Many-core Algorithms for Combinatorial Optimization

Combinatorial Optimization is becoming ever more crucial, in these days. From natural sciences to economics, passing through urban centers administration and personnel management, methodologies and algorithms with a strong theoretical background and a consolidated real-word effectiveness is more and more requested, in order to find, quickly, good solutions to complex strategical problems. Resource optimization is, nowadays, a fundamental ground for building the basements of successful projects. From the theoretical point of view, Combinatorial Optimization rests on stable and strong foundations, that allow researchers to face ever more challenging problems. However, from the application point of view, it seems that the rate of theoretical developments cannot cope with that enjoyed by modern hardware technologies, especially with reference to the one of processors industry. In this work we propose new parallel algorithms, designed for exploiting the new parallel architectures available on the market. We found that, exposing the inherent parallelism of some resolution techniques (like Dynamic Programming), the computational benefits are remarkable, lowering the execution times by more than an order of magnitude, and allowing to address instances with dimensions not possible before. We approached four Combinatorial Optimization’s notable problems: Packing Problem, Vehicle Routing Problem, Single Source Shortest Path Problem and a Network Design problem. For each of these problems we propose a collection of effective parallel solution algorithms, either for solving the full problem (Guillotine Cuts and SSSPP) or for enhancing a fundamental part of the solution method (VRP and ND). We endorse our claim by presenting computational results for all problems, either on standard benchmarks from the literature or, when possible, on data from real-world applications, where speed-ups of one order of magnitude are usually attained, not uncommonly scaling up to 40 X factors.

Response threshold models, stochastic learning automata and ant colony optimization-based decentralized self-coordination algorithms for heterogeneous multi-tasks distribution in multi-robot systems

In recent decades, there has been an increasing interest in systems comprised of several autonomous mobile robots, and as a result, there has been a substantial amount of development in the eld of Articial Intelligence, especially in Robotics. There are several studies in the literature by some researchers from the scientic community that focus on the creation of intelligent machines and devices capable to imitate the functions and movements of living beings. Multi-Robot Systems (MRS) can often deal with tasks that are dicult, if not impossible, to be accomplished by a single robot. In the context of MRS, one of the main challenges is the need to control, coordinate and synchronize the operation of multiple robots to perform a specic task. This requires the development of new strategies and methods which allow us to obtain the desired system behavior in a formal and concise way. This PhD thesis aims to study the coordination of multi-robot systems, in particular, addresses the problem of the distribution of heterogeneous multi-tasks. The main interest in these systems is to understand how from simple rules inspired by the division of labor in social insects, a group of robots can perform tasks in an organized and coordinated way. We are mainly interested on truly distributed or decentralized solutions in which the robots themselves, autonomously and in an individual manner, select a particular task so that all tasks are optimally distributed. In general, to perform the multi-tasks distribution among a team of robots, they have to synchronize their actions and exchange information. Under this approach we can speak of multi-tasks selection instead of multi-tasks assignment, which means, that the agents or robots select the tasks instead of being assigned a task by a central controller. The key element in these algorithms is the estimation ix of the stimuli and the adaptive update of the thresholds. This means that each robot performs this estimate locally depending on the load or the number of pending tasks to be performed. In addition, it is very interesting the evaluation of the results in function in each approach, comparing the results obtained by the introducing noise in the number of pending loads, with the purpose of simulate the robot's error in estimating the real number of pending tasks. The main contribution of this thesis can be found in the approach based on self-organization and division of labor in social insects. An experimental scenario for the coordination problem among multiple robots, the robustness of the approaches and the generation of dynamic tasks have been presented and discussed. The particular issues studied are: Threshold models: It presents the experiments conducted to test the response threshold model with the objective to analyze the system performance index, for the problem of the distribution of heterogeneous multitasks in multi-robot systems; also has been introduced additive noise in the number of pending loads and has been generated dynamic tasks over time. Learning automata methods: It describes the experiments to test the learning automata-based probabilistic algorithms. The approach was tested to evaluate the system performance index with additive noise and with dynamic tasks generation for the same problem of the distribution of heterogeneous multi-tasks in multi-robot systems. Ant colony optimization: The goal of the experiments presented is to test the ant colony optimization-based deterministic algorithms, to achieve the distribution of heterogeneous multi-tasks in multi-robot systems. In the experiments performed, the system performance index is evaluated by introducing additive noise and dynamic tasks generation over time.

Column Generation Algorithms for Nonlinear Optimization II: Numerical Investigations

García et al. present a class of column generation (CG) algorithms for nonlinear programs. Its main motivation from a theoretical viewpoint is that under some circumstances, finite convergence can be achieved, in much the same way as for the classic simplicial decomposition method; the main practical motivation is that within the class there are certain nonlinear column generation problems that can accelerate the convergence of a solution approach which generates a sequence of feasible points. This algorithm can, for example, accelerate simplicial decomposition schemes by making the subproblems nonlinear. This paper complements the theoretical study on the asymptotic and finite convergence of these methods given in [1] with an experimental study focused on their computational efficiency. Three types of numerical experiments are conducted. The first group of test problems has been designed to study the parameters involved in these methods. The second group has been designed to investigate the role and the computation of the prolongation of the generated columns to the relative boundary. The last one has been designed to carry out a more complete investigation of the difference in computational efficiency between linear and nonlinear column generation approaches. In order to carry out this investigation, we consider two types of test problems: the first one is the nonlinear, capacitated single-commodity network flow problem of which several large-scale instances with varied degrees of nonlinearity and total capacity are constructed and investigated, and the second one is a combined traffic assignment model

Aerodynamic Optimization of the Nose Shape of a High-speed Train

La influencia de la aerodinámica en el diseño de los trenes de alta velocidad, unida a la necesidad de resolver nuevos problemas surgidos con el aumento de la velocidad de circulación y la reducción de peso del vehículo, hace evidente el interés de plantear un estudio de optimización que aborde tales puntos. En este contexto, se presenta en esta tesis la optimización aerodinámica del testero de un tren de alta velocidad, llevada a cabo mediante el uso de métodos de optimización avanzados. Entre estos métodos, se ha elegido aquí a los algoritmos genéticos y al método adjunto como las herramientas para llevar a cabo dicha optimización. La base conceptual, las características y la implementación de los mismos se detalla a lo largo de la tesis, permitiendo entender los motivos de su elección, y las consecuencias, en términos de ventajas y desventajas que cada uno de ellos implican. El uso de los algorimos genéticos implica a su vez la necesidad de una parametrización geométrica de los candidatos a óptimo y la generación de un modelo aproximado que complementa al método de optimización. Estos puntos se describen de modo particular en el primer bloque de la tesis, enfocada a la metodología seguida en este estudio. El segundo bloque se centra en la aplicación de los métodos a fin de optimizar el comportamiento aerodinámico del tren en distintos escenarios. Estos escenarios engloban los casos más comunes y también algunos de los más exigentes a los que hace frente un tren de alta velocidad: circulación en campo abierto con viento frontal o viento lateral, y entrada en túnel. Considerando el caso de viento frontal en campo abierto, los dos métodos han sido aplicados, permitiendo una comparación de las diferentes metodologías, así como el coste computacional asociado a cada uno, y la minimización de la resistencia aerodinámica conseguida en esa optimización. La posibilidad de evitar parametrizar la geometría y, por tanto, reducir el coste computacional del proceso de optimización es la característica más significativa de los métodos adjuntos, mientras que en el caso de los algoritmos genéticos se destaca la simplicidad y capacidad de encontrar un óptimo global en un espacio de diseño multi-modal o de resolver problemas multi-objetivo. El caso de viento lateral en campo abierto considera nuevamente los dos métoxi dos de optimización anteriores. La parametrización se ha simplificado en este estudio, lo que notablemente reduce el coste numérico de todo el estudio de optimización, a la vez que aún recoge las características geométricas más relevantes en un tren de alta velocidad. Este análisis ha permitido identificar y cuantificar la influencia de cada uno de los parámetros geométricos incluídos en la parametrización, y se ha observado que el diseño de la arista superior a barlovento es fundamental, siendo su influencia mayor que la longitud del testero o que la sección frontal del mismo. Finalmente, se ha considerado un escenario más a fin de validar estos métodos y su capacidad de encontrar un óptimo global. La entrada de un tren de alta velocidad en un túnel es uno de los casos más exigentes para un tren por el pico de sobrepresión generado, el cual afecta a la confortabilidad del pasajero, así como a la estabilidad del vehículo y al entorno próximo a la salida del túnel. Además de este problema, otro objetivo a minimizar es la resistencia aerodinámica, notablemente superior al caso de campo abierto. Este problema se resuelve usando algoritmos genéticos. Dicho método permite obtener un frente de Pareto donde se incluyen el conjunto de óptimos que minimizan ambos objetivos. ABSTRACT Aerodynamic design of trains influences several aspects of high-speed trains performance in a very significant level. In this situation, considering also that new aerodynamic problems have arisen due to the increase of the cruise speed and lightness of the vehicle, it is evident the necessity of proposing an optimization study concerning the train aerodynamics. Thus, the aerodynamic optimization of the nose shape of a high-speed train is presented in this thesis. This optimization is based on advanced optimization methods. Among these methods, genetic algorithms and the adjoint method have been selected. A theoretical description of their bases, the characteristics and the implementation of each method is detailed in this thesis. This introduction permits understanding the causes of their selection, and the advantages and drawbacks of their application. The genetic algorithms requirethe geometrical parameterization of any optimal candidate and the generation of a metamodel or surrogate model that complete the optimization process. These points are addressed with a special attention in the first block of the thesis, focused on the methodology considered in this study. The second block is referred to the use of these methods with the purpose of optimizing the aerodynamic performance of a high-speed train in several scenarios. These scenarios englobe the most representative operating conditions of high-speed trains, and also some of the most exigent train aerodynamic problems: front wind and cross-wind situations in open air, and the entrance of a high-speed train in a tunnel. The genetic algorithms and the adjoint method have been applied in the minimization of the aerodynamic drag on the train with front wind in open air. The comparison of these methods allows to evaluate the methdology and computational cost of each one, as well as the resulting minimization of the aerodynamic drag. Simplicity and robustness, the straightforward realization of a multi-objective optimization, and the capability of searching a global optimum are the main attributes of genetic algorithm. However, the requirement of geometrically parameterize any optimal candidate is a significant drawback that is avoided with the use of the adjoint method. This independence of the number of design variables leads to a relevant reduction of the pre-processing and computational cost. Considering the cross-wind stability, both methods are used again for the minimization of the side force. In this case, a simplification of the geometric parameterization of the train nose is adopted, what dramatically reduces the computational cost of the optimization process. Nevertheless, some of the most important geometrical characteristics are still described with this simplified parameterization. This analysis identifies and quantifies the influence of each design variable on the side force on the train. It is observed that the A-pillar roundness is the most demanding design parameter, with a more important effect than the nose length or the train cross-section area. Finally, a third scenario is considered for the validation of these methods in the aerodynamic optimization of a high-speed train. The entrance of a train in a tunnel is one of the most exigent train aerodynamic problems. The aerodynamic consequences of high-speed trains running in a tunnel are basically resumed in two correlated phenomena, the generation of pressure waves and an increase in aerodynamic drag. This multi-objective optimization problem is solved with genetic algorithms. The result is a Pareto front where a set of optimal solutions that minimize both objectives.

Distributed Estimation of Distribution Algorithms for continuous optimization: how does the exchanged information influence their behavior?

One of the most promising areas in which probabilistic graphical models have shown an incipient activity is the field of heuristic optimization and, in particular, in Estimation of Distribution Algorithms. Due to their inherent parallelism, different research lines have been studied trying to improve Estimation of Distribution Algorithms from the point of view of execution time and/or accuracy. Among these proposals, we focus on the so-called distributed or island-based models. This approach defines several islands (algorithms instances) running independently and exchanging information with a given frequency. The information sent by the islands can be either a set of individuals or a probabilistic model. This paper presents a comparative study for a distributed univariate Estimation of Distribution Algorithm and a multivariate version, paying special attention to the comparison of two alternative methods for exchanging information, over a wide set of parameters and problems ? the standard benchmark developed for the IEEE Workshop on Evolutionary Algorithms and other Metaheuristics for Continuous Optimization Problems of the ISDA 2009 Conference. Several analyses from different points of view have been conducted to analyze both the influence of the parameters and the relationships between them including a characterization of the configurations according to their behavior on the proposed benchmark.

Aerodynamic Optimization of a High-speed Train using Genetic Algorithms

Genetic algorithms (GA) have been used for the minimization of the aerodynamic drag of a train subject to front wind. The significant importance of the external aerodynamic drag on the total resistance a train experiments as the cruise speed is increased highlights the interest of this study. A complete description of the methodology required for this optimization method is introduced here, where the parameterization of the geometry to be optimized and the metamodel used to speed up the optimization process are detailed. A reduction of about a 25% of the initial aerodynamic drag is obtained in this study, what confirms GA as a proper method for this optimization problem. The evolution of the nose shape is consistent with the literature. The advantage of using metamodels is stressed thanks to the information of the whole design space extracted from it. The influence of each design variable on the objective function is analyzed by means of an ANOVA test.

Approximation algorithms

Increasing global competition, rapidly changing markets, and greater consumer awareness have altered the way in which corporations do business. To become more efficient, many industries have sought to model some operational aspects by gigantic optimization problems. It is not atypical to encounter models that capture 106 separate “yes” or “no” decisions to be made. Although one could, in principle, try all 2106 possible solutions to find the optimal one, such a method would be impractically slow. Unfortunately, for most of these models, no algorithms are known that find optimal solutions with reasonable computation times. Typically, industry must rely on solutions of unguaranteed quality that are constructed in an ad hoc manner. Fortunately, for some of these models there are good approximation algorithms: algorithms that produce solutions quickly that are provably close to optimal. Over the past 6 years, there has been a sequence of major breakthroughs in our understanding of the design of approximation algorithms and of limits to obtaining such performance guarantees; this area has been one of the most flourishing areas of discrete mathematics and theoretical computer science.

One Approach for the Optimization of Estimates Calculating Algorithms

In this article the new approach for optimization of estimations calculating algorithms is suggested. It can be used for finding the correct algorithm of minimal complexity in the context of algebraic approach for pattern recognition.

Optimization of adaptive traffic signal control with transit signal priority at isolated intersections using parallel genetic algorithms

Optimization of adaptive traffic signal timing is one of the most complex problems in traffic control systems. This dissertation presents a new method that applies the parallel genetic algorithm (PGA) to optimize adaptive traffic signal control in the presence of transit signal priority (TSP). The method can optimize the phase plan, cycle length, and green splits at isolated intersections with consideration for the performance of both the transit and the general vehicles. Unlike the simple genetic algorithm (GA), PGA can provide better and faster solutions needed for real-time optimization of adaptive traffic signal control. ^ An important component in the proposed method involves the development of a microscopic delay estimation model that was designed specifically to optimize adaptive traffic signal with TSP. Macroscopic delay models such as the Highway Capacity Manual (HCM) delay model are unable to accurately consider the effect of phase combination and phase sequence in delay calculations. In addition, because the number of phases and the phase sequence of adaptive traffic signal may vary from cycle to cycle, the phase splits cannot be optimized when the phase sequence is also a decision variable. A "flex-phase" concept was introduced in the proposed microscopic delay estimation model to overcome these limitations. ^ The performance of PGA was first evaluated against the simple GA. The results show that PGA achieved both faster convergence and lower delay for both under- or over-saturated traffic conditions. A VISSIM simulation testbed was then developed to evaluate the performance of the proposed PGA-based adaptive traffic signal control with TSP. The simulation results show that the PGA-based optimizer for adaptive TSP outperformed the fully actuated NEMA control in all test cases. The results also show that the PGA-based optimizer was able to produce TSP timing plans that benefit the transit vehicles while minimizing the impact of TSP on the general vehicles. The VISSIM testbed developed in this research provides a powerful tool to design and evaluate different TSP strategies under both actuated and adaptive signal control. ^

Efficient algorithms to solve scheduling problems with a variety of optimization criteria

La programmation par contraintes est une technique puissante pour résoudre, entre autres, des problèmes d’ordonnancement de grande envergure. L’ordonnancement vise à allouer dans le temps des tâches à des ressources. Lors de son exécution, une tâche consomme une ressource à un taux constant. Généralement, on cherche à optimiser une fonction objectif telle la durée totale d’un ordonnancement. Résoudre un problème d’ordonnancement signifie trouver quand chaque tâche doit débuter et quelle ressource doit l’exécuter. La plupart des problèmes d’ordonnancement sont NP-Difficiles. Conséquemment, il n’existe aucun algorithme connu capable de les résoudre en temps polynomial. Cependant, il existe des spécialisations aux problèmes d’ordonnancement qui ne sont pas NP-Complet. Ces problèmes peuvent être résolus en temps polynomial en utilisant des algorithmes qui leur sont propres. Notre objectif est d’explorer ces algorithmes d’ordonnancement dans plusieurs contextes variés. Les techniques de filtrage ont beaucoup évolué dans les dernières années en ordonnancement basé sur les contraintes. La proéminence des algorithmes de filtrage repose sur leur habilité à réduire l’arbre de recherche en excluant les valeurs des domaines qui ne participent pas à des solutions au problème. Nous proposons des améliorations et présentons des algorithmes de filtrage plus efficaces pour résoudre des problèmes classiques d’ordonnancement. De plus, nous présentons des adaptations de techniques de filtrage pour le cas où les tâches peuvent être retardées. Nous considérons aussi différentes propriétés de problèmes industriels et résolvons plus efficacement des problèmes où le critère d’optimisation n’est pas nécessairement le moment où la dernière tâche se termine. Par exemple, nous présentons des algorithmes à temps polynomial pour le cas où la quantité de ressources fluctue dans le temps, ou quand le coût d’exécuter une tâche au temps t dépend de t.

Combination of Evolutionary Algorithms with Experimental Design, Traditional Optimization and Machine Learning

Evolutionary algorithms alone cannot solve optimization problems very efficiently since there are many random (not very rational) decisions in these algorithms. Combination of evolutionary algorithms and other techniques have been proven to be an efficient optimization methodology. In this talk, I will explain the basic ideas of our three algorithms along this line (1): Orthogonal genetic algorithm which treats crossover/mutation as an experimental design problem, (2) Multiobjective evolutionary algorithm based on decomposition (MOEA/D) which uses decomposition techniques from traditional mathematical programming in multiobjective optimization evolutionary algorithm, and (3) Regular model based multiobjective estimation of distribution algorithms (RM-MEDA) which uses the regular property and machine learning methods for improving multiobjective evolutionary algorithms.

I. Quantal effects in biochemical cooperativity and a proposed mechanism for the differentiation of calcium signaling in synaptic plasticity. II. Evolutionary algorithms for the optimization of methods in computational chemistry

In Part 1 of this thesis, we propose that biochemical cooperativity is a fundamentally non-ideal process. We show quantal effects underlying biochemical cooperativity and highlight apparent ergodic breaking at small volumes. The apparent ergodic breaking manifests itself in a divergence of deterministic and stochastic models. We further predict that this divergence of deterministic and stochastic results is a failure of the deterministic methods rather than an issue of stochastic simulations.

Ergodic breaking at small volumes may allow these molecular complexes to function as switches to a greater degree than has previously been shown. We propose that this ergodic breaking is a phenomenon that the synapse might exploit to differentiate Ca$^{2+}$ signaling that would lead to either the strengthening or weakening of a synapse. Techniques such as lattice-based statistics and rule-based modeling are tools that allow us to directly confront this non-ideality. A natural next step to understanding the chemical physics that underlies these processes is to consider \textit{in silico} specifically atomistic simulation methods that might augment our modeling efforts.

In the second part of this thesis, we use evolutionary algorithms to optimize \textit{in silico} methods that might be used to describe biochemical processes at the subcellular and molecular levels. While we have applied evolutionary algorithms to several methods, this thesis will focus on the optimization of charge equilibration methods. Accurate charges are essential to understanding the electrostatic interactions that are involved in ligand binding, as frequently discussed in the first part of this thesis.

Optimization of Pattern Matching Algorithms for Multi- and Many-Core Platforms

Image and video compression play a major role in the world today, allowing the storage and transmission of large multimedia content volumes. However, the processing of this information requires high computational resources, hence the improvement of the computational performance of these compression algorithms is very important. The Multidimensional Multiscale Parser (MMP) is a pattern-matching-based compression algorithm for multimedia contents, namely images, achieving high compression ratios, maintaining good image quality, Rodrigues et al. [2008]. However, in comparison with other existing algorithms, this algorithm takes some time to execute. Therefore, two parallel implementations for GPUs were proposed by Ribeiro [2016] and Silva [2015] in CUDA and OpenCL-GPU, respectively. In this dissertation, to complement the referred work, we propose two parallel versions that run the MMP algorithm in CPU: one resorting to OpenMP and another that converts the existing OpenCL-GPU into OpenCL-CPU. The proposed solutions are able to improve the computational performance of MMP by 3 and 2:7 , respectively. The High Efficiency Video Coding (HEVC/H.265) is the most recent standard for compression of image and video. Its impressive compression performance, makes it a target for many adaptations, particularly for holoscopic image/video processing (or light field). Some of the proposed modifications to encode this new multimedia content are based on geometry-based disparity compensations (SS), developed by Conti et al. [2014], and a Geometric Transformations (GT) module, proposed by Monteiro et al. [2015]. These compression algorithms for holoscopic images based on HEVC present an implementation of specific search for similar micro-images that is more efficient than the one performed by HEVC, but its implementation is considerably slower than HEVC. In order to enable better execution times, we choose to use the OpenCL API as the GPU enabling language in order to increase the module performance. With its most costly setting, we are able to reduce the GT module execution time from 6.9 days to less then 4 hours, effectively attaining a speedup of 45 .

Modelling and Optimization of Energy Management Strategies for Hybrid Vehicles

In the last decades the automotive sector has seen a technological revolution, due mainly to the more restrictive regulation, the newly introduced technologies and, as last, to the poor resources of fossil fuels remaining on Earth. Promising solution in vehicles’ propulsion are represented by alternative architectures and energy sources, for example fuel-cells and pure electric vehicles. The automotive transition to new and green vehicles is passing through the development of hybrid vehicles, that usually combine positive aspects of each technology. To fully exploit the powerful of hybrid vehicles, however, it is important to manage the powertrain’s degrees of freedom in the smartest way possible, otherwise hybridization would be worthless. To this aim, this dissertation is focused on the development of energy management strategies and predictive control functions. Such algorithms have the goal of increasing the powertrain overall efficiency and contextually increasing the driver safety. Such control algorithms have been applied to an axle-split Plug-in Hybrid Electric Vehicle with a complex architecture that allows more than one driving modes, including the pure electric one. The different energy management strategies investigated are mainly three: the vehicle baseline heuristic controller, in the following mentioned as rule-based controller, a sub-optimal controller that can include also predictive functionalities, referred to as Equivalent Consumption Minimization Strategy, and a vehicle global optimum control technique, called Dynamic Programming, also including the high-voltage battery thermal management. During this project, different modelling approaches have been applied to the powertrain, including Hardware-in-the-loop, and diverse powertrain high-level controllers have been developed and implemented, increasing at each step their complexity. It has been proven the potential of using sophisticated powertrain control techniques, and that the gainable benefits in terms of fuel economy are largely influenced by the chose energy management strategy, even considering the powerful vehicle investigated.

Models and algorithms for real-world optimization problems

This thesis deals with efficient solution of optimization problems of practical interest. The first part of the thesis deals with bin packing problems. The bin packing problem (BPP) is one of the oldest and most fundamental combinatorial optimiza- tion problems. The bin packing problem and its generalizations arise often in real-world ap- plications, from manufacturing industry, logistics and transportation of goods, and scheduling. After an introductory chapter, I will present two applications of two of the most natural extensions of the bin packing: Chapter 2 will be dedicated to an application of bin packing in two dimension to a problem of scheduling a set of computational tasks on a computer cluster, while Chapter 3 deals with the generalization of BPP in three dimensions that arise frequently in logistic and transportation, often com- plemented with additional constraints on the placement of items and characteristics of the solution, like, for example, guarantees on the stability of the items, to avoid potential damage to the transported goods, on the distribution of the total weight of the bins, and on compatibility with loading and unloading operations. The second part of the thesis, and in particular Chapter 4 considers the Trans- mission Expansion Problem (TEP), where an electrical transmission grid must be expanded so as to satisfy future energy demand at the minimum cost, while main- taining some guarantees of robustness to potential line failures. These problems are gaining importance in a world where a shift towards renewable energy can impose a significant geographical reallocation of generation capacities, resulting in the ne- cessity of expanding current power transmission grids.