東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

FindBook

Google Book

Amazon

博客來

Provably Safe Learning-Based Robot Control Via Anomaly Detection and Generalization Theory.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Provably Safe Learning-Based Robot Control Via Anomaly Detection and Generalization Theory./
作者:	Farid, Alec Jacob.
面頁冊數:	1 online resource (279 pages)
附註:	Source: Dissertations Abstracts International, Volume: 84-10, Section: B.
Contained By:	Dissertations Abstracts International84-10B.
標題:	Robotics. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30247508click for full text (PQDT)
ISBN:	9798379403836

Provably Safe Learning-Based Robot Control Via Anomaly Detection and Generalization Theory.
Farid, Alec Jacob.

Provably Safe Learning-Based Robot Control Via Anomaly Detection and Generalization Theory. - 1 online resource (279 pages)

Source: Dissertations Abstracts International, Volume: 84-10, Section: B.

Thesis (Ph.D.)--Princeton University, 2023.

Includes bibliographical references

How can we ensure the safety of a control policy for a robot that operates using high-dimensional sensor observations such as vision (e.g., an autonomous drone or a home robot)? Machine learning techniques provide an appealing option because of their ability to deal with rich sensory inputs. However, policies that have learning-based components as part of the perception and control pipeline typically struggle to generalize to novel environments. Consider a home robot which is trained to find, grasp, and move objects using thousands of different environments, objects, and goal locations. It is inevitable that such a complex robot will encounter new settings for which it is unprepared. State-of-the-art approaches for synthesizing policies for this robotic system (e.g., based on deep reinforcement learning or imitation learning) generally do not provide assurances of safety, and can lead to policies that fail catastrophically in new environments.In this dissertation, we will present techniques we have developed to analyze the safety of robotic systems when they are deployed in new and potentially unsafe settings. We use and develop tools from generalization theory in order to leverage statistical guarantees for learning in robotics to make progress in two key research directions. (i) We provide methods for guaranteed detection of when a robotic system is unprepared for the setting it is operating in and could lead to a failure. With this knowledge, an emergency maneuver or backup safety policy can be deployed to keep the robot safe. (ii) We also develop performance bounds for settings which are out-of-distribution with respect to the training dataset. In both cases, we apply the techniques on challenging problems including vision-based drone navigation and autonomous vehicle planning and demonstrate the ability to provide strong safety guarantees for robotic systems.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2023

Mode of access: World Wide Web

ISBN: 9798379403836Subjects--Topical Terms:

519753
Robotics.
Subjects--Index Terms:

Control policyIndex Terms--Genre/Form:

542853
Electronic books.

Provably Safe Learning-Based Robot Control Via Anomaly Detection and Generalization Theory.
LDR:03526nmm a2200493K 4500 001 2360713
005 20231015184502.5
006 m o d
007 cr mn ---uuuuu
008 241011s2023 xx obm 000 0 eng d
020 $a 9798379403836
035 $a (MiAaPQ)AAI30247508
035 $a AAI30247508
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Farid, Alec Jacob. $3 3701344
245 1 0 $a Provably Safe Learning-Based Robot Control Via Anomaly Detection and Generalization Theory.
264 0 $c 2023
300 $a 1 online resource (279 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: 84-10, Section: B.
500 $a Advisor: Majumdar, Anirudha.
502 $a Thesis (Ph.D.)--Princeton University, 2023.
504 $a Includes bibliographical references
520 $a How can we ensure the safety of a control policy for a robot that operates using high-dimensional sensor observations such as vision (e.g., an autonomous drone or a home robot)? Machine learning techniques provide an appealing option because of their ability to deal with rich sensory inputs. However, policies that have learning-based components as part of the perception and control pipeline typically struggle to generalize to novel environments. Consider a home robot which is trained to find, grasp, and move objects using thousands of different environments, objects, and goal locations. It is inevitable that such a complex robot will encounter new settings for which it is unprepared. State-of-the-art approaches for synthesizing policies for this robotic system (e.g., based on deep reinforcement learning or imitation learning) generally do not provide assurances of safety, and can lead to policies that fail catastrophically in new environments.In this dissertation, we will present techniques we have developed to analyze the safety of robotic systems when they are deployed in new and potentially unsafe settings. We use and develop tools from generalization theory in order to leverage statistical guarantees for learning in robotics to make progress in two key research directions. (i) We provide methods for guaranteed detection of when a robotic system is unprepared for the setting it is operating in and could lead to a failure. With this knowledge, an emergency maneuver or backup safety policy can be deployed to keep the robot safe. (ii) We also develop performance bounds for settings which are out-of-distribution with respect to the training dataset. In both cases, we apply the techniques on challenging problems including vision-based drone navigation and autonomous vehicle planning and demonstrate the ability to provide strong safety guarantees for robotic systems.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2023
538 $a Mode of access: World Wide Web
650 4 $a Robotics. $3 519753
650 4 $a Automotive engineering. $3 2181195
650 4 $a Aerospace engineering. $3 1002622
650 4 $a Mechanical engineering. $3 649730
653 $a Control policy
653 $a Autonomous drone
653 $a Machine learning
653 $a Environments
653 $a Sensor observations
653 $a Home robot
653 $a Reinforcement learning
653 $a Robotic systems
653 $a Anomaly detection
653 $a Generalization theory
653 $a Safe learning
655 7 $a Electronic books. $2 lcsh $3 542853
690 $a 0771
690 $a 0800
690 $a 0548
690 $a 0538
690 $a 0540
710 2 $a ProQuest Information and Learning Co. $3 783688
710 2 $a Princeton University. $b Mechanical and Aerospace Engineering. $3 2102828
773 0 $t Dissertations Abstracts International $g 84-10B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30247508 $z click for full text (PQDT)