國立虎尾科技大學 |

Task and Motion Planning for Robots in Open Worlds.

紀錄類型:	書目-語言資料,手稿 : Monograph/item
正題名/作者:	Task and Motion Planning for Robots in Open Worlds./
作者:	Ding, Yan.
面頁冊數:	1 online resource (160 pages)
附註:	Source: Dissertations Abstracts International, Volume: 85-12, Section: B.
Contained By:	Dissertations Abstracts International85-12B.
標題:	Computer engineering. -
電子資源:	click for full text (PQDT)
ISBN:	9798383097182

Task and Motion Planning for Robots in Open Worlds.
Ding, Yan.

Task and Motion Planning for Robots in Open Worlds. - 1 online resource (160 pages)

Source: Dissertations Abstracts International, Volume: 85-12, Section: B.

Thesis (Ph.D.)--State University of New York at Binghamton, 2024.

Includes bibliographical references

To complete long-horizon tasks, real-world robots generally need to plan at both task and motion levels. Task-level planning involves generating a sequence of symbolic actions in a discrete space for long-term goals. In contrast, motion-level planning implements these actions by computing trajectories in a continuous space. Real-world robots usually operate within the context of "open-world" environments, characterized by their unpredictable and dynamic nature. Existing task and motion planning frameworks, however, tend to focus on closed-world settings, which assume the robot is provided with complete world knowledge. This discrepancy poses several challenges. At times, robots may receive underspecified requests from users. In these instances, robots should understand the requests and generate plans that meet user preferences. Additionally, robots might encounter new objects during task execution. Adapting their plans in response to these new objects is crucial. Moreover, robots must manage uncertainties from other agents, such as humans or robots, in open-world settings. Despite having a plan, they may face unexpected situations that hinder task completion.The primary contribution of this research is to develop task and motion planning algorithms and systems that are efficient, feasible, and adaptable for robots operating in open-world environments. This dissertation comprises several studies. I begin by introducing TMPUD, a task and motion planning framework for urban autonomous driving under full observability. TMPUD is designed to handle the uncertainty from other agents, aiming to enhance the safety and efficiency of urban driving. In contrast, the second framework, GLAD, operates under partial observability. The third framework, TMOC, focuses on planning with novel objects. This framework incorporates a physics engine to ground objects, learn about their physical attributes, and improve task-motion planning skills based on gathered learning experiences. Next, I discuss LLM-GROP, a method that leverages commonsense knowledge for planning to handle underspecified user requests in tasks involving multi-object rearrangement. Finally, I present COWP, a system that combines task-oriented commonsense knowledge derived from Large Language Models (LLMs) with classical AI planning methods to effectively handle unforeseen situations.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2024

Mode of access: World Wide Web

ISBN: 9798383097182Subjects--Topical Terms:

569006
Computer engineering.
Subjects--Index Terms:

Large Language ModelsIndex Terms--Genre/Form:

554714
Electronic books.

Task and Motion Planning for Robots in Open Worlds.
LDR:03724ntm a22003977 4500 001 1150141
005 20241022111557.5
006 m o d
007 cr bn ---uuuuu
008 250605s2024 xx obm 000 0 eng d
020 $a 9798383097182
035 $a (MiAaPQ)AAI30994302
035 $a AAI30994302
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Ding, Yan. $3 1476574
245 1 0 $a Task and Motion Planning for Robots in Open Worlds.
264 0 $c 2024
300 $a 1 online resource (160 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: Dissertations Abstracts International, Volume: 85-12, Section: B.
500 $a Advisor: Zhang, Shiqi.
502 $a Thesis (Ph.D.)--State University of New York at Binghamton, 2024.
504 $a Includes bibliographical references
520 $a To complete long-horizon tasks, real-world robots generally need to plan at both task and motion levels. Task-level planning involves generating a sequence of symbolic actions in a discrete space for long-term goals. In contrast, motion-level planning implements these actions by computing trajectories in a continuous space. Real-world robots usually operate within the context of "open-world" environments, characterized by their unpredictable and dynamic nature. Existing task and motion planning frameworks, however, tend to focus on closed-world settings, which assume the robot is provided with complete world knowledge. This discrepancy poses several challenges. At times, robots may receive underspecified requests from users. In these instances, robots should understand the requests and generate plans that meet user preferences. Additionally, robots might encounter new objects during task execution. Adapting their plans in response to these new objects is crucial. Moreover, robots must manage uncertainties from other agents, such as humans or robots, in open-world settings. Despite having a plan, they may face unexpected situations that hinder task completion.The primary contribution of this research is to develop task and motion planning algorithms and systems that are efficient, feasible, and adaptable for robots operating in open-world environments. This dissertation comprises several studies. I begin by introducing TMPUD, a task and motion planning framework for urban autonomous driving under full observability. TMPUD is designed to handle the uncertainty from other agents, aiming to enhance the safety and efficiency of urban driving. In contrast, the second framework, GLAD, operates under partial observability. The third framework, TMOC, focuses on planning with novel objects. This framework incorporates a physics engine to ground objects, learn about their physical attributes, and improve task-motion planning skills based on gathered learning experiences. Next, I discuss LLM-GROP, a method that leverages commonsense knowledge for planning to handle underspecified user requests in tasks involving multi-object rearrangement. Finally, I present COWP, a system that combines task-oriented commonsense knowledge derived from Large Language Models (LLMs) with classical AI planning methods to effectively handle unforeseen situations.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2024
538 $a Mode of access: World Wide Web
650 4 $a Computer engineering. $3 569006
650 4 $a Robotics. $3 561941
653 $a Large Language Models
653 $a Open worlds
653 $a Robots
653 $a Motion planning
653 $a Task planning
655 7 $a Electronic books. $2 local $3 554714
690 $a 0800
690 $a 0771
690 $a 0464
710 2 $a State University of New York at Binghamton. $b Computer Science. $3 845523
710 2 $a ProQuest Information and Learning Co. $3 1178819
773 0 $t Dissertations Abstracts International $g 85-12B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30994302 $z click for full text (PQDT)