi - Uluslararası - Kurum Öğretim Elemanıen_US
dc.rightsinfo:eu-repo/semantics/openAccessen_US
dc.snmzKA_Scopus_20250206
dc.subjectMobile Robot Autonomyen_US
dc.subjectReinforcement Learningen_US
dc.subjectSimultaneous Localization And Mapping (SLAM)en_US
dc.titleApplication of Efficient Mobile Robot Navigation through Machine Learning Techniqueen_US
dc.typeConference Objecten_US

Dosyalar