Learning for scheduling a portfolio of constraint solvers

Nel lavoro di tesi qui presentato si indaga l'applicazione di tecniche di apprendimento mirate ad una più efficiente esecuzione di un portfolio di risolutore di vincoli (constraint solver). Un constraint solver è un programma che dato in input un problema di vincoli, elabora una soluzione mediante l'utilizzo di svariate tecniche. I problemi di vincoli sono altamente presenti nella vita reale. Esempi come l'organizzazione dei viaggi dei treni oppure la programmazione degli equipaggi di una compagnia aerea, sono tutti problemi di vincoli. Un problema di vincoli è formalizzato da un problema di soddisfacimento di vincoli(CSP). Un CSP è descritto da un insieme di variabili che possono assumere valori appartenenti ad uno specico dominio ed un insieme di vincoli che mettono in relazione variabili e valori assumibili da esse. Una tecnica per ottimizzare la risoluzione di tali problemi è quella suggerita da un approccio a portfolio. Tale tecnica, usata anche in am- biti come quelli economici, prevede la combinazione di più solver i quali assieme possono generare risultati migliori di un approccio a singolo solver. In questo lavoro ci preoccupiamo di creare una nuova tecnica che combina un portfolio di constraint solver con tecniche di machine learning. Il machine learning è un campo di intelligenza articiale che si pone l'obiettivo di immettere nelle macchine una sorta di `intelligenza'. Un esempio applicativo potrebbe essere quello di valutare i casi passati di un problema ed usarli in futuro per fare scelte. Tale processo è riscontrato anche a livello cognitivo umano. Nello specico, vogliamo ragionare in termini di classicazione. Una classicazione corrisponde ad assegnare ad un insieme di caratteristiche in input, un valore discreto in output, come vero o falso se una mail è classicata come spam o meno. La fase di apprendimento sarà svolta utilizzando una parte di CPHydra, un portfolio di constraint solver sviluppato presso la University College of Cork (UCC). Di tale algoritmo a portfolio verranno utilizzate solamente le caratteristiche usate per descrivere determinati aspetti di un CSP rispetto ad un altro; queste caratteristiche vengono altresì dette features. Creeremo quindi una serie di classicatori basati sullo specifico comportamento dei solver. La combinazione di tali classicatori con l'approccio a portfolio sara nalizzata allo scopo di valutare che le feature di CPHydra siano buone e che i classicatori basati su tali feature siano affidabili. Per giusticare il primo risultato, eettueremo un confronto con uno dei migliori portfolio allo stato dell'arte, SATzilla. Una volta stabilita la bontà delle features utilizzate per le classicazioni, andremo a risolvere i problemi simulando uno scheduler. Tali simulazioni testeranno diverse regole costruite con classicatori precedentemente introdotti. Prima agiremo su uno scenario ad un processore e successivamente ci espanderemo ad uno scenario multi processore. In questi esperimenti andremo a vericare che, le prestazioni ottenute tramite l'applicazione delle regole create appositamente sui classicatori, abbiano risultati migliori rispetto ad un'esecuzione limitata all'utilizzo del migliore solver del portfolio. I lavoro di tesi è stato svolto in collaborazione con il centro di ricerca 4C presso University College Cork. Su questo lavoro è stato elaborato e sottomesso un articolo scientico alla International Joint Conference of Articial Intelligence (IJCAI) 2011. Al momento della consegna della tesi non siamo ancora stati informati dell'accettazione di tale articolo. Comunque, le risposte dei revisori hanno indicato che tale metodo presentato risulta interessante.

A study on set variable representations in constraint programming

Il lavoro presentato in questa tesi si colloca nel contesto della programmazione con vincoli, un paradigma per modellare e risolvere problemi di ricerca combinatoria che richiedono di trovare soluzioni in presenza di vincoli. Una vasta parte di questi problemi trova naturale formulazione attraverso il linguaggio delle variabili insiemistiche. Dal momento che il dominio di tali variabili può essere esponenziale nel numero di elementi, una rappresentazione esplicita è spesso non praticabile. Recenti studi si sono quindi focalizzati nel trovare modi efficienti per rappresentare tali variabili. Pertanto si è soliti rappresentare questi domini mediante l'uso di approssimazioni definite tramite intervalli (d'ora in poi rappresentazioni), specificati da un limite inferiore e un limite superiore secondo un'appropriata relazione d'ordine. La recente evoluzione della ricerca sulla programmazione con vincoli sugli insiemi ha chiaramente indicato che la combinazione di diverse rappresentazioni permette di raggiungere prestazioni di ordini di grandezza superiori rispetto alle tradizionali tecniche di codifica. Numerose proposte sono state fatte volgendosi in questa direzione. Questi lavori si differenziano su come è mantenuta la coerenza tra le diverse rappresentazioni e su come i vincoli vengono propagati al fine di ridurre lo spazio di ricerca. Sfortunatamente non esiste alcun strumento formale per paragonare queste combinazioni. Il principale obiettivo di questo lavoro è quello di fornire tale strumento, nel quale definiamo precisamente la nozione di combinazione di rappresentazioni facendo emergere gli aspetti comuni che hanno caratterizzato i lavori precedenti. In particolare identifichiamo due tipi possibili di combinazioni, una forte ed una debole, definendo le nozioni di coerenza agli estremi sui vincoli e sincronizzazione tra rappresentazioni. Il nostro studio propone alcune interessanti intuizioni sulle combinazioni esistenti, evidenziandone i limiti e svelando alcune sorprese. Inoltre forniamo un'analisi di complessità della sincronizzazione tra minlex, una rappresentazione in grado di propagare in maniera ottimale vincoli lessicografici, e le principali rappresentazioni esistenti.

Modellazione multibody di veicoli ferroviari: sviluppo ed implementazione di modelli innovativi per l'analisi del contatto ruota - rotaia

The wheel - rail contact analysis plays a fundamental role in the multibody modeling of railway vehicles. A good contact model must provide an accurate description of the global contact phenomena (contact forces and torques, number and position of the contact points) and of the local contact phenomena (position and shape of the contact patch, stresses and displacements). The model has also to assure high numerical efficiency (in order to be implemented directly online within multibody models) and a good compatibility with commercial multibody software (Simpack Rail, Adams Rail). The wheel - rail contact problem has been discussed by several authors and many models can be found in the literature. The contact models can be subdivided into two different categories: the global models and the local (or differential) models. Currently, as regards the global models, the main approaches to the problem are the so - called rigid contact formulation and the semi – elastic contact description. The rigid approach considers the wheel and the rail as rigid bodies. The contact is imposed by means of constraint equations and the contact points are detected during the dynamic simulation by solving the nonlinear algebraic differential equations associated to the constrained multibody system. Indentation between the bodies is not permitted and the normal contact forces are calculated through the Lagrange multipliers. Finally the Hertz’s and the Kalker’s theories allow to evaluate the shape of the contact patch and the tangential forces respectively. Also the semi - elastic approach considers the wheel and the rail as rigid bodies. However in this case no kinematic constraints are imposed and the indentation between the bodies is permitted. The contact points are detected by means of approximated procedures (based on look - up tables and simplifying hypotheses on the problem geometry). The normal contact forces are calculated as a function of the indentation while, as in the rigid approach, the Hertz’s and the Kalker’s theories allow to evaluate the shape of the contact patch and the tangential forces. Both the described multibody approaches are computationally very efficient but their generality and accuracy turn out to be often insufficient because the physical hypotheses behind these theories are too restrictive and, in many circumstances, unverified. In order to obtain a complete description of the contact phenomena, local (or differential) contact models are needed. In other words wheel and rail have to be considered elastic bodies governed by the Navier’s equations and the contact has to be described by suitable analytical contact conditions. The contact between elastic bodies has been widely studied in literature both in the general case and in the rolling case. Many procedures based on variational inequalities, FEM techniques and convex optimization have been developed. This kind of approach assures high generality and accuracy but still needs very large computational costs and memory consumption. Due to the high computational load and memory consumption, referring to the current state of the art, the integration between multibody and differential modeling is almost absent in literature especially in the railway field. However this integration is very important because only the differential modeling allows an accurate analysis of the contact problem (in terms of contact forces and torques, position and shape of the contact patch, stresses and displacements) while the multibody modeling is the standard in the study of the railway dynamics. In this thesis some innovative wheel – rail contact models developed during the Ph. D. activity will be described. Concerning the global models, two new models belonging to the semi – elastic approach will be presented; the models satisfy the following specifics: 1) the models have to be 3D and to consider all the six relative degrees of freedom between wheel and rail 2) the models have to consider generic railway tracks and generic wheel and rail profiles 3) the models have to assure a general and accurate handling of the multiple contact without simplifying hypotheses on the problem geometry; in particular the models have to evaluate the number and the position of the contact points and, for each point, the contact forces and torques 4) the models have to be implementable directly online within the multibody models without look - up tables 5) the models have to assure computation times comparable with those of commercial multibody software (Simpack Rail, Adams Rail) and compatible with RT and HIL applications 6) the models have to be compatible with commercial multibody software (Simpack Rail, Adams Rail). The most innovative aspect of the new global contact models regards the detection of the contact points. In particular both the models aim to reduce the algebraic problem dimension by means of suitable analytical techniques. This kind of reduction allows to obtain an high numerical efficiency that makes possible the online implementation of the new procedure and the achievement of performance comparable with those of commercial multibody software. At the same time the analytical approach assures high accuracy and generality. Concerning the local (or differential) contact models, one new model satisfying the following specifics will be presented: 1) the model has to be 3D and to consider all the six relative degrees of freedom between wheel and rail 2) the model has to consider generic railway tracks and generic wheel and rail profiles 3) the model has to assure a general and accurate handling of the multiple contact without simplifying hypotheses on the problem geometry; in particular the model has to able to calculate both the global contact variables (contact forces and torques) and the local contact variables (position and shape of the contact patch, stresses and displacements) 4) the model has to be implementable directly online within the multibody models 5) the model has to assure high numerical efficiency and a reduced memory consumption in order to achieve a good integration between multibody and differential modeling (the base for the local contact models) 6) the model has to be compatible with commercial multibody software (Simpack Rail, Adams Rail). In this case the most innovative aspects of the new local contact model regard the contact modeling (by means of suitable analytical conditions) and the implementation of the numerical algorithms needed to solve the discrete problem arising from the discretization of the original continuum problem. Moreover, during the development of the local model, the achievement of a good compromise between accuracy and efficiency turned out to be very important to obtain a good integration between multibody and differential modeling. At this point the contact models has been inserted within a 3D multibody model of a railway vehicle to obtain a complete model of the wagon. The railway vehicle chosen as benchmark is the Manchester Wagon the physical and geometrical characteristics of which are easily available in the literature. The model of the whole railway vehicle (multibody model and contact model) has been implemented in the Matlab/Simulink environment. The multibody model has been implemented in SimMechanics, a Matlab toolbox specifically designed for multibody dynamics, while, as regards the contact models, the CS – functions have been used; this particular Matlab architecture allows to efficiently connect the Matlab/Simulink and the C/C++ environment. The 3D multibody model of the same vehicle (this time equipped with a standard contact model based on the semi - elastic approach) has been then implemented also in Simpack Rail, a commercial multibody software for railway vehicles widely tested and validated. Finally numerical simulations of the vehicle dynamics have been carried out on many different railway tracks with the aim of evaluating the performances of the whole model. The comparison between the results obtained by the Matlab/ Simulink model and those obtained by the Simpack Rail model has allowed an accurate and reliable validation of the new contact models. In conclusion to this brief introduction to my Ph. D. thesis, we would like to thank Trenitalia and the Regione Toscana for the support provided during all the Ph. D. activity. Moreover we would also like to thank the INTEC GmbH, the society the develops the software Simpack Rail, with which we are currently working together to develop innovative toolboxes specifically designed for the wheel rail contact analysis.

Hybrid constraint programming and metaheuristic methods for large scale optimization problems

This work presents hybrid Constraint Programming (CP) and metaheuristic methods for the solution of Large Scale Optimization Problems; it aims at integrating concepts and mechanisms from the metaheuristic methods to a CP-based tree search environment in order to exploit the advantages of both approaches. The modeling and solution of large scale combinatorial optimization problem is a topic which has arisen the interest of many researcherers in the Operations Research field; combinatorial optimization problems are widely spread in everyday life and the need of solving difficult problems is more and more urgent. Metaheuristic techniques have been developed in the last decades to effectively handle the approximate solution of combinatorial optimization problems; we will examine metaheuristics in detail, focusing on the common aspects of different techniques. Each metaheuristic approach possesses its own peculiarities in designing and guiding the solution process; our work aims at recognizing components which can be extracted from metaheuristic methods and re-used in different contexts. In particular we focus on the possibility of porting metaheuristic elements to constraint programming based environments, as constraint programming is able to deal with feasibility issues of optimization problems in a very effective manner. Moreover, CP offers a general paradigm which allows to easily model any type of problem and solve it with a problem-independent framework, differently from local search and metaheuristic methods which are highly problem specific. In this work we describe the implementation of the Local Branching framework, originally developed for Mixed Integer Programming, in a CP-based environment. Constraint programming specific features are used to ease the search process, still mantaining an absolute generality of the approach. We also propose a search strategy called Sliced Neighborhood Search, SNS, that iteratively explores slices of large neighborhoods of an incumbent solution by performing CP-based tree search and encloses concepts from metaheuristic techniques. SNS can be used as a stand alone search strategy, but it can alternatively be embedded in existing strategies as intensification and diversification mechanism. In particular we show its integration within the CP-based local branching. We provide an extensive experimental evaluation of the proposed approaches on instances of the Asymmetric Traveling Salesman Problem and of the Asymmetric Traveling Salesman Problem with Time Windows. The proposed approaches achieve good results on practical size problem, thus demonstrating the benefit of integrating metaheuristic concepts in CP-based frameworks.

Constraint based methods for allocation and scheduling of periodic applications

This work presents exact algorithms for the Resource Allocation and Cyclic Scheduling Problems (RA&CSPs). Cyclic Scheduling Problems arise in a number of application areas, such as in hoist scheduling, mass production, compiler design (implementing scheduling loops on parallel architectures), software pipelining, and in embedded system design. The RA&CS problem concerns time and resource assignment to a set of activities, to be indefinitely repeated, subject to precedence and resource capacity constraints. In this work we present two constraint programming frameworks facing two different types of cyclic problems. In first instance, we consider the disjunctive RA&CSP, where the allocation problem considers unary resources. Instances are described through the Synchronous Data-flow (SDF) Model of Computation. The key problem of finding a maximum-throughput allocation and scheduling of Synchronous Data-Flow graphs onto a multi-core architecture is NP-hard and has been traditionally solved by means of heuristic (incomplete) algorithms. We propose an exact (complete) algorithm for the computation of a maximum-throughput mapping of applications specified as SDFG onto multi-core architectures. Results show that the approach can handle realistic instances in terms of size and complexity. Next, we tackle the Cyclic Resource-Constrained Scheduling Problem (i.e. CRCSP). We propose a Constraint Programming approach based on modular arithmetic: in particular, we introduce a modular precedence constraint and a global cumulative constraint along with their filtering algorithms. Many traditional approaches to cyclic scheduling operate by fixing the period value and then solving a linear problem in a generate-and-test fashion. Conversely, our technique is based on a non-linear model and tackles the problem as a whole: the period value is inferred from the scheduling decisions. The proposed approaches have been tested on a number of non-trivial synthetic instances and on a set of realistic industrial instances achieving good results on practical size problem.

Portfolio Approaches in Constraint Programming

Recent research has shown that the performance of a single, arbitrarily efficient algorithm can be significantly outperformed by using a portfolio of —possibly on-average slower— algorithms. Within the Constraint Programming (CP) context, a portfolio solver can be seen as a particular constraint solver that exploits the synergy between the constituent solvers of its portfolio for predicting which is (or which are) the best solver(s) to run for solving a new, unseen instance. In this thesis we examine the benefits of portfolio solvers in CP. Despite portfolio approaches have been extensively studied for Boolean Satisfiability (SAT) problems, in the more general CP field these techniques have been only marginally studied and used. We conducted this work through the investigation, the analysis and the construction of several portfolio approaches for solving both satisfaction and optimization problems. We focused in particular on sequential approaches, i.e., single-threaded portfolio solvers always running on the same core. We started from a first empirical evaluation on portfolio approaches for solving Constraint Satisfaction Problems (CSPs), and then we improved on it by introducing new data, solvers, features, algorithms, and tools. Afterwards, we addressed the more general Constraint Optimization Problems (COPs) by implementing and testing a number of models for dealing with COP portfolio solvers. Finally, we have come full circle by developing sunny-cp: a sequential CP portfolio solver that turned out to be competitive also in the MiniZinc Challenge, the reference competition for CP solvers.

Longitudinal dynamics modeling and handling qualities of a conventional jet aircraft

Un noto centro di ricerca europea ha recentemente modificato un jet convenzionale di classe CS-25 in una piattaforma scientifica. Durante il processo di certificazione delle modifiche, l’impatto delle stesse sulle prestazioni è stato studiato in modo esaustivo. Per lo studio delle qualità di volo, i piloti collaudatori hanno sviluppato una procedura di certificazione ad hoc che consiste in test qualitativi separati della stabilità longitudinale, laterale e direzionale. L’obiettivo della tesi è analizzare i dati di volo, registrati durante i test di collaudo, con l'obiettivo di estrarre informazioni di carattere quantitativo circa la stabilità longitudinale del velivolo modificato. In primo luogo sono state analizzate tre diverse modifiche apportate all’aeromobile e successivamente i risultati sono stati messi a confronto per capirne l’influenza sulle qualità di volo dell’aeromobile. Le derivate aerodinamiche sono state stimate utilizzando la cosiddetta “identificazione dei parametri”, che mira a replicare le variabili registrate durante i test di volo, variando un dato insieme di coefficienti all’interno del modello linearizzato della dinamica dell’aeromobile. L'identificazione del modo di corto periodo ha consentito l'estrazione dei suoi parametri caratteristici, quali il rapporto di smorzamento e la frequenza naturale. La procedura ha consentito inoltre di calcolare il cosiddetto “Control Anticipation Parameter” (CAP), parametro caratterizzante delle qualità di volo di un aeroplano. I risultati ottenuti sono stati messi a confronto con i requisiti prescritti dalla normativa MIL-STD-1797-A, risultando conformi al livello più alto di qualità di volo.

A Morphological Box for Handling Temporal Data in B2C Systems

User interfaces are key properties of Business-to-Consumer (B2C) systems, and Web-based reservation systems are an important class of B2C systems. In this paper we show that these systems use a surprisingly broad spectrum of different approaches to handling temporal data in their Web inter faces. Based on these observations and on a literature analysis we develop a Morphological Box to present the main options for handling temporal data and give examples. The results indicate that the present state of developing and maintaining B2C systems has not been much influenced by modern Web Engi neering concepts and that there is considerable potential for improvement.

Remodeling of calcium handling in skeletal muscle through PGC-1?: impact on force, fatigability, and fiber type

Regular endurance exercise remodels skeletal muscle, largely through the peroxisome proliferator-activated receptor-γ coactivator-1α (PGC-1α). PGC-1α promotes fiber type switching and resistance to fatigue. Intracellular calcium levels might play a role in both adaptive phenomena, yet a role for PGC-1α in the adaptation of calcium handling in skeletal muscle remains unknown. Using mice with transgenic overexpression of PGC-1α, we now investigated the effect of PGC-1α on calcium handling in skeletal muscle. We demonstrate that PGC-1α induces a quantitative reduction in calcium release from the sarcoplasmic reticulum by diminishing the expression of calcium-releasing molecules. Concomitantly, maximal muscle force is reduced in vivo and ex vivo. In addition, PGC-1α overexpression delays calcium clearance from the myoplasm by interfering with multiple mechanisms involved in calcium removal, leading to higher myoplasmic calcium levels following contraction. During prolonged muscle activity, the delayed calcium clearance might facilitate force production in mice overexpressing PGC-1α. Our results reveal a novel role of PGC-1α in altering the contractile properties of skeletal muscle by modulating calcium handling. Importantly, our findings indicate PGC-1α to be both down- as well as upstream of calcium signaling in this tissue. Overall, our findings suggest that in the adaptation to chronic exercise, PGC-1α reduces maximal force, increases resistance to fatigue, and drives fiber type switching partly through remodeling of calcium transients, in addition to promoting slow-type myofibrillar protein expression and adequate energy supply.

Development of a model-based transient calibration process for diesel engine electronic control module tables – Part 2: modelling and optimization

This is the second part of a study investigating a model-based transient calibration process for diesel engines. The first part addressed the data requirements and data processing required for empirical transient emission and torque models. The current work focuses on modelling and optimization. The unexpected result of this investigation is that when trained on transient data, simple regression models perform better than more powerful methods such as neural networks or localized regression. This result has been attributed to extrapolation over data that have estimated rather than measured transient air-handling parameters. The challenges of detecting and preventing extrapolation using statistical methods that work well with steady-state data have been explained. The concept of constraining the distribution of statistical leverage relative to the distribution of the starting solution to prevent extrapolation during the optimization process has been proposed and demonstrated. Separate from the issue of extrapolation is preventing the search from being quasi-static. Second-order linear dynamic constraint models have been proposed to prevent the search from returning solutions that are feasible if each point were run at steady state, but which are unrealistic in a transient sense. Dynamic constraint models translate commanded parameters to actually achieved parameters that then feed into the transient emission and torque models. Combined model inaccuracies have been used to adjust the optimized solutions. To frame the optimization problem within reasonable dimensionality, the coefficients of commanded surfaces that approximate engine tables are adjusted during search iterations, each of which involves simulating the entire transient cycle. The resulting strategy, different from the corresponding manual calibration strategy and resulting in lower emissions and efficiency, is intended to improve rather than replace the manual calibration process.

Model Predictive Control for Vehicle Stabilization at the Limits of Handling

Recent developments in vehicle steering systems offer new opportunities to measure the steering torque and reliably estimate the vehicle sideslip and the tire-road friction coefficient. This paper presents an approach to vehicle stabilization that leverages these estimates to define state boundaries that exclude unstable vehicle dynamics and utilizes a model predictive envelope controller to bound the vehicle motion within this stable region of the state space. This approach provides a large operating region accessible by the driver and smooth interventions at the stability boundaries. Experimental results obtained with a steer-by-wire vehicle and a proof of envelope invariance demonstrate the efficacy of the envelope controller in controlling the vehicle at the limits of handling.

The influence of handling and exposure to a ferret on body temperature and running wheel activity of golden hamsters (Mesocricetus auratus)

In order to determine a stress response, two groups of twenty male golden hamsters were either exposed to a ferret or handled by a human. The hamsters' body temperature and running wheel activity were measured as stress correlates. Half of the hamsters' cages were equipped with a functional running wheel to determine whether the presence of a running wheel might reduce stress. Exposure to the ferret was followed by a significant increase in body temperature and running wheel revolutions: however, running wheel activity did not change after handling. Body temperature increased less after handling in hamsters living in a cage with a functional running wheel than in those with a non-revolving running wheel. This suggests that hamsters with a functional running wheel reacted less strongly to acute stress caused by handling. On the other hand, temperature increase after the exposure to a ferret was not affected by the presence of a running wheel. Both exposure to a ferret and handling caused stress in golden hamsters, as demonstrated by an increase in body temperature (emotional fever). Stress caused by handling was much milder than stress caused by the ferret. (C) 2011 Elsevier B.V. All rights reserved.