GRASP is a multi-start metaheuristic for combinatorial optimization problems, in which each iteration consists basically of two phases: construction and local search. We discuss some advances and extensions of the basic GRASP procedure, such as path-relinking as a memory-based intensification strategy and restart strategies to speedup the search. Successful implementation techniques are discussed and illustrated by numerical results obtained for different applications. We also illustrate the use of multiple time-to-target plots as a useful tool to characterize, evaluate, and compare the behavior of randomized heuristics.
The Fourth Industrial Revolution or the Industry 4.0 (Ind4.0) dismantles the existing production methods and breaks the boundaries between industries, thereby creating new demands and reshaping the markets. The new production systems in the era of Ind4.0 can be identified as manufacturing-service synthesis or ‘servicification’ due to the underlying mega trend of integration between products and services. The platform is actually at the heart of this change. However, the existing analytical frameworks developed based on the linear value chain model may not properly capture the value production mechanisms in the new era. Therefore, the academia has been trying to develop new perspectives together with novel analytical tools. One of the representative academic achievements is the notion of the business ecosystem organized by a multi-sided platform. While the economic characteristics of the multi-sided platforms and their ecosystems have been widely studied, however, the operations and their value creation mechanisms in the platform ecosystems have not been actively studied. In this background, I first introduce a couple of studies addressing important operational issues in the multi-sided platform ecosystems. They include the newsvendor model extended in the context of the two-sided markets (Chou et al, 2012), platform operations in the sharing economy (Kostami et al., 2017), and the nonlinear value creation mechanisms in the value ecosystems (Kim, 2017; 2018). Except for these studies, the lack of research on the nonlinear production systems in the platform-based servicification could be found in that we have not yet fully figured out the key issues on the current changes of the production mechanisms. Accordingly, I suggest some important issues and research points on the production system evolution led by platform-based servicification in the era of Ind4.0. These are the subjects such as complementarity and flexibility in resource reconfiguration, of which prior studies already exist in the field of management sciences in the 1980s and 1990s (e.g., Milgrom & Roberts, 1990; Roth, 1982). In sum, management science plays an important role in developing a more systematic approach and research framework for the nonlinear production mechanisms in the value ecosystems as well as the integrated manufacturing-service operations of the platform businesses in the era of Ind4.0.
Dos temas que han aparecido crecientemente en el manejo forestal de cosechas0 son los posibles efectos de cambio climático por un lado y el control de incendios por otro. En ambos casos hay un nivel de incertidumbre importante. Para el caso de cambio climático se han desarrollado escenarios de climas futuros, que a su vez han dado origen a escenarios de crecimiento forestal. Presentamos modelos estocásticos y de múltiples objetivos que optimizan decisiones de cosecha considerando estos escenarios. Se busca soluciones robustas, es decir, que tengan buen comportamiento para cualquiera de los escenarios posibles. Debe notarse la importancia de los escenarios ‘cisnes negros’ , que son aquellos de baja probabilidad de ocurrencia, que causan mucho daño si ocurren. Para el caso de incendios, se considera manejos que impidan que incendios se propaguen demasiado. Esto se logra con planes de cosecha de unidades que puedan hacer de cortafuegos, es decir, ralenticen el avance del fuego, en particular hacia zonas de alto valor de bosque, o más importante, zonas habitadas. Presentamos modelos de simulación y optimización estocásticos que apoyan decisiones de cosecha para estos efectos. Ilustramos los modelos en ambos casos con resolución de casos basados en bosques en Portugal y Canadá.
Increased competition in a globalized economy, real-time access to a wealth of transparent information, the rise of a more knowledgeable and pragmatic generation of consumers is currently changing the perception and nature of optimal pricing. Modeling the customer reaction to price is nowadays a central issue for most companies.
Bilevel models are well suited to capture such situations where two decision makers act non cooperatively and in a sequential way. More precisely, bilevel programs allow modeling situations in which a main agent, theleader, strives to optimize a given quantity but controls only a subset of the decision variables. The remaining variables fall under the control of a second agent, the follower, who solves its own problem by taking into account the decisions taken by the leader. In other words, bilevel programs can be considered as demand-offer equilibrium models where the demand is the result of another mathematical problem.
In this talk we first present a brief theoretical study of bilevel problems. Next we focus on two special cases : a price setting problem on a network and a demand side and revenue management problem in the energy field.For each of these two applications, we define the models, study their properties and sketch solution methods based on the structures of the problems.
The ability of traffic flows to adapt their rate and fairly use all available resources is one of the Internet's pillars. However, this traffic characteristic, often referred to as elasticity, has not yet been fully considered in the optimization literature.
We present a new approach to network routing with elastic demands, where the interaction between the network operator and the congestion control scheme is modeled as a Stackelberg game, whose equilibrium can be computed by solving a bilevel optimization problem. In the first level, the network operator chooses a routing path for each origin-destination pair so as to maximize a utility function while, in the second level, the congestion control scheme determines a bandwidth allocation to the flows by maximizing a fairness measure.
After discussing structural properties and complexity issues, we describe single-level mixed-integer programming formulations or approximations for two known fairness measures. Then we summarize our work on exact and heuristic methods and report some computational results. The numerical experiments confirm the substantial difference between the advocated traffic engineering approach and the standard ones which neglect the bilevel nature of Internet routing.
The talk is based on joint works with Antonio Capone, Stefano Coniglio, Luca Gianoli and Leonardo Taccari.
Considering the importance of the seminal work "Partitioning Procedures for Solving Mixed Variables Programming Problems" (published by J. F. Benders in 1962) the lecture presents a review of the use of the theories of partition and decomposition of Benders, and its variations, from 1962 to date. As central part it discussed the use of multilevel parallelism allowed by Benders methodology using modern computational architectures (based on multiple CPUs, multiples GPUs and large storages of RAM memory), it is of utmost importance to solve very large problems in "reasonable" solution times. This is the basis of the "Asynchronous Parallel Optimization" and the "Real-Time Distributed Optimization" as fundamental approaches to the optimization of the future, both based on Cross Decomposition, the integration of Benders Theory with Lagrangeana Relaxation, atomizing the problem in multiple types of sub-problems based on the concepts of Dynamic Programming and Optimal Discrete Control. These concepts will be presented considering its applicability in the solution of large scale real problems.
Green supply chain management and sustainable operations management arose long time ago with a growing public pressure that resulted in strong regulations. Peak attention was reached in 2010 and is continuously thriving. The extension of classical supply chains to include “greening concepts” is a complex endeavor for companies especially when outreaching to their supply chain(s). Greening concepts can be widely described as reducing all forms of waste and thus using all kinds of resources with such care, so that they are available for future generations – which is sustainable. A plethora of literature and best practice has been developed throughout the years advising companies how to enhance performances in their supply chain(s) with respect to the greening movement. Areas include green supplier selection, environmental emissions reduction, waste reduction, recycling, remanufacturing & refurbishing, reverse logistics – just to mention a few – and how to measure green performance building up environmental management systems in accordance with developed standards such as ISO 14000.
With the recent development in politics and practice, it is time to revise how far industries have come in their efforts to greening their supply chains and take their stake in operating in a sustainable manner. In this talk, we will review the past, look at the present developments under the green label and give thoughts on what is (still) needed to enable every stakeholder – which we all are as consumers and/or producers – to partake in a sustainable future.
In this talk, I present some aspects of continuous minimization on Riemannian manifolds. It is generally accepted that the first attempt to apply optimization techniques to problems on manifolds was presented by Luenberger in 1972.
After that, there were an extensive development of that area, mostly on Riemannian manifolds, but also on more general nonlinear spaces. The main objectives are the extension of the necessary theory, as convex analysis (Udriste, 1994), as of classical methods of continuous optimization in linear spaces. Some essential concepts as geodesics and curvature of the manifold are presented in a descriptive form. They will be necessary to show methods as geodesic (gradient) descent and proximal methods.