國立虎尾科技大學 |

Deep Generative Models for Competency Awareness and Uncertainty Quantification.

Record Type:	Language materials, manuscript : Monograph/item
Title/Author:	Deep Generative Models for Competency Awareness and Uncertainty Quantification./
Author:	Acharya, Aastha.
Description:	1 online resource (129 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 85-06, Section: B.
Contained By:	Dissertations Abstracts International85-06B.
Subject:	Aerospace engineering. -
Online resource:	click for full text (PQDT)
ISBN:	9798381165593

Deep Generative Models for Competency Awareness and Uncertainty Quantification.
Acharya, Aastha.

Deep Generative Models for Competency Awareness and Uncertainty Quantification. - 1 online resource (129 pages)

Source: Dissertations Abstracts International, Volume: 85-06, Section: B.

Thesis (Ph.D.)--University of Colorado at Boulder, 2023.

Includes bibliographical references

As they become increasingly advanced, autonomous agents are slowly integrating in the real world and working alongside human users. To establish appropriate levels of trust between the human users and the autonomous agents, autonomous agents can communicate their competencies. One approach to competency awareness is to provide the agents with the ability to self-assess their own behaviors and outcomes using trajectories, which are highly impacted by the uncertainties faced in the environment. This work focuses on generating forecasted trajectories using a probabilistic world model, i.e. a model of the environment, which fully accounts for the encountered uncertainties. To equip an autonomous agent with the ability to forecast accurate and reliable trajectories, we use deep generative models as our probabilistic world model. This thesis contains several key contributions, each of which build up to a robust and efficient probabilistic world model with uncertainties quantified. First, in Chapter 3, we provide a feasibility study on the usage of trajectories to form human-interpretable abstract behavior models that identify the conditions leading to different strategies and outcomes for the agent. In Chapter 4, two different forms of deep generative models, variational autoencoders and flow-based models, are analysed on the task of conditional density estimation. Chapter 5 presents the development of probabilistic world model using variational autoencoder, which is then used by the autonomous agent to perform trajectory forecasting. Finally, in chapter 6, we incorporate model, or epistemic, uncertainty into trajectory forecasting and show how deep generative models can be used to form an end-to-end uncertainty quantification and trajectory forecasting system. The contributions of this thesis take us one step closer to creating trustworthy autonomous agents with the ability to effectively analyze and communicate their competencies.

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

Mode of access: World Wide Web

ISBN: 9798381165593Subjects--Topical Terms:

686400
Aerospace engineering.
Subjects--Index Terms:

Deep generative modelIndex Terms--Genre/Form:

554714
Electronic books.

Deep Generative Models for Competency Awareness and Uncertainty Quantification.
LDR:03344ntm a22003857 4500 001 1145199
005 20240617111407.5
006 m o d
007 cr mn ---uuuuu
008 250605s2023 xx obm 000 0 eng d
020 $a 9798381165593
035 $a (MiAaPQ)AAI30810791
035 $a AAI30810791
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Acharya, Aastha. $3 1470439
245 1 0 $a Deep Generative Models for Competency Awareness and Uncertainty Quantification.
264 0 $c 2023
300 $a 1 online resource (129 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: 85-06, Section: B.
500 $a Advisor: Ahmed, Nisar.
502 $a Thesis (Ph.D.)--University of Colorado at Boulder, 2023.
504 $a Includes bibliographical references
520 $a As they become increasingly advanced, autonomous agents are slowly integrating in the real world and working alongside human users. To establish appropriate levels of trust between the human users and the autonomous agents, autonomous agents can communicate their competencies. One approach to competency awareness is to provide the agents with the ability to self-assess their own behaviors and outcomes using trajectories, which are highly impacted by the uncertainties faced in the environment. This work focuses on generating forecasted trajectories using a probabilistic world model, i.e. a model of the environment, which fully accounts for the encountered uncertainties. To equip an autonomous agent with the ability to forecast accurate and reliable trajectories, we use deep generative models as our probabilistic world model. This thesis contains several key contributions, each of which build up to a robust and efficient probabilistic world model with uncertainties quantified. First, in Chapter 3, we provide a feasibility study on the usage of trajectories to form human-interpretable abstract behavior models that identify the conditions leading to different strategies and outcomes for the agent. In Chapter 4, two different forms of deep generative models, variational autoencoders and flow-based models, are analysed on the task of conditional density estimation. Chapter 5 presents the development of probabilistic world model using variational autoencoder, which is then used by the autonomous agent to perform trajectory forecasting. Finally, in chapter 6, we incorporate model, or epistemic, uncertainty into trajectory forecasting and show how deep generative models can be used to form an end-to-end uncertainty quantification and trajectory forecasting system. The contributions of this thesis take us one step closer to creating trustworthy autonomous agents with the ability to effectively analyze and communicate their competencies.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2024
538 $a Mode of access: World Wide Web
650 4 $a Aerospace engineering. $3 686400
650 4 $a Robotics. $3 561941
653 $a Deep generative model
653 $a Probabilistic world model
653 $a Uncertainty quantification
653 $a Variational autoencoder
655 7 $a Electronic books. $2 local $3 554714
690 $a 0538
690 $a 0800
690 $a 0771
710 2 $a ProQuest Information and Learning Co. $3 1178819
710 2 $a University of Colorado at Boulder. $b Aerospace Engineering. $3 1179005
773 0 $t Dissertations Abstracts International $g 85-06B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30810791 $z click for full text (PQDT)