Theoretical Developments and Applications of Physics-Informed Neural Networks for Solid Mechanics Analysis

Le, Duc Thang 세종대학교 대학원 2025 국내박사

Despite great achievements in accurately simulating solid mechanics problems by discretizing partial differential equations (PDEs), numerical methods often confront various obstacles relating to their poor scalability and huge computing requirements in achieving high-quality results when dealing with high-dimensional problems because of the requirement of fine mesh generation. In addition, parametric PDEs cannot be solved directly by these traditional techniques, significantly limiting their applicability for tackling iterative tasks due to the curse of dimensionality. Thus, conventional data-driven machine learning techniques have arisen as alternatives for effectively simulating mechanical responses in solid mechanics, but they need big data that are not always available to configure the models and potentially have unsatisfactory generalization and extrapolation performance. In order to overcome this considerable limitation, physics-informed neural networks (PINNs) have been proposed recently in the literature. In this approach, physics laws are encoded and embedded into PINN training process to produce predictive models. The PINN methodology, despite its proliferation in successfully tackling PDE issues in recent years, encounters several arduous difficulties in terms of its trainability and applicability for analyzing real-world and non-trivial solid mechanics problems. Firstly, it is observed that poor convergence in training vanilla PINN approaches caused by ill-conditioned loss function and gradient failures is usually a critical restriction and emerged significantly when dealing with PDEs having non-trivial parameters. Secondly, original PINNs could produce good prediction quality for trivial PDE problems but might fail to efficiently tackle these issues with uncommon coefficients, relieving the scalability of PINN methods for handling practical issues in solid mechanics. Thirdly, some preliminary works experimentally pointed out that the preciseness of numerical integration calculation of PINN loss function plays a vital role in stabilizing PINN training process and thereby producing high-quality prediction, but this topic hitherto has been overlooked in the literature. Finally, the extension of PINNs to solve parametric problems in solid mechanics is currently limited in practice due to its poor scalability and applicability to complex geometries and nonlinearity, restricting the PINNs for other iterative assignments such as optimal and/or inverse design and uncertainty quantification. Throughout this dissertation, underlying knowledge and intrinsic limitations of PINN methods drawn from both theoretical and experimental analyses are validated and discussed by appropriate examinations. They are solid foundations to develop advanced PINN techniques in this thesis to effectively analyze both single and parametric solid mechanics issues for real-world applications.

Physics-Informed Neural Networks for 4D Flow MRI Enhancement: Towards Super-Resolution, Denoising, and Flow Rate Consistency

강지훈 강원대학교 대학원 2026 국내박사

Four-dimensional flow magnetic resonance imaging (4D Flow MRI) provides time- resolved, three-dimensional visualization of cardiovascular hemodynamics, but its accuracy is limited by noise, low resolution, and phase artifacts. These deficiencies hinder reliable estimation of critical biomarkers such as wall shear stress and volumetric flow rate. This study presents Physics-Informed Neural Networks (PINNs) for reconstructing high-fidelity, physically consistent velocity fields from degraded 4D flow MRI data. The network integrates the Navier–Stokes, continuity equations, and flow rate constraints into a composite loss function that enforces data fidelity and physical constraints. Loss normalization and projecting conflicting gradient (PCGrad) optimization maintain balance among multiple objectives, while a learnable turbulent viscosity term improves stability in turbulent flows. Validation across synthetic, in-vitro, and in-vivo datasets—including aortic stenosis and regurgitation cases—demonstrated substantial improvements in flow rate consistency and velocity reconstruction accuracy. By embedding physical laws into deep learning optimization, the proposed PINNs achieve simultaneous denoising, super-resolution, and physiological coherence. This physics-constrained paradigm bridges computational fluid dynamics and medical imaging, advancing 4D flow MRI toward robust and clinically reliable hemodynamic assessment. 4차원 유동 자기공명영상(4D Flow MRI)은 시간에 따른 3차원 심혈관 혈류역학을 시각화할 수 있으나, 잡음, 낮은 공간 해상도, 위상(phase) 아티팩트로 인해 정확도가 제한된다. 이러한 결함은 벽전단응력(wall shear stress)과 체적 유량(volumetric flow rate)과 같은 핵심 바이오마커를 신뢰성 있게 추정하는 데 장애가 된다. 본 연구는 열화된 4D Flow MRI 데이터로부터 고충실도이면서 물리적으로 일관된 속도장을 재구성하기 위해 물리 정보 신경망(Physics-Informed Neural Networks, PINNs)을 제안한다. 제안한 네트워크는 Navier–Stokes 방정식, 연속방정식, 그리고 유량 제약을 복합 손실함수에 통합하여 데이터 적합성과 물리 제약을 동시에 만족하도록 한다. 손실 정규화와 상충 그래디언트 투영(PCGrad) 최적화를 통해 다중 목적 간 균형을 유지하며, 학습 가능한 난류 점성 항을 도입하여 난류 유동에서의 학습 안정성을 향상시킨다. 합성(synthetic), in-vitro, in-vivo 데이터셋(대동맥 협착 및 역류 사례 포함) 전반에 대한 검증 결과, 유량 일관성과 속도 재구성 정확도가 유의미하게 개선됨을 확인하였다. 물리 법칙을 딥러닝 최적화 과정에 내재화함으로써, 제안한 PINNs는 잡음 제거, 초해상도, 생리학적 정합성을 동시에 달성한다. 이러한 물리 제약 기반 패러다임은 전산유체역학(CFD)과 의료영상의 간극을 연결하여, 4D Flow MRI가 보다 견고하고 임상적으로 신뢰 가능한 혈류역학 평가로 발전하는 데 기여한다.

Making theory: I. Producing physics and physicists in postwar America. II. Post-inflation reheating in an expanding universe

Kaiser, David Isaac Harvard University 2000 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

This dissertation examines the reinvention of theoretical physics in the United States through pedagogical means after World War II. Physics graduate student enrollments ballooned immediately after the war. The unprecedented enrollments forced questions of procedures and standards for graduate training as never before. At the same time, the crush of numbers spurred an increased bureaucratization and, at least some American physicists feared, a different system of values than what had prevailed during the quieter interwar period. Out of these new bureaucratic and pedagogical developments, theoretical physics became a recognized specialty within American physics, surrounded by new ideas about what theory was for and how students should be trained to do it. Two case studies focus on developments within theoretical physics after the war, using pedagogy as a lens through which to understand the links between practices and practitioners. Within nuclear and particle physics, as Part II discusses, young graduate students and postdoctoral fellows puzzled over how to calculate with, and how to interpret, the simple line-drawings introduced by Richard Feynman in 1948. The number of distinct pictorial forms, calculational roles, and attributed meanings for the simple stick-figures quickly multiplied: rather than commanding a single use or interpretation, the diagrams came to be used for a wide variety of distinct tasks. Some theorists clung to the diagrams even as they declared the original theoretical framework from which the diagrams had sprung to be “sterile” and “dead.” These young theorists drew the diagrams much the same way as Feynman had, yet read content into them which had no correlate in the older approaches. Part III uses pedagogy to make sense of a similar series of changes within the long-dormant field of gravitational physics. Einstein's gravitational field equations proved to be no more obvious or auto-interpreting than Feynman's diagrams had been. Physicists after the war crafted new theoretical practices with which to approach questions within gravitational theory. During the middle decades of this century, American graduate students in physics learned to treat Einstein's famous theory in ways which would have been totally unrecognizable to Einstein himself—even as these new recruits were being trained to “do” gravitational physics.

The Role of Gender in Physics Peer Recognition

Sundstrom, Meagan Cornell University ProQuest Dissertations & Theses 2024 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

The under-representation of women in undergraduate science courses is well-documented. One significant challenge is that women may both perceive and receive less recognition from their science peers about their abilities as scientists than men. Here we investigate the presence and nature of such gender biases in peer recognition in the discipline of physics specifically.First, we examine the extent to which three different instructional physics contexts exhibit a gender bias in received peer recognition by asking students to list their strong physics peers on a survey. We find that there is a gender bias (in which students disproportionately recognize men as strong in their physics course more than women) in physics courses aimed at first-year, but not beyond first-year, students.We then analyze possible mechanisms underlying this gender bias. Asking students to both nominate their strong physics peers and explain their reasons for these nominations, we find an effect of gender on what skills students are recognized for in lab, but not lecture, physics courses. In both kinds of courses, we find a strong association between peer interactions and peer recognition: of the peers with whom students interact, students disproportionately select peers of their same gender to nominate as a strong student.In the third chapter, we investigate received peer recognition over a two-semester introductory physics course sequence at a mostly-women institution. We observe that while general patterns of recognition are stable over time for the same cohort of students, the most highly nominated students are subject to fluctuations that are closely tied to changes in student outspokenness.Finally, we directly compare students' received recognition (the number of nominations they receive from peers as strong in their physics course) and perceived recognition (the extent to which they feel recognized by their peers as a physics person) across student gender. We find that for men and women receiving the same amount of peer recognition (and having the same race or ethnicity, academic year, and academic major), men report significantly higher perceptions of their recognition than women.Together, these four studies provide a strong foundation for our understanding of who and what gets recognized in physics peer recognition, with a focus on the role of gender in such recognition. This body of research lays the groundwork for future studies that design, implement, and evaluate instructional activities aimed at mitigating gender differences in peer recognition. Such interventions have the potential to retain more women and other marginalized groups in physics.

Development of physics-based neural network model for centrifugal chiller

라선중 서울대학교 대학원 2024 국내박사

Modeling approaches are categorized into the physics-based model, hybrid- based model, and data-driven model. The modeling approach for HVAC systems has been increasingly shifting towards data-driven methods because the data-driven models can be utilized for various purposes, such as fault detection and diagnosis, model predictive control, etc. However, the data-driven approach sometimes leads to reduced reliability of the model. Since a machine learning model only learns patterns from the training data, various issues can arise depending on the conditions of the training data. Additionally, machine learning models may forecast physically unrealistic or inconsistent results due to a lack of extrapolation ability. Therefore, the author proposes physics-based neural networks to model HVAC systems, focusing on water-cooled chillers in office buildings. PbNNs can be categorized into three types: Physics-guided neural networks (PgNNs), Physics-informed neural networks (PiNNs), and Physics-encoded neural networks (PeNNs). The author developed the three models to predict the energy consumption of centrifugal chillers in both a DOE reference large office building and an existing building. The author verified the reliability of models in terms of accuracy metrics, generalization performance, and adherence to the physical laws of the model’s inherent prediction (Kendall’s tau coefficient). As a result, PiNNs show superior performance across all metrics (virtual building: MAPE 3.9%, Kendall’s tau coefficient: 0.85, existing building: MAPE 8.4%, Kendall’s tau coefficient: 0.66). Finally, it is highlighted in the paper that PiNNs can improve the reliability of HVAC system models. HVAC 시스템의 모델링 접근법은 물리 법칙 기반 모델, 데이터 기반 모델, 하이브리드 기반 모델로 분류된다. 최근, HVAC 시스템의 주로 활용되는 모델링 접근 방법으로써, 데이터 기반 모델은 측정된 운영 데이터의 학습을 기반으로 고장 감지 및 진단(FDD), 모델 예측 제어(MPC) 등 다양한 목적으로 활발히 활용되고 있다. 그러나 데이터 기반 모델은 물리 법칙을 직접 학습하지 않고 측정된 데이터에 대한 통계적인 학습에 의존하며, 이에 따라 훈련 데이터 외의 예측 정확도가 낮아지는 경향이 존재한다. 또한, 기계 학습 모델은 외삽 능력이 부족하여 물리적으로 비현실적이거나 일관성 없는 결과를 예측할 수 있다. 따라서 저자는 터보 냉동기를 모델링하기 위해 물리 지식 기반 신경망(physics-based neural networks, PbNNs)을 제안한다. 물리 지식 기반 신경망은 물리 지식 가이드 신경망(physics-guided neural networks, PgNNs), 물리 지식 정보 신경망(physics-informed neural networks, PiNNs), 물리 지식 인코딩 신경망(physics-encoded neural networks, PeNNs)으로 분류할 수 있다. 저자는 실제 건물과 미국 에너지성의 가상 대형 사무소 건물에서 터보 냉동기의 동적 거동(에너지 소비량)을 예측하기 위해, 물리 지식 가이드 신경망, 물리 지식 정보 신경망, 물리 지식 인코딩 신경망을 개발하였다. 저자는 기존 정확도 지표(MBE, MAPE, CVRMSE, R2)를 활용하여 모델의 일반화 성능을 평가하고 다양한 제어 환경에서 모델의 성능을 물리 법칙과의 유사성(Kendall’s tau coefficient)과 시각화 기법을 활용하여 모델의 신뢰성을 검증했다. 그 결과, PiNNs은 예측 성능과 물리 법칙 준수 성능에서 모두 우수한 결과를 도출하였다(가상 건물: MAPE 3.9%, Kendall’s tau coefficient: 0.85, 실제 건물: MAPE 8.4%, Kendall’s tau coefficient: 0.66). 결과적으로, 본 연구에서 개발한 PiNNs은 건물 운영 단계에서 활용되는 시뮬레이션 모델의 신뢰성을 향상시켜 HVAC 시스템의 운영 효율성을 개선하는 데 기여할 수 있는 모델임을 증명한다. 또한, HVAC 시스템에 대한 기계학습 모델의 정량적 성능 평가를 통해, 모델 개발 과정에서 목적에 맞는 신뢰할 수 있는 모델을 평가하고 개발하는 데 기여한다.

고등학교 양자물리 수업 지도안 개발

구성현 창원대학교 2019 국내석사

Physics is composed of various disciplines. At the beginning of the 20th century, the emergence of a new way of thinking that can explain the micro-world has led to the birth of the discipline 'quantum physics'. Various observations became possible by explaining the motion of particles in the micro-world. Quantum physics has become an essential science for the development of modern society. As the importance of modern technology emerged, the proportion of quantum physics increased in the curriculum. However, the revised curriculum in 2009 caused a lot of confusion in the field due to the large amount and content of excessive applications. The revised curriculum in 2015 solved the preceding problems, but lacked enough information to understand the basic concepts of quantum physics. In textbooks, there is no mention of why quantum physics has occurred. In addition, there is no intriguing introduction, and basic information on quantum physics is omitted. Through this study, we developed quantum physics auxiliary textbook and teaching guide plan to complement the problems. I have created a quantum physics auxiliary textbook that will stimulate students' interest and give a definition of 'quantum physics' and the overall flow. Based on this, I developed a teaching guide plan. Also, in order to achieve the nature and goals of the 2015 revised curriculum, we were careful when developing teaching materials and teaching plans. I introduced the introductory essay to induce students' interest, and designed a discussion class based on a simple question, so that I could develop the science core competence of thinking and communicating ability. It also explains precisely the definition of 'quantum physics' and the exact meaning of 'quantum', thereby enhancing understanding of science. This course introduces the typical characteristics of quantum physics so that students can understand the basic contents before introducing the unit. I described the history of science through the history and scientists of quantum physics. Then, I introduce the application field of quantum dot and understand the principle and phenomenon through 'inquiry activity'. This enabled us to develop science core competence, inquiry skills. In the case of physics Ⅱ, since the curriculum has not yet been applied, it is not possible to teach the students to the actual school. However, it is expected that the auxiliary textbook and teaching guide plan developed in this study will be very helpful to students and teachers in the revised curriculum in 2015. 물리학은 다양한 학문으로 이루어져 있다. 20세기 초, 미시세계를 설명할 수 있는 새로운 사고방식의 등장으로 ‘양자물리’ 라는 학문이 탄생하게 되었다. 미시세계의 입자의 운동을 설명하게 되면서 다양한 관측이 가능해졌다. 양자물리는 현대사회의 발전에 필수적인 학문으로 자리 잡았다. 이러한 현대기술의 중요성이 대두되면서 교육과정에서 양자물리의 비중이 증가하게 되었다. 하지만 2009년 개정 교육과정의 경우, 많은 분량과 과도한 응용분야에 대한 내용 때문에 현장에서 많은 혼란을 야기하였다. 2015년 개정 교육과정은 앞선 문제점들을 해결하였으나, 양자물리에 대한 기본 개념을 이해할 수 있는 내용이 부족하였다. 교과서의 경우, 왜 양자물리가 생기게 되었는지에 대한 내용이 없다. 또한 흥미를 유발하는 도입내용이 없으며, 양자물리에 대한 기본적인 내용이 생략되었다. 본 연구를 통해, 앞선 문제점을 보완할 수 있는 양자물리 부교재와 수업지도안을 개발하였다. 학생들의 흥미를 유발하고 ‘양자물리’에 대한 정의와 전체적인 흐름을 알 수 있는 양자물리 부교재를 만들었다. 그리고 이를 바탕으로 수업지도안을 개발하였다. 또, 2015 개정 교육과정의 성격과 목표를 달성 할 수 있는 수업을 위해 부교재와 수업지도안의 개발 시 유의하였다. 학생들의 흥미를 유발하기 위한 도입 글을 삽입하였으며, 간단한 질문을 바탕으로 토론수업을 진행하여 과학과 핵심역량인 사고력과 의사소통능력을 키울 수 있도록 설계하였다. 또한 ‘양자물리’에 대한 정확한 정의와 ‘양자’의 정확한 뜻을 설명하여 학문에 대한 이해를 높인다. 양자물리의 대표적인 특징을 소개하여, 단원 도입 전에 기본적인 내용을 이해할 수 있도록 한다. 과학자와 양자물리의 역사를 설명하여 과학사의 내용을 담았다. 이후 응용분야인 양자점을 소개하고 ‘탐구활동’을 통해 그 원리와 현상을 이해하도록 한다. 이것을 통해 과학과 핵심역량인 탐구능력을 키울 수 있도록 하였다. 물리학Ⅱ의 경우, 아직 교육과정이 적용되지 않았으므로 실제 학교에서 학생들을 대상으로 수업을 실시하지 못하였다. 하지만 앞으로 도입될 2015년 개정 교육과정에서 본 연구에서 개발된 부교재와 수업지도안이 학생과 교사에게 많은 도움이 될 것으로 예상된다.

Studies of Higgs Boson Self-Interaction via Higgs Boson Pair Production in the ATLAS Experiment and the Development of Prior-Assisted Anomaly Detection Methods for New Physics Searches

Cheng, Chi Lung The University of Wisconsin - Madison ProQuest Dis 2025 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

Particle physics stands at the forefront of humanity's endeavor to understand our physical reality, studying the elementary particles and the forces that govern their interactions. This thesis presents contributions to this scientific effort, detailing experimental research conducted as part of the ATLAS Collaboration at the CERN Large Hadron Collider (LHC). It also introduces novel machine learning methods designed for model-agnostic searches for new physics. The work presented herein addresses fundamental questions at the heart of the Standard Model (SM) and explores avenues for discovering physics beyond it.The first part focuses on an investigation into the Higgs boson's self-interaction via Higgs boson pair production (HH) processes. Measuring the Higgs self-coupling is essential for experimentally reconstructing the Higgs potential, providing important insights into the mechanism of electroweak symmetry breaking and offering a window into physics beyond the Standard Model. This work presents a dedicated search in the HH to bb final state using the full 140fb-1 Run 2 dataset collected by the ATLAS experiment at √s=13 TeV. The analysis strategy was optimized for sensitivity to both the dominant gluon-gluon fusion and sub-dominant vector-boson fusion production modes. Furthermore, this thesis details the statistical combination of this search with four other ATLAS Run 2 HH analyses. At the time of publication, this combination yielded the most stringent constraints derived from the LHC Run 2 data: an observed (expected) 95% CL upper limit is set on the SM HH production signal strength modifier of µHH < 2.9 (2.4). The combination constrains the Higgs boson trilinear self-coupling modifier, κλ = λHHH/λHHH SM, to the interval [-1.2, 7.2] (expected [-1.6, 7.2]) as well as the quartic HHVV coupling modifier, κ2V, to [0.6, 1.5] (expected [0.4, 1.6]) at 95% CL. These results are consistent with the Standard Model predictions.The second part shifts focus to model-agnostic search strategies, addressing the challenge of discovering unexpected new physics phenomena. Novel machine learning techniques for anomaly detection are introduced. This work presents a method called Prior Assisted Weak Supervision (PAWS), developed to significantly enhance search sensitivity by incorporating physics knowledge to guide the learning process. This physics knowledge, serving as a prior, defines a restricted, physically motivated space of potential signal functions, which focuses the search when looking for potential anomalies in the data. On benchmark simulated dijet resonance datasets, PAWS demonstrates over an order-of-magnitude improvement in sensitivity compared to conventional weakly supervised approaches. This method is further extended into a complete framework for statistical inference, Generator Based Inference (GBI), enabling not only the detection of anomalies but also the direct, quantitative measurement of their physical properties (e.g., particle mass, signal fraction) with well-defined confidence intervals. Studies on simulated datasets show the GBI-PAWS framework can detect signals with significances as low as 0.1σ and provide accurate, unbiased parameter estimates. While these methods show significant promise, their application to experimental data remains future work.Together, these efforts contribute to the ongoing exploration of fundamental physics, providing experimental constraints on the Higgs boson's self-interaction while introducing effective anomaly detection methodologies for future discoveries.

Physics-Enhanced Neural Networks for Autonomous Driving: Design Methodologies and Applications

양진호 한양대학교 대학원 2026 국내박사

High Energy Physics from Low Energy Physics

Farrell, Roland Carlos ProQuest Dissertations & Theses University of Wash 2024 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

The separation between physics at low and high energies is essential for physics to have any utility; the details of quantum gravity are not necessary to calculate the trajectory of a cannon ball. However, physics at low and high energies are not completely independent, and this thesis explores two ways that they are related.The first is through a UV/IR symmetry that relates scattering processes at low and high energies. This UV/IR symmetry manifests in geometrical properties of the S-matrix, and of the RG flow of the coupling constants in the corresponding effective field theory. Low energy nuclear physics nearly realizes this UV/IR symmetry, providing an explanation for the smallness of shape parameters in the effective range expansion of nucleon-nucleon scattering, and inspiring a new way to organize the interactions between neutrons and protons.The second is through the use of quantum computers to simulate lattice gauge theories. Quantum simulations rely on the universality of the rules of quantum mechanics, which can be applied equally well to describe a (low energy) transmon qubit at 15 milli-Kelvin as a (high energy) 1 TeV quark. This thesis presents the first simulations of one dimensional lattice quantum chromodynamics on a quantum computer, culminating in a real-time simulation of beta-decay. Results from the first simulations of a lattice gauge theory on 100+ qubits of a quantum computer are also presented. The methods developed in this thesis for quantum simulation are "physics-aware", and are guided by the symmetries and hierarchies in length scales of the systems being studied. Without these physics-aware methods, 100+ qubit simulations of lattice gauge theories would not have been possible on the noisy quantum computers that are presently available.

A Physics Enhanced Residual Learning (PERL) Framework for Autonomous Vehicle

Long, Keke The University of Wisconsin - Madison ProQuest Dis 2024 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

Autonomous vehicles (AV) driving in mixed traffic, comprising human-driven vehicles (HVs) and AVs, is notoriously challenging, underscored by complex driving environments and a multitude of uncertainties. Further, the AV system involves complex interactions cascading across the interconnected modules for perception, planning, and control. An error in the higher chain of these interconnected modules can propagate downstream, inducing instability in vehicle control. Such errors can also cascade through following vehicles and impact traffic dynamics at a broader level. To overcome these challenges, we propose the novel Physics-Enhanced Residual Learning (PERL) framework for AV operations in mixed traffic. PERL comprises two components: a physics-based model and residual learning. The physics model provides primary outputs, including prediction outputs or control outputs, and then its residuals are learned by a learning-based approach as corrections to the physics model to enhance the results.This dissertation comprises three major research thrusts: (1) Development of PERL frameworks for vehicle trajectory prediction (2) Validate the contribution of PERL-based prediction in AV control, and (3) Development of Physics-Enhanced Residual Policy Learning (PERPL) framework for vehicle control.The first part proposed the PERL framework and applied it to a vehicle trajectory prediction problem with real-world trajectory data of both HV and AV, using an adapted Newell car-following model as the physics model, and four kinds of neural networks (GRU, Convolution Long Short-Term Memory (CLSTM), VAE and Informer model) as the residual learning model. We compare this PERL model with pure physics models, NN models, and other physics-informed neural network (PINN) models. The result reveals that the PERL model yields the best prediction with limited training data and it has fast convergence during training. Moreover, the PERL model requires fewer parameters to achieve similar predictive performance compared to NN and PINN models. The second part proposes a PERL-based vehicle control method to mitigate traffic oscillation in the mixed traffic environment of connected and autonomous vehicles (CAVs) and HVs. This model includes the PERL prediction model and a controller. The PERL-based prediction model precisely predicts the behavior of the preceding vehicle, especially downstream speed fluctuations, to allow sufficient time for the driver to respond to these speed fluctuations. For the controller, we employ a Model Predictive Control (MPC) model that considers the dynamics of the CAV and its following vehicles, improving safety and comfort for the platoon formed including the CAV and following vehicles. The proposed model is validated through a Vehicle-in-the-loop (ViL) field test. Results validate the proposed method in damping traffic oscillation and enhancing the safety and fuel efficiency of the CAV and the following vehicles in mixed traffic with the presence of uncertain vehicle dynamics and actuator lag. The third part proposes the Physics-Enhanced Residual Policy Learning (PERPL) framework for vehicle control, leveraging the advantages of both physics-based models (data-efficient and interpretable) and RL methods (flexible to multiple objectives and fast computing). The physics component provides model interpretability and stability and the learning-based Residual Policy adjusts the physics-based policy to adapt to the changing environment, thereby refining the decisions of the physics model. This model is applied in decentralized control of a mixed traffic platoon of CAVs and HVs using a constant time gap (CTG) strategy, with actuator lag and communication delays. Experimental results demonstrate that this model has high extrapolation ability, achieving smaller headway errors and better oscillation dampening than the linear control model and reinforcement learning (RL) model in artificial extreme scenarios. At the macroscopic level, overall traffic oscillations are also reduced as the penetration rate of CAVs employing the PERPL-based controller increases.