Self-adaptive and self-managing systems optimize their own behaviour according to high-level objectives and constraints. One way for human administrators to effectively specify goals for such optimization problems is using policies. Over the past decade, researchers produced various approaches, models and techniques for policy specification in different areas including distributed systems, communication networks, web services, autonomic computing, and cloud computing. Research challenges range from characterizing policies for ease of specification in particular application domains to categorizing policies for achieving good solution qualities for particular algorithmic techniques. The contributions of this thesis are threefold. Firstly, we give a mathematical formulation for each of the three policy types, action, goal and utility function policies, introduced in the policy framework by Kephart and Walsh. In particular, we introduce a first precise characterization of goal policies for optimization problems. Secondly, this thesis introduces a mathematical framework that adds structure to the underlying optimization problem for different types of policies. Structure is added either to the objective function or the constraints of the optimization problem. These mathematical structures, imposed on the underlying problem, progressively increase the quality of the solutions obtained when using the greedy optimization technique. Thirdly, we show the applicability of our framework through case studies by analyzing several optimization problems encountered in self-adaptive and self-managing systems, such as resource allocation, quality of service management, and Service Level Agreement (SLA) profit optimization to provide quality guarantees for their solutions. Our approach combines the algorithmic results by Edmonds, Fisher et al., and Mestre, and the policy framework of Kephart and Walsh. Our characterization and approach will help designers of self-adaptive and self-managing systems formulate optimization problems, decide on algorithmic strategies based on policy requirements, and reason about solution qualities. / Graduate / 0984

Magister Scientiae - MSc / Essentially this project report is a discussion of mathematical modelling in pension funds, presenting sections from Cairns, A.J.D., Blake, D., Dowd, K., Stochastic lifestyling: Optimal dynamic asset allocation for defined contribution pension plans, Journal of Economic Dynamics and Control, Volume 30, Issue 2006, Pages 843-877, with added details and background material in order to demonstrate the mathematical methods. In the investigation of the management of the investment portfolio, we only use one risky asset together with a bond and cash as other assets in a continuous time framework. The particular model is very much designed according to the members’ preference and then the funds are invested by the fund manager in the financial market. At the end, we are going to show various simulations of these models. Our methods include stochastic control for utility maximisation among others. The optimisation problem entails the optimal investment portfolio to maximise a certain power utility function. We use MATLAB and MAPLE programming languages to generate results in the form of graphs and tables. / South Africa

In einem wichtigen Beitrag „on self-interest and greed“ hat Gebhard Kirchgässner im Journal of Business Economics die zu beobachtende Verhaltensvielfalt von Wirtschaftssubjekten hinterfragt. Die ökonomische Standardannahme rationalen, egoistischen Verhaltens trifft die Realität nicht immer. Menschen handeln auch irrational, teils sozialorientiert und teils sozial desinteressiert, teils unmoralisch-gierig, teils aber auch altruistisch-großzügig. Welches Modell erklärt die Vielfalt? In diesem Beitrag wird die von Kirchgässner aufgezeigte Verhaltensvielfalt mit den Erkenntnissen abgeglichen, die sich aus der Hamiltonschen Fitnessthese und den Ergänzungen durch Trivers, Fiske und de Botton ergeben. Damit kann gezeigt werden, dass sich je nach Gruppe, in welcher sich ein Mensch bewegt, Verhaltensweisen einstellen, die von extrem prosozialem Verhalten (z.B. zur Statussicherung durch Vorbild oder aus Angst vor Sanktionen) bis zu sehr unsozial „gierigem“ Verhalten reichen. Der Nutzen dieser Sichtweise für die Ökonomik wird aufgezeigt. Es sind nur drei exogene Mechanismen nötig, die Verhaltensvielfalt zu erzeugen. / In an important article „on self-interest and greed” in the Journal of Business Economics, Gebhard Kirchgässner questioned the variety observed in the behavior of economic subjects. The economic standard assumption of rational, egoist behavior does not comply with reality in many cases. People act irrationally, socially oriented as well as socially uninterested, immorally-greedy as well as altruistically-generously. The variety is immense. The question occurs: which model is capable of explaining the diversity? In this article, the variety of human behavior is being compared to findings that emerge from Hamilton’s Fitness Thesis and extensions by Trivers, Fiske and de Botton. The Fitness Thesis lays the ground. Depending on the social group to which an individual belongs in a certain situation, behavior can vary from extremely prosocial (e. g. to maintain status or to avoid sanctions) to very unsocial-greedy. As any individual is part of many groups he or she can act prosocially with respect to one group and unsocially-greedy with respect to another, will say: greed and altruism is not a question of character but a consequence of the specific situation in which an individual acts. Merely three exogenous mechanisms are required to create the observed variety in behavior.

In this thesis, we conduct an experimental simulation of 131 students from a university in Hong Kong and investigate the relationship between pay changes and the perceived values (i.e., utility). Applying traditional psychophysical methods, we measure the utility of pay changes (i.e., pay raises and pay cuts) of different sizes by individual responses (i.e., happiness/unhappiness). Drawing on utility theory and expectancy theory, we examine the function that best fits this relationship by considering common function forms including linear, quadratic, logarithmic, and power functions. Using regression techniques, we find that a quadratic function best fits the data, and the utility function is concave in the pay change. When we examine the best form of utility functions for pay raises and pay cuts separately, we find that the utility of pay raises and that of pay cuts are best described by a quadratic function and a linear function, respectively. We further show that a single model involving all pay changes better describes the utility than two separate models for pay raises and pay cuts. In addition, our best-fit utility model reveals that a sufficiently small amount of pay increase may generate a negative value of utility, and we calculate the percentage of smallest meaningful pay increase that results in non-negative utility. We also discuss the theoretical contributions of our findings to the literature and their implications to practitioners.

Background Manufacturers have to adopt modern technologies to compete at the top of their field. However, adopting new technologies can be expensive and difficult to validate prior to implementation. One technology that has difficulties receiving a wider application is bin picking. Bin picking uses vision technology to communicate with an industrial robot. Consequently, the technology enables robots to pick randomly sorted objects. Research finds that the difficulty in assessing its performance can explain the lacklustre application of bin picking. In addition, research on bin picking is primarily focused on its technical difficulties and neglects information that can be valuable for potential adopters.  Objectives This thesis aims to aid decision-makers in assessing the implications and opportunities of bin picking. Furthermore, the thesis desire to inspire potential adopters by analysing the viability of bin picking through feasibility and the tangible and intangible benefits in a real-world setting. Methods A utility function is developed and assigned categories based on interviews with suppliers and adopters of the technology and literature review. The utility function highlights the feasibility and the intangible benefits of a bin picking solution and enables ranking among alternatives. The highest scoring article is used to conduct a feasibility study in collaboration with suppliers of bin picking technology. Based on the feasibility studies, a DES is created to highlight the implications that may arise in a real manufacturing environment. Finally, financial calculations through NPV, IRR, PP and DPP are created to evaluate the solution. Results All NPV calculations (excluding a 12.5 \% discount rate) are positive with enough years. The IRR is positive when the time span exceeds 11 years. The PP is 10-11 years while the DPP is 12-13, 14-15, 19-20, 32-33 years at a discount rate of 2.5 \%, 5 \%, 7.5 \% respectively 10 \%. However, the investment is never recoupable at a discount rate of 12.5 \%. The categories of the utility function have a clear impact on the feasibility and intangible benefits of the technology in a real-world setting. Bin picking relieves MMH tasks for operators, frees up facility space and reduces the collision risk. However, there are several risks with a bin picking solution. Conclusions Bin picking can become financially viable through automating MMH processes. However, how much capital is released depends on the man-hours spent in the previous process. The feasibility of bin picking implementation is dependent on the geometric complexity of the article, the sorting method inside the bin, the surrounding environment and the time margin. Decision-makers need to account for these factors prior to implementation. The intangible benefits can incentivise decision-makers to implement bin picking, even if the financial calculations show a net loss on the investment. / Bakgrund Tillverkare måste anta moderna tekniker för att kunna konkurrera vid toppen av sin bransch. Det kan dock vara både dyrt och svårt att validera ny teknologi före implementation. En teknologi som har haft det svårt att nå en bredare tillämpning är bin picking. Bin picking använder visionteknik för att kommunicera med industrirobotar. Teknologin gör det möjligt för robotar att plocka slumpmässigt sorterade objekt. Den låga tillämpningen av teknologin kan enligt forskare bero på svårigheterna med att bedöma dess prestanda. Forskning fokuserar dessutom främst på de tekniska problemen med bin picking och försummar information som är värdefull för potentiella användare. Syftet Syftet med denna studie är att tillhandahålla underlag för beslutsfattare att bedöma konsekvenserna och möjligheterna med bin picking. Vidare avser studien att inspirera potentiella användare genom att analysera lönsamheten av bin picking via dess genomförbarhet och dess materiella samt immateriella förmåner i en verklig miljö. Metod En nyttofunktion utvecklas och tilldelas kategorier baserat på intervjuer med leverantörer och antagare av teknologin samt från tidigare litteratur. Nyttofunktionen lyfter fram genomförbarheten samt de immateriella förmånerna i en bin  picking lösning, dessutom möjliggör den rangording mellan alternativ. Artikeln som rankas högst används för att genomföra en förstudie tillsammans med leverantörer av bin picking teknologi. En DES som baseras på förstudierna skapas för att lyfta fram de implikationer som kan uppstå i en verklig produktionsmiljö. Slutligen utvärderas lösningen genom finansiella medel, innefattande NPV, IRR, PP och DPP. Resultat Alla NPV-beräkningar (exklusive vid en diskonteringsränta på 12,5 \%) är positiva efter tillräckligt många år. IRR är positivt när tiden överstiger 11 år. PP är 10-11 år, medans DPP är 12-13, 14-15, 19-20, 32-33 år med en diskonteringsränta på 2,5 \%, 5 \%, 7,5 \% respektive 10 \%. Investeringen är dock aldrig återbetalningsbar vid en diskonteringsränta på 12,5 \%. Nyttofunktionens kategorier har en tydlig påverkan gällande teknologins genomförbarhet och immateriella fördelar i en verklig produktionsmiljö. Bin picking avlastar operatörer från MMH-uppgifter, frigör fabriksyta och minskar kollisionsrisken. Det finns dock flera risker med en bin picking lösning. Slutsats Bin picking kan vara ekonomiskt lönsamt genom att automatisera MMH processer. Hur mycket kapital som frigörs beror dock på det antal arbetstimmar som spenderas vid den manuella processen. Genomförbarheten vid implementeringen av bin picking är beroende av artikelns geometriska komplexitet, sorteringsmetod, omgivningen och tidsmarginal. Detta är faktorer som beslutsfattare måste ta hänsyn till före implementation. De immateriella fördelarna kan ge beslutsfattare incitament att införa bin picking, även om de finansiella beräkningar visar en förlust vid en investering.

In this thesis, I take a theoretical dynamic stochastic general equilibrium (DSGE) approach to investigate optimal aggregate dividend policy. I make the following contribution: 1. I extend the standard DSGE model to incorporate a residual dividend policy, external financing and default and find that simulated optimal aggregate payouts are much more volatile than the observed data when other variables are close to the values observed in the data. 2. I examine the sensitivity of optimal aggregate dividend policy to the level of the representative agent¿s habit motive. My results show that, when the habit motive gets stronger, the volatility of optimal aggregate payouts increases while the volatility of aggregate consumption decreases. This is consistent with the hypothesis that investors use cash payouts from well diversified portfolios to help smooth consumption. 3. I demonstrate that the variability of optimal aggregate payouts is sensitive to capital adjustment costs. My simulated results show that costly frictions from changing the capital base of the firm cause optimal aggregate dividends and real investments to be smooth and share prices to be volatile. This finding is consistent with prior empirical observations. 4. I run simulations that support the hypothesis that optimal aggregate dividend policy is similar when the representative firm is risk averse to when it has capital adjustment costs. In both cases, optimal aggregate dividends volatility is very low. 5. In all calibrated DSGE models, apart from case 4, optimal aggregate payouts are found to be countercyclical. This supports the hypothesis that corporations prefer to hold more free cash flows for potential investment opportunities instead of paying dividends when the economy is booming, but is inconsistent with observed data. Keywords: Dynamic Stochastic General Equilibrium (DSGE), real business cycle, utility function, habits, dividends

<p>For prevention and treatment of thrombotic diseases with an anticoagulant drug it is important that an adequate dose is given to avoid occurrence or recurrence of thrombosis, without increasing the risk of bleeding and other adverse events to unacceptable levels. The aim of this thesis was to develop mathematical models that describe the dose-concentration (pharmacokinetic) and concentration-response (pharmacodynamic) relationships of direct thrombin inhibitors, in order to estimate optimal dosages for treatment and long-term secondary prevention of venous thromboembolism (VTE).</p><p>Population pharmacokinetic-pharmacodynamic models were developed, based on data from clinical investigations in healthy volunteers and patients receiving intravenous inogatran, subcutaneous melagatran and/or its oral prodrug ximelagatran. The benefit-risk profiles of different ximelagatran dosages were estimated using clinical utility functions. These functions were based on the probabilities and fatal consequences of thrombosis, bleeding and elevation of the hepatic enzyme alanine aminotransferase (ALAT).</p><p>The studies demonstrate that the pharmacokinetics of melagatran and ximelagatran were predictable and well correlated to renal function. The coagulation marker, activated partial thromboplastin time (APTT), increased non-linearly with increasing thrombin inhibitor plasma concentration. Overall, the systemic melagatran exposure (AUC) and APTT were similarly predictive of thrombosis and bleedings. The identified relationship between the risk of ALAT-elevation and melagatran AUC suggests that the incidence approaches a maximum at high exposures. The estimated clinical utility was favourable compared to placebo in the overall study population and in special subgroups of patients following fixed dosing of ximelagatran for long-term secondary prevention of VTE. Individualized dosing was predicted to add limited clinical benefit in this indication.</p><p>The models developed can be used to support the studied dosage and for selection of alternative dosing strategies that may improve the clinical outcome of ximelagatran treatment. In addition, the models may be extrapolated to aid the dose selection in clinical trials with other direct thrombin inhibitors.</p>

For prevention and treatment of thrombotic diseases with an anticoagulant drug it is important that an adequate dose is given to avoid occurrence or recurrence of thrombosis, without increasing the risk of bleeding and other adverse events to unacceptable levels. The aim of this thesis was to develop mathematical models that describe the dose-concentration (pharmacokinetic) and concentration-response (pharmacodynamic) relationships of direct thrombin inhibitors, in order to estimate optimal dosages for treatment and long-term secondary prevention of venous thromboembolism (VTE). Population pharmacokinetic-pharmacodynamic models were developed, based on data from clinical investigations in healthy volunteers and patients receiving intravenous inogatran, subcutaneous melagatran and/or its oral prodrug ximelagatran. The benefit-risk profiles of different ximelagatran dosages were estimated using clinical utility functions. These functions were based on the probabilities and fatal consequences of thrombosis, bleeding and elevation of the hepatic enzyme alanine aminotransferase (ALAT). The studies demonstrate that the pharmacokinetics of melagatran and ximelagatran were predictable and well correlated to renal function. The coagulation marker, activated partial thromboplastin time (APTT), increased non-linearly with increasing thrombin inhibitor plasma concentration. Overall, the systemic melagatran exposure (AUC) and APTT were similarly predictive of thrombosis and bleedings. The identified relationship between the risk of ALAT-elevation and melagatran AUC suggests that the incidence approaches a maximum at high exposures. The estimated clinical utility was favourable compared to placebo in the overall study population and in special subgroups of patients following fixed dosing of ximelagatran for long-term secondary prevention of VTE. Individualized dosing was predicted to add limited clinical benefit in this indication. The models developed can be used to support the studied dosage and for selection of alternative dosing strategies that may improve the clinical outcome of ximelagatran treatment. In addition, the models may be extrapolated to aid the dose selection in clinical trials with other direct thrombin inhibitors.

In this thesis, I take a theoretical dynamic stochastic general equilibrium (DSGE) approach to investigate optimal aggregate dividend policy. I make the following contribution: 1. I extend the standard DSGE model to incorporate a residual dividend policy, external financing and default and find that simulated optimal aggregate payouts are much more volatile than the observed data when other variables are close to the values observed in the data. 2. I examine the sensitivity of optimal aggregate dividend policy to the level of the representative agent's habit motive. My results show that, when the habit motive gets stronger, the volatility of optimal aggregate payouts increases while the volatility of aggregate consumption decreases. This is consistent with the hypothesis that investors use cash payouts from well diversified portfolios to help smooth consumption. 3. I demonstrate that the variability of optimal aggregate payouts is sensitive to capital adjustment costs. My simulated results show that costly frictions from changing the capital base of the firm cause optimal aggregate dividends and real investments to be smooth and share prices to be volatile. This finding is consistent with prior empirical observations. 4. I run simulations that support the hypothesis that optimal aggregate dividend policy is similar when the representative firm is risk averse to when it has capital adjustment costs. In both cases, optimal aggregate dividends volatility is very low. 5. In all calibrated DSGE models, apart from case 4, optimal aggregate payouts are found to be countercyclical. This supports the hypothesis that corporations prefer to hold more free cash flows for potential investment opportunities instead of paying dividends when the economy is booming, but is inconsistent with observed data. Keywords: Dynamic Stochastic General Equilibrium (DSGE), real business cycle, utility function, habits, dividends.

[pt] Redes de comutação em pacotes vêm substituindo gradativamente as redes de comutação em circuitos, unificando assim, a forma como voz e dados são trafegados. Devido a esse novo conceito de redes comutadas por pacotes, novos desafios referentes ao surgimento de novas aplicações necessitando requisitos de qualidade cada vez mais estritos, vêm recentemente atraindo a atenção de pesquisadores da área de redes de computação. Entre os mecanismos existentes para determinar a qualidade de serviço oferecida por tais redes, o mecanismochave considerado, dos roteadores que desempenham tal função, é o escalonador de pacotes. O conceito que dita à operação de tais escalonadores é o conceito de disciplinas de serviço. No presente trabalho, é proposta uma nova disciplina de serviço onde os pacotes são escalonados de acordo com um indicador de desempenho que mede a satisfação dos usuários baseado no serviço recebido em relação ao serviço contratado. Esse indicador é trazido da economia e se chama Função Utilidade. O sistema considerado no trabalho leva em consideração o funcionamento do roteador de borda de uma provedora de serviços de telecomunicações onde os usuários requerem serviços de VoIP, FTP e HTTP. Os resultados de simulação do sistema mostram que o desempenho da disciplina proposta é comparável, sendo em alguns casos levemente insatisfatória, quando comparada a outras disciplinas existentes na literatura. A lista de trabalhos futuros sugere possibilidades de melhoria deste desempenho, mas que não foram testadas por falta de tempo. / [en] Packet-switched networks have recently replaced circuit-switched networks unifying the way voice and data are carried through those networks. Due to such a new switching concept, challenges related to new applications with different quality requirements have arisen and attracted much attention of computer network researchers. Among the mechanisms used to provide quality of service in those networks, the one chosen in this research is the router’s packet scheduler. The concept that governs packet schedulers is queuing scheduling discipline. In this research, a new queuing scheduling discipline is proposed where packets are scheduled based on a performance indicator that evaluates the user’s satisfaction with the service contracted. That performance indicator is a well-known economic function called utility function. The system considered on this research is focused on the operation of a telecommunications service provider network’s border router where users request FTP, VoIP and HTTP services. Simulation results show that the performance of the proposed service discipline is comparable, but not always to a satisfactory degree, when compared with other service disciplines found in the literature. Some suggestions for future research are added in this dissertation, in order to improve the performance of the proposed service discipline, but they have not been tested due to space and time limitations.