서브메뉴
검색
Online Adaptation for Safe Control of Constrained Dynamical Systems
Online Adaptation for Safe Control of Constrained Dynamical Systems
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202103639
- ISBN
- 9798314873311
- DDC
- 629.8
- 저자명
- Parwana, Hardik.
- 서명/저자
- Online Adaptation for Safe Control of Constrained Dynamical Systems
- 발행사항
- [Sl] : University of Michigan, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 219 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-11, Section: B.
- 주기사항
- Advisor: Panagou, Dimitra.
- 학위논문주기
- Thesis (Ph.D.)--University of Michigan, 2025.
- 초록/해제
- 요약Advances in sensing modalities and computational power have led to the prospect of a widespread deployment of robots in our society. Central to this objective is developing control and navigation stacks that avoid conservatism, presumed to be measured by a performance metric, while being provably and practically safe. A crucial element that must be accounted for is that controllers, which are typically designed for and tuned in laboratory or highly monitored industrial settings for a specific scenario, may experience a drop in performance and lose their safety guarantees when used elsewhere. It is of paramount importance therefore to import robots with the capability to adapt their controllers online to customize responses to a priori untested environments.In brief, this dissertation presents (1) tools to adapt any parametric controller using a model-based approach to achieve simultaneous satisfaction of multiple state constraints and enhanced performance; (2) a numerical scheme for predicting future state distributions in systems governed by stochastic dynamics with state-dependent disturbances, which can be utilized in model-predictive approaches; and (3) a method to assist decision-making on dropping (disregarding) constraints when it is not feasible to satisfy all constraints simultaneously.A significant part of the dissertation focuses on a specific safety-critical control method called control barrier functions (CBF). The CBF-based controllers have garnered interest in recent years due to their ease of implementation. However, finding a theoretically valid CBF remains a challenge and in practice, they are prone to performance degradation and safety violations, especially when multiple CBFs are imposed together. This dissertation introduces a new notion of CBFs, called Rate-Tunable CBFs, that allows for time-varying parameters in theory and online tuning in practice.The dissertation starts with an introductory chapter reviewing safety-critical controllers and planners under deterministic and stochastic settings. The second chapter provides a thorough technical review of CBFs as well as concepts from optimization, probability and set invariance relevant to this work. The third and fourth chapters focus almost exclusively on control barrier functions. The third chapter formally introduces the notion of compatibility of multiple CBF constraints and then follows up with a metric - the volume of the feasible solution space of the QP - to quantify distance to the infeasibility of CBF-QP controllers. A new CBF is designed to prevent the volume from going to zero thereby ensuring the existence of a solution to the CBF-QP controller at all times. The fourth chapter then introduces our notion of Rate-Tunable CBFs that allow the parameters of class-K function to vary with time. This allows tuning of the response of CBF-based controllers. Two methods, one instantaneously locally optimal and another based on a model-predictive approach employing gradient-descent on parameters are introduced to design the parameter dynamics.The fifth chapter examines the effects of uncertain dynamics with state-dependent disturbances. A numerical scheme called Expansion-Compression (EC) Layers, based on the Unscented Transform (UT), is proposed to predict future state distributions. The UT is a weighted, particle-based method; the expansion layer increases the number of particles to represent increased uncertainty due to state-dependent disturbances, while the compression layer uses moment-matching to consolidate these particles into fewer, representative points, resulting in a scalable scheme. Applications of the EC-UT are shown in the gradient-based model predictive auto-tuning framework and the model predictive path integral controller, which advances the state of the art in sample efficiency.The sixth chapter takes a departure from parameter adaptation and imparts the controller the capability to permanently drop a constraint from its optimization problem. Such scenarios are of interest when some low-priority task specifications, imposed as state-input constraints, conflict with the safety or high-priority task constraints and need to be sacrificed. An algorithm is proposed to drop the minimum number of constraints under an additive priority scheme, a problem that is known to be NP-Hard but unexplored in the context of dynamical systems where the optimization plays the role of a controller. Towards this, a Lagrange multiplier-based heuristic that keeps track of active constraints in the past is introduced to form a more informed prior for solving the NP-Hard problem. It is shown empirically that the proposed heuristic outperforms the existing slack variable-based heuristics.
- 일반주제명
- Robotics
- 일반주제명
- Computer engineering
- 키워드
- Safe control
- 기타저자
- University of Michigan Robotics
- 기본자료저록
- Dissertations Abstracts International. 86-11B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017358068
■00520260202103639
■006m o d
■007cr#unu||||||||
■020 ▼a9798314873311
■035 ▼a(MiAaPQ)AAI32092498
■035 ▼a(MiAaPQ)umichrackham006051
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a629.8
■1001 ▼aParwana, Hardik.
■24510▼aOnline Adaptation for Safe Control of Constrained Dynamical Systems
■260 ▼a[Sl]▼bUniversity of Michigan▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a219 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-11, Section: B.
■500 ▼aAdvisor: Panagou, Dimitra.
■5021 ▼aThesis (Ph.D.)--University of Michigan, 2025.
■520 ▼aAdvances in sensing modalities and computational power have led to the prospect of a widespread deployment of robots in our society. Central to this objective is developing control and navigation stacks that avoid conservatism, presumed to be measured by a performance metric, while being provably and practically safe. A crucial element that must be accounted for is that controllers, which are typically designed for and tuned in laboratory or highly monitored industrial settings for a specific scenario, may experience a drop in performance and lose their safety guarantees when used elsewhere. It is of paramount importance therefore to import robots with the capability to adapt their controllers online to customize responses to a priori untested environments.In brief, this dissertation presents (1) tools to adapt any parametric controller using a model-based approach to achieve simultaneous satisfaction of multiple state constraints and enhanced performance; (2) a numerical scheme for predicting future state distributions in systems governed by stochastic dynamics with state-dependent disturbances, which can be utilized in model-predictive approaches; and (3) a method to assist decision-making on dropping (disregarding) constraints when it is not feasible to satisfy all constraints simultaneously.A significant part of the dissertation focuses on a specific safety-critical control method called control barrier functions (CBF). The CBF-based controllers have garnered interest in recent years due to their ease of implementation. However, finding a theoretically valid CBF remains a challenge and in practice, they are prone to performance degradation and safety violations, especially when multiple CBFs are imposed together. This dissertation introduces a new notion of CBFs, called Rate-Tunable CBFs, that allows for time-varying parameters in theory and online tuning in practice.The dissertation starts with an introductory chapter reviewing safety-critical controllers and planners under deterministic and stochastic settings. The second chapter provides a thorough technical review of CBFs as well as concepts from optimization, probability and set invariance relevant to this work. The third and fourth chapters focus almost exclusively on control barrier functions. The third chapter formally introduces the notion of compatibility of multiple CBF constraints and then follows up with a metric - the volume of the feasible solution space of the QP - to quantify distance to the infeasibility of CBF-QP controllers. A new CBF is designed to prevent the volume from going to zero thereby ensuring the existence of a solution to the CBF-QP controller at all times. The fourth chapter then introduces our notion of Rate-Tunable CBFs that allow the parameters of class-K function to vary with time. This allows tuning of the response of CBF-based controllers. Two methods, one instantaneously locally optimal and another based on a model-predictive approach employing gradient-descent on parameters are introduced to design the parameter dynamics.The fifth chapter examines the effects of uncertain dynamics with state-dependent disturbances. A numerical scheme called Expansion-Compression (EC) Layers, based on the Unscented Transform (UT), is proposed to predict future state distributions. The UT is a weighted, particle-based method; the expansion layer increases the number of particles to represent increased uncertainty due to state-dependent disturbances, while the compression layer uses moment-matching to consolidate these particles into fewer, representative points, resulting in a scalable scheme. Applications of the EC-UT are shown in the gradient-based model predictive auto-tuning framework and the model predictive path integral controller, which advances the state of the art in sample efficiency.The sixth chapter takes a departure from parameter adaptation and imparts the controller the capability to permanently drop a constraint from its optimization problem. Such scenarios are of interest when some low-priority task specifications, imposed as state-input constraints, conflict with the safety or high-priority task constraints and need to be sacrificed. An algorithm is proposed to drop the minimum number of constraints under an additive priority scheme, a problem that is known to be NP-Hard but unexplored in the context of dynamical systems where the optimization plays the role of a controller. Towards this, a Lagrange multiplier-based heuristic that keeps track of active constraints in the past is introduced to form a more informed prior for solving the NP-Hard problem. It is shown empirically that the proposed heuristic outperforms the existing slack variable-based heuristics.
■590 ▼aSchool code: 0127.
■650 4▼aRobotics
■650 4▼aComputer engineering
■653 ▼aSafe control
■653 ▼aControl under uncertainty
■653 ▼aOnline controller adaptation
■653 ▼aState constraints
■653 ▼aStochastic dynamics
■690 ▼a0771
■690 ▼a0464
■690 ▼a0800
■71020▼aUniversity of Michigan▼bRobotics.
■7730 ▼tDissertations Abstracts International▼g86-11B.
■790 ▼a0127
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17358068▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


