國立虎尾科技大學 |

Digital Twin Synchronization and Control of Robot Arm-Based Manufacturing via Reinforcement Learning and Unity

紀錄類型:	書目-語言資料,印刷品 : Monograph/item
正題名/作者:	Digital Twin Synchronization and Control of Robot Arm-Based Manufacturing via Reinforcement Learning and Unity/ M. Matsive Ali.
作者:	Ali, M. Matsive,
面頁冊數:	1 electronic resource (122 pages)
附註:	Source: Masters Abstracts International, Volume: 87-02.
Contained By:	Masters Abstracts International87-02.
標題:	Computer engineering. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=32172638
ISBN:	9798290970790

Digital Twin Synchronization and Control of Robot Arm-Based Manufacturing via Reinforcement Learning and Unity
Ali, M. Matsive,

Digital Twin Synchronization and Control of Robot Arm-Based Manufacturing via Reinforcement Learning and Unity [electronic resource] /M. Matsive Ali. - 1 electronic resource (122 pages)

Source: Masters Abstracts International, Volume: 87-02.

The concept of Digital Twin (DT) for robotic systems represents a significant advancement toward smart manufacturing in Industry 5.0, poised to revolutionize the industrial market. At the core of this transformation is Artificial Intelligence (AI), which enables DT to enhance various aspects of industry, including production efficiency, safety, process scheduling, human-robot collaboration (HRC), and operational flexibility. However, hard programming is often time-consuming when aiming to achieve optimal performance. To address this, soft programming approaches such as Reinforcement Learning (RL) and virtual reality (VR) simulator platforms like Unity provide substantial assistance. A comprehensive literature review was conducted to examine the advancements in robotic manipulation using RL within Unity, aiming to extract and compare relevant characteristics. This comparison focused on the employed RL algorithms, hyperparameters, reward policies, and the validation of DT applicability. Key trends and challenges were identified, particularly in the domain of robotic DT. This research presents an innovative approach that integrates the Soft Actor-Critic (SAC) RL algorithm with DT technology to enable real-time adaptive control and path planning in robotic manufacturing. The experimental setup consists of a ViperX 300s robot arm, with two distinct in-situ defect mitigation environments: (1) static target reaching and (2) dynamic target following. The system architecture combines the Unity game engine's simulation environment with ROS2, enabling seamless and innovative DT synchronization. A hierarchical reward structure was implemented to address common RL challenges, including local minima avoidance, convergence acceleration, and training stability. Additionally, transfer learning was leveraged to efficiently adapt trained behavior models across different task domains. Experimental results demonstrated rapid policy convergence and robust task execution, with performance metrics, including cumulative reward, value loss, policy loss, and entropy confirming the effectiveness of the proposed approach. This work advances the integration of RL with realistic DT in industrial and manufacturing robotics, offering a framework for enhanced adaptive real-time control in smart additive manufacturing (AM) processes.

English

ISBN: 9798290970790Subjects--Topical Terms:

569006
Computer engineering.
Subjects--Index Terms:

Control systems

Digital Twin Synchronization and Control of Robot Arm-Based Manufacturing via Reinforcement Learning and Unity
LDR:03927nam a22004573i 4500 001 1172940
005 20260622113227.5
006 m o d
007 cr|nu||||||||
008 260803s2025 miu||||||m |||||||eng d
020 $a 9798290970790
035 $a (MiAaPQD)AAI32172638
035 $a AAI32172638
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Ali, M. Matsive, $e author. $0 (orcid)0000-0002-7345-5247 $3 1503586
245 1 0 $a Digital Twin Synchronization and Control of Robot Arm-Based Manufacturing via Reinforcement Learning and Unity $c M. Matsive Ali. $h [electronic resource] /
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2025
300 $a 1 electronic resource (122 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Masters Abstracts International, Volume: 87-02.
500 $a Includes supplementary digital materials.
500 $a Advisors: Liu, Sen Committee members: Barhorst, Alan A.; Gottumukkala, Raju; Zhang, Boyang.
502 $b M.S. $c University of Louisiana at Lafayette $d 2025.
520 # $a The concept of Digital Twin (DT) for robotic systems represents a significant advancement toward smart manufacturing in Industry 5.0, poised to revolutionize the industrial market. At the core of this transformation is Artificial Intelligence (AI), which enables DT to enhance various aspects of industry, including production efficiency, safety, process scheduling, human-robot collaboration (HRC), and operational flexibility. However, hard programming is often time-consuming when aiming to achieve optimal performance. To address this, soft programming approaches such as Reinforcement Learning (RL) and virtual reality (VR) simulator platforms like Unity provide substantial assistance. A comprehensive literature review was conducted to examine the advancements in robotic manipulation using RL within Unity, aiming to extract and compare relevant characteristics. This comparison focused on the employed RL algorithms, hyperparameters, reward policies, and the validation of DT applicability. Key trends and challenges were identified, particularly in the domain of robotic DT. This research presents an innovative approach that integrates the Soft Actor-Critic (SAC) RL algorithm with DT technology to enable real-time adaptive control and path planning in robotic manufacturing. The experimental setup consists of a ViperX 300s robot arm, with two distinct in-situ defect mitigation environments: (1) static target reaching and (2) dynamic target following. The system architecture combines the Unity game engine's simulation environment with ROS2, enabling seamless and innovative DT synchronization. A hierarchical reward structure was implemented to address common RL challenges, including local minima avoidance, convergence acceleration, and training stability. Additionally, transfer learning was leveraged to efficiently adapt trained behavior models across different task domains. Experimental results demonstrated rapid policy convergence and robust task execution, with performance metrics, including cumulative reward, value loss, policy loss, and entropy confirming the effectiveness of the proposed approach. This work advances the integration of RL with realistic DT in industrial and manufacturing robotics, offering a framework for enhanced adaptive real-time control in smart additive manufacturing (AM) processes.
546 $a English
590 $a School code: 1363
650 # 4 $a Computer engineering. $3 569006
650 # 4 $a Mechanical engineering. $3 557493
650 # 4 $a Robotics. $3 561941
653 # # $a Control systems
653 # # $a Digital twin
653 # # $a Reinforcement learning
653 # # $a Smart manufacturing
653 # # $a Unity
690 $a 0771
690 $a 0800
690 $a 0548
690 $a 0464
710 2 # $a University of Louisiana at Lafayette. $b Mechanical Engineering. $3 1180308
720 1 $a Liu, Sen $e degree supervisor.
773 0 # $t Masters Abstracts International $g 87-02.
790 $a 1363
791 $a M.S.
792 $a 2025
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=32172638