This work focuses on the robustness issues related to distributed model predictive control (DMPC) strategies in the presence of model uncertainty. The robustness of DMPC with respect to model uncertainty has been identified by researchers as a key factor in the successful application of DMPC. A first task towards the formulation of robust DMPC strategy was to propose a new systematic methodology for the selection of a control structure in the context of DMPC. The methodology is based on the trade-off between performance and simplicity of structure (e.g., a centralized versus decentralized structure) and is formulated as a multi-objective mixed-integer nonlinear program (MINLP). The multi-objective function is composed of the contribution of two indices: 1) closed-loop performance index computed as an upper bound on the variability of the closed-loop system due to the effect on the output error of either set-point or disturbance input, and 2) a connectivity index used as a measure of the simplicity of the control structure. The parametric uncertainty in the models of the process is also considered in the methodology and it is described by a polytopic representation whereby the actual process’s states are assumed to evolve within a polytope whose vertices are defined by linear models that can be obtained from either linearizing a nonlinear model or from their identification in the neighborhood of different operating conditions. The system’s closed-loop performance and stability are formulated as Linear Matrix Inequalities (LMI) problems so that efficient interior-point methods can be exploited. To solve the MINLP a multi-start approach is adopted in which many starting points are generated in an attempt to obtain global optima. The efficiency of the proposed methodology is shown through its application to benchmark simulation examples. The simulation results are consistent with the conclusions obtained from the analysis. The proposed methodology can be applied at the design stage to select the best control configuration in the presence of model errors. A second goal accomplished in this research was the development of a novel online algorithm for robust DMPC that explicitly accounts for parametric uncertainty in the model. This algorithm requires the decomposition of the entire system’s model into N subsystems and the solution of N convex corresponding optimization problems in parallel. The objective of this parallel optimizations is to minimize an upper bound on a robust performance objective by using a time-varying state-feedback controller for each subsystem. Model uncertainty is explicitly considered through the use of polytopic description of the model. The algorithm employs an LMI approach, in which the solutions are convex and obtained in polynomial time. An observer is designed and embedded within each controller to perform state estimations and the stability of the observer integrated with the controller is tested online via LMI conditions. An iterative design method is also proposed for computing the observer gain. This algorithm has many practical advantages, the first of which is the fact that it can be implemented in real-time control applications and thus has the benefit of enabling the use of a decentralized structure while maintaining overall stability and improving the performance of the system. It has been shown that the proposed algorithm can achieve the theoretical performance of centralized control. Furthermore, the proposed algorithm can be formulated using a variety of objectives, such as Nash equilibrium, involving interacting processing units with local objective functions or fully decentralized control in the case of communication failure. Such cases are commonly encountered in the process industry. Simulations examples are considered to illustrate the application of the proposed method. Finally, a third goal was the formulation of a new algorithm to improve the online computational efficiency of DMPC algorithms. The closed-loop dual-mode paradigm was employed in order to perform most of the heavy computations offline using convex optimization to enlarge invariant sets thus rendering the iterative online solution more efficient. The solution requires the satisfaction of only relatively simple constraints and the solution of problems each involving a small number of decision variables. The algorithm requires solving N convex LMI problems in parallel when cooperative scheme is implemented. The option of using Nash scheme formulation is also available for this algorithm. A relaxation method was incorporated with the algorithm to satisfy initial feasibility by introducing slack variables that converge to zero quickly after a small number of early iterations. Simulation case studies have illustrated the applicability of this approach and have demonstrated that significant improvement can be achieved with respect to computation times. Extensions of the current work in the future should address issues of communication loss, delays and actuator failure and their impact on the robustness of DMPC algorithms. In addition, integration of the proposed DMPC algorithms with other layers in automation hierarchy can be an interesting topic for future work.

This work focuses on the robustness issues related to distributed model predictive control (DMPC) strategies in the presence of model uncertainty. The robustness of DMPC with respect to model uncertainty has been identified by researchers as a key factor in the successful application of DMPC. A first task towards the formulation of robust DMPC strategy was to propose a new systematic methodology for the selection of a control structure in the context of DMPC. The methodology is based on the trade-off between performance and simplicity of structure (e.g., a centralized versus decentralized structure) and is formulated as a multi-objective mixed-integer nonlinear program (MINLP). The multi-objective function is composed of the contribution of two indices: 1) closed-loop performance index computed as an upper bound on the variability of the closed-loop system due to the effect on the output error of either set-point or disturbance input, and 2) a connectivity index used as a measure of the simplicity of the control structure. The parametric uncertainty in the models of the process is also considered in the methodology and it is described by a polytopic representation whereby the actual process’s states are assumed to evolve within a polytope whose vertices are defined by linear models that can be obtained from either linearizing a nonlinear model or from their identification in the neighborhood of different operating conditions. The system’s closed-loop performance and stability are formulated as Linear Matrix Inequalities (LMI) problems so that efficient interior-point methods can be exploited. To solve the MINLP a multi-start approach is adopted in which many starting points are generated in an attempt to obtain global optima. The efficiency of the proposed methodology is shown through its application to benchmark simulation examples. The simulation results are consistent with the conclusions obtained from the analysis. The proposed methodology can be applied at the design stage to select the best control configuration in the presence of model errors. A second goal accomplished in this research was the development of a novel online algorithm for robust DMPC that explicitly accounts for parametric uncertainty in the model. This algorithm requires the decomposition of the entire system’s model into N subsystems and the solution of N convex corresponding optimization problems in parallel. The objective of this parallel optimizations is to minimize an upper bound on a robust performance objective by using a time-varying state-feedback controller for each subsystem. Model uncertainty is explicitly considered through the use of polytopic description of the model. The algorithm employs an LMI approach, in which the solutions are convex and obtained in polynomial time. An observer is designed and embedded within each controller to perform state estimations and the stability of the observer integrated with the controller is tested online via LMI conditions. An iterative design method is also proposed for computing the observer gain. This algorithm has many practical advantages, the first of which is the fact that it can be implemented in real-time control applications and thus has the benefit of enabling the use of a decentralized structure while maintaining overall stability and improving the performance of the system. It has been shown that the proposed algorithm can achieve the theoretical performance of centralized control. Furthermore, the proposed algorithm can be formulated using a variety of objectives, such as Nash equilibrium, involving interacting processing units with local objective functions or fully decentralized control in the case of communication failure. Such cases are commonly encountered in the process industry. Simulations examples are considered to illustrate the application of the proposed method. Finally, a third goal was the formulation of a new algorithm to improve the online computational efficiency of DMPC algorithms. The closed-loop dual-mode paradigm was employed in order to perform most of the heavy computations offline using convex optimization to enlarge invariant sets thus rendering the iterative online solution more efficient. The solution requires the satisfaction of only relatively simple constraints and the solution of problems each involving a small number of decision variables. The algorithm requires solving N convex LMI problems in parallel when cooperative scheme is implemented. The option of using Nash scheme formulation is also available for this algorithm. A relaxation method was incorporated with the algorithm to satisfy initial feasibility by introducing slack variables that converge to zero quickly after a small number of early iterations. Simulation case studies have illustrated the applicability of this approach and have demonstrated that significant improvement can be achieved with respect to computation times. Extensions of the current work in the future should address issues of communication loss, delays and actuator failure and their impact on the robustness of DMPC algorithms. In addition, integration of the proposed DMPC algorithms with other layers in automation hierarchy can be an interesting topic for future work.

An increasing number of applications ranging from multi-vehicle systems, large-scale process control systems, transportation systems to smart grids call for the development of cooperative control theory. Meanwhile, when designing the cooperative controller, the state and control constraints, ubiquitously existing in the physical system, have to be respected. Model predictive control (MPC) is one of a few techniques that can explicitly and systematically handle the state and control constraints. This dissertation studies the robust MPC and distributed MPC strategies, respectively. Specifically, the problems we investigate are: the robust MPC for linear or nonlinear systems, distributed MPC for constrained decoupled systems and distributed MPC for constrained nonlinear systems with coupled system dynamics. In the robust MPC controller design, three sub-problems are considered. Firstly, a computationally efficient multi-stage suboptimal MPC strategy is designed by exploiting the j-step admissible sets, where the j-step admissible set is the set of system states that can be steered to the maximum positively invariant set in j control steps. Secondly, for nonlinear systems with control constraints and external disturbances, a novel robust constrained MPC strategy is designed, where the cost function is in a non-squared form. Sufficient conditions for the recursive feasibility and robust stability are established, respectively. Finally, by exploiting the contracting dynamics of a certain type of nonlinear systems, a less conservative robust constrained MPC method is designed. Compared to robust MPC strategies based on Lipschitz continuity, the strategy employed has the following advantages: 1) it can tolerate larger disturbances; and 2) it is feasible for a larger prediction horizon and enlarges the feasible region accordingly. For the distributed MPC of constrained continuous-time nonlinear decoupled systems, the cooperation among each subsystems is realized by incorporating a coupling term in the cost function. To handle the effect of the disturbances, a robust control strategy is designed based on the two-layer invariant set. Provided that the initial state is feasible and the disturbance is bounded by a certain level, the recursive feasibility of the optimization is guaranteed by appropriately tuning the design parameters. Sufficient conditions are given ensuring that the states of each subsystem converge to the robust positively invariant set. Furthermore, a conceptually less conservative algorithm is proposed by exploiting the controllability set instead of the positively invariant set, which allows the adoption of a shorter prediction horizon and tolerates a larger disturbance level. For the distributed MPC of a large-scale system that consists of several dynamically coupled nonlinear systems with decoupled control constraints and disturbances, the dynamic couplings and the disturbances are accommodated through imposing new robustness constraints in the local optimizations. Relationships among, and design procedures for the parameters involved in the proposed distributed MPC are derived to guarantee the recursive feasibility and the robust stability of the overall system. It is shown that, for a given bound on the disturbances, the recursive feasibility is guaranteed if the sampling interval is properly chosen. / Graduate / 0548 / 0544 / 0546 / liuxiaotao1982@gmail.com

No description available.

No description available.

学位授与大学：京都大学 ; 取得学位: 博士(工学) ; 学位授与年月日: 2007-09-25 ; 学位の種類: 新制・課程博士 ; 学位記番号: 工博第2849号 ; 請求記号: 新制/工/1419 ; 整理番号: 25534 / Understanding river water quantity and quality variation is one of the fundamental requirements for the integrated watershed management. Monitoring is usually preferred to examine and understand the river water quantity and quality, especially focusing on pre-specified objectives. Although monitoring is invaluable in many instances, it is of less use to forecast the foreseeable changes, especially, for the long-term prediction that is usually required by the decision-makers. Therefore, for the decision-making, modeling is widely practiced. Due to the limited understanding of hydrological processes inside a watershed, models often fail to estimate properly, which in worst case could often mislead the targeted plans. Among several aspects, spatial variability such as land cover, topography, soil, geology is believed to affect the overall performance of the model. Such thought lead to the concept of distributed models that were supposed to represent spatial variability through modeling specific variations inside the watershed by using several representative units or grids. In that meaning, distributed models required to identify and assign the values of its parameters to represent the physical processes defined by the governing equations for each grid. Due to the unavailability of required spatial information at appropriate grid sizes, even physically based and conceptually sound distributed models fail to estimate properly thereby offsetting the credibility of distributed models. Therefore, in this study, we set a major objective to develop a new distributed water quantity and water quality model to address some of the stated issues. Major emphasis was given to conceptually sound but simple structure of the model. In addition to that, model aimed to utilize the potential of recent advances in spatial information, such as remote sensing and GIS, to generate and process the spatial data, and to determine the values of its essential parameters. The approach was expected to provide an example that the complexity of the model should be preferred only if the defined processes could be ascertained within some reasonable limit. At the initial stage, several spatial data were collected from different sources and they were processed into raster format, which was one of the essential requirements for the distributed model. Analysis of spatial database indicated that the watershed was characterized by forested parts in the hills, and densely populated urban areas in plains. Rainfall occurred quite frequently but they were of short duration. Besides constructing spatial database, several water quantity and quality surveys were also conducted at different spatial and temporal conditions from 2000 to 2006. The data were mainly used to understand variation patterns of water quantity and quality at both spatial and temporal conditions. Later on, some of the data were also used for the verification of model in study area. 28 water quality indices (WQIs) were observed for each observation, which were mainly utilized to understand the overall variation pattern of river water quality. Initial analysis of flow rate condition of the river showed that the rainfall-runoff responses were quite rapid after the rainfall but such effect appear for very short duration (< 2 days). Then, analysis of variance (ANOVA) and two multivariate analysis techniques (MVA), namely, principle component analysis (PCA) and cluster analysis (CA) were used to explore effectively the river water quality datasets. Analysis showed that the observed covariation among majority of WQIs could be due to the inter-linkages among rainfall pattern, atmospheric deposition of acidic ions, soil and geology of dominant forest areas, topography, and climatic conditions. The identified pattern indicated that there could be close relationship between the biogeochemical processes in the forest areas with both river water quantity and quality variation. A new distributed water quantity and quality model was developed especially focusing on the biophysical characteristics of the watershed. Basic structure of the model was similar to the concept of lumped tank model, which was often credited for its simple and sound conceptual structure. Two storey tanks were conceptualized for each grid, but model also took into consideration of drainage channels in urban areas and natural river channels as rapidly conveying structures. Besides, the model considered all major aspects affecting the estimation of water quantity, such as interception of the rainfall, evapotranspiration loss, surface runoff, sub-surface runoff, and ground water runoff. Compared with the original tank model, major emphasis was given to assign the values major parameters, such as coefficients and storage heights of the outlets, by relating them with the hilly topography of the study area and the variation in land cover, soil, and geology. The model was further integrated with water quality component, which was based on two fundamental assumptions of build-up and wash-off of the WQIs in the environment. Build-up was based on the land cover type and population, while wash off was based on the estimated runoff volume. Remote sensing and GIS techniques were used to assist in the modeling process. At first, remote sensing was mainly focused in the classification of land cover by utilizing seasonal Landsat ETM+ images. In addition to urban and vegetated urban categories, four major forest categories (shaded, deciduous, mixed, and evergreen) were identified. Then leaf area index (Lai) was determined for each vegetation category. Lai was mainly used to determine the rainfall interception by the canopy in the forest areas. In this study, forest areas showed the capacity to intercept as high as 1.2 mm of rainfall, which could be quite important during smaller rainfall events. Remote sensing was further used to determine the transpiration coefficient of the vegetations, which was a major requirement for the estimation of evapotranspiration (Et) loss by the FAO Penman- Monteith method used in the model simulation. Et was estimated even reached more than 4 mm/d in summer months, but it was relatively lower (< 2 mm/d) in the winter months. These facts suggested that consideration of both interception and Et loss in a forested watershed could have significant influence on the estimation of flow rates by the model. At the final stage, model was applied in the study area. Mainly three approaches were considered to assess the estimation by the model. First was conventional approach in which comparison between the observed and estimated data were done considering different spatial and temporal contexts. Assigned values of the parameters gave satisfactory prediction for both water quantity and quality for the selected grid size of 50 m in which the relative error was usually less than 1. The second approach evaluated the model by considering different scale of the grids ranging from 100m to 500m. It was observed that grid resizing usually affected the basin attributed such as slope, outlet height, drainage characteristics following nearly proportionate pattern than other categorical variables such as land cover or geology. Usually same parameter values gave very different prediction level for both magnitude and shape of the hydrographs (or pollutographs), in which increasing grid size was accompanied by the increasing peak event estimation or overall error. The effects were further assessed by changing the values of key parameters for each grid size targeting the minimum differences between the observed and estimated values. Interestingly, the parameters also showed some identifiable (increasing or decreasing) trend with the change in grid size. Particularly, due to the direct effect of predicted runoff on the reference WQIs, its showed more complex variation pattern at different grid sizes. Overall assessment of the distributed model indicated that the model was quite sensitive to the selection of key parameters for different grid sizes. It indicated that the values of calibrated parameters might not give stable result if the scale of input data were changed. It could further indicate that the choice of grid size should be assessed before the actual application of the model considering the spatial variability of the watershed. In the third approach, model was utilized to estimate at different scenarios, namely, rainfall variation and land cover changes. The differences in the estimated results could indicate that the model could be available for the watershed management at different runoff and land cover scenarios in future. / Kyoto University (京都大学) / 0048 / 新制・課程博士 / 博士(工学) / 甲第13378号 / 工博第2849号 / 新制||工||1419(附属図書館) / 25534 / UT51-2007-Q779 / 京都大学大学院工学研究科都市環境工学専攻 / (主査)教授 田中 宏明, 教授 藤井 滋穂, 教授 清水 芳久 / 学位規則第4条第1項該当

Une vallée hydroélectrique est constituée d'un nombre important de centrales interconnectées du fait de l'utilisation de la même ressource en eau. Pour pouvoir optimiser en temps réel sa production, il a été proposé dans cette thèse d'utiliser les méthodes associées aux systèmes à grande échelle pour développer les outils nécessaires. Cette étude de la commande globale du système a été orientée vers l'utilisation des méthodes de décomposition-coordination. Ces méthodes ont été examinées et appliquées à un cas d'étude simplifié (une partie de la vallée hydraulique) mis à disposition par EDF. Plus particulièrement, les méthodes de décomposition-coordination par les prix, ou encore par les prédictions des interactions, s'appuyant sur des commandes MPC, ont été considérées et comparées avec une commande centralisée. En vue d'une implémentation temps-réel, nous nous sommes intéressés à exprimer les problèmes d'optimisation comme des problèmes QP, pour ensuite obtenir des solutions explicites en utilisant une méthodologie de caractérisation géométrique. Nous avons proposé des formulations complètement explicites (niveau coordinateur et sous-systèmes) pour les deux méthodes. Des résultats de simulation avec des données réelles mises à disposition par EDF sont présentés. Afin de valider les méthodes conçues, une première phase d'implantation sur la plate-forme Supervision NG d'EDF permettant la communication avec un modèle de la vallée hydroélectrique (basé sur les équations de Saint Venant et la bathymétrie de la rivière), est enfin incluse dans ce mémoire. / This study is mainly about the hydroelectric production problem. What we aim to do, is to develop optimization tools for a chain of hydroelectric plants, using appropriate control methodologies. A hydroelectric valley is a large scale system, made up of interconnected plants. The study of the global control system has been focused to the use of decomposition-coordination methods. Those methods have been examined and applied to a simplified case study (a part of a hydroelectric valley) given by EDF. To be more specific, the price decomposition - coordination method and the interactions prediction method, based on MPC controls, have been considered and compared to a centralized control. Because of the need of implementation in real time, we have expressed the optimization problems as QP problems, so as to obtain explicit solutions using the geometric characterization methodology. We have proposed a completely explicit formulation (both at the coordinator level and at the subsystems level) for both methods. Simulation results with real data information given by EDF are also presented. To verify and validate the designed methods, a first step of implementation on the supervision platform NG by EDF, that allows the communication with a model of the hydroelectric valley (based on the equations of Saint Venant and on the river bathymetry) is finally also included in this thesis.

À l’heure actuelle, les stratégies de gestion de l’énergie pour les bâtiments sontprincipalement basées sur une concaténation de règles logiques. Bien que cette approcheoffre des avantages certains, particulièrement lors de sa mise en oeuvre sur des automatesprogrammables, elle peine à traiter la diversité de situations complexes quipeuvent être rencontrées (prix de l’énergie variable, limitations de puissance, capacitéde stockage d’énergie, bâtiments de grandes dimension).Cette thèse porte sur le développement et l’évaluation d’une commande prédictivepour la gestion de l’énergie dans le bâtiment ainsi que l’étude de l’embarcabilité del’algorithme de contrôle sur une cible temps-réel (Roombox - Schneider-Electric).La commande prédictive est basée sur l’utilisation d’un modèle du bâtiment ainsique des prévisions météorologiques et d’occupation afin de déterminer la séquencede commande optimale à mettre en oeuvre sur un horizon de prédiction glissant.Seul le premier élément de cette séquence est en réalité appliqué au bâtiment. Cetteséquence de commande optimale est obtenue par la résolution en ligne d’un problèmed’optimisation. La capacité de la commande prédictive à gérer des systèmes multivariablescontraints ainsi que des objectifs économiques, la rend particulièrementadaptée à la problématique de la gestion de l’énergie dans le bâtiment.Cette thèse propose l’élaboration d’un schéma de commande distribué pour contrôlerles conditions climatiques dans chaque zone du bâtiment. L’objectif est de contrôlersimultanément: la température intérieure, le taux de CO2 ainsi que le niveaud’éclairement dans chaque zone en agissant sur les équipements présents (CVC, éclairage,volets roulants). Par ailleurs, le cas des bâtiments multi-sources (par exemple:réseau électrique + production locale solaire), dans lequel chaque source d’énergie estcaractérisée par son propre prix et une limitation de puissance, est pris en compte.Dans ce contexte, les décisions relatives à chaque zone ne peuvent plus être effectuéesde façon indépendante. Pour résoudre ce problème, un mécanisme de coordinationbasé sur une décomposition du problème d’optimisation centralisé est proposé. Cettethèse CIFRE 1 a été préparée au sein du laboratoire Gipsa-lab en partenariat avecSchneider-Electric dans le cadre du programme HOMES (www.homesprogramme.com). / Currently, energy management strategies for buildings are mostly based on a concatenationof logical rules. Despite the fact that such rule based strategy can be easilyimplemented, it suffers from some limitations particularly when dealing with complexsituations. This thesis is concerned with the development and assessment ofModel Predictive Control (MPC) algorithms for energy management in buildings. Inthis work, a study of implementability of the control algorithm on a real-time hardwaretarget is conducted beside yearly simulations showing a substantial energy savingpotential. The thesis explores also the ability of MPC to deal with the diversity ofcomplex situations that could be encountered (varying energy price, power limitations,local storage capability, large scale buildings).MPC is based on the use of a model of the building as well as weather forecasts andoccupany predictions in order to find the optimal control sequence to be implementedin the future. Only the first element of the sequence is actually applied to the building.The best control sequence is found by solving, at each decision instant, an on lineoptimization problem. MPC’s ability to handle constrained multivariable systems aswell as economic objectives makes this paradigm particularly well suited for the issueof energy management in buildings.This thesis proposes the design of a distributed predictive control scheme to controlthe indoor conditions in each zone of the building. The goal is to control thefollowing simultaneously in each zone of the building: indoor temperature, indoorCO2 level and indoor illuminance by acting on all the actuators of the zone (HVAC,lighting, shading). Moreover, the case of multi-source buildings is also explored, (e.g.power from grid + local solar production), in which each power source is characterizedby its own dynamic tariff and upper limit. In this context, zone decisions can nolonger be performed independently. To tackle this issue, a coordination mechanismis proposed. A particular attention is paid to computational effectiveness of the proposedalgorithms. This CIFRE2 Ph.D. thesis was prepared within the Gipsa-lab laboratoryin partnership with Schneider-Electric in the scope of the HOMES program(www.homesprogramme.com).

Recent advancements in the development of high-efficiency multijunction solar cells have led to a renewed interest in the design and implementation of high concentration photovoltaic systems. With the emergence of novel materials and design structures, understanding the operation of multijunction solar cells has become a challenging task. Modeling and simulation hence play an important role in the analysis of such devices. In this dissertation, techniques for accurate optoelectrical modeling of concentrating photovoltaic systems, based on multijunction solar cells, are proposed. A 2-dimensional, distributed circuit model is proposed, parametrized to values obtained by numerical modeling of three multijunction cell designs, namely: a three-junction, lattice matched design, a three-junction lattice-mismatched, inverted metamorphic design, and a four-junction,lattice matched design. Cell performance for all the three designs is evaluated under both uniform and nonuniform illumination profiles at high concentrations and efficiency enhancement by optimizing finger spacing is proposed. The effect of luminescent coupling from higher bandgap subcells is also determined.Fresnel-lens based, refractive concentrating optical systems are modeled and optimized using an optical ray-tracing simulator at two different concentrations, with and without a secondary optical element. The corresponding optical efficiency, acceptance angle, and the degree of nonuniformity are determined for each optical system. An integrated approach,combining optical design with electrical modeling is proposed for optimizing the multijunction solar cell in tandem with the corresponding concentrating optics. The approach is validated by on-sun, acceptance angle measurements, using a three-junction,lattice-matched cell. Also, temperature effects are modeled and are experimentally validated for a three-junction, lattice-matched design. Experimental results with a single-junction, dilute-nitride cell, targeted for four-junction operation, are presented as well. A modified distributed circuit model is used for analysis of temperature effects in a four-junction solar cell, and the results under both uniform and nonuniform temperature profiles are presented. When implemented, the designs and their corresponding analyses, may result in new insights into the development of CPV systems, thereby enabling enhanced efficiencies at higher concentrations.

京都大学 / 新制・課程博士 / 博士(工学) / 甲第23479号 / 工博第4891号 / 新制||工||1764(附属図書館) / 京都大学大学院工学研究科社会基盤工学専攻 / (主査)教授 立川 康人, 准教授 市川 温, 教授 角 哲也 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM