Refine
Year of publication
Document Type
- ZIB-Report (13)
- Doctoral Thesis (2)
- Master's Thesis (1)
Language
- English (16)
Is part of the Bibliography
- no (16)
Keywords
- column generation (16) (remove)
Institute
Generic Branch-Cut-and-Price
(2010)
Robust tail assignment
(2014)
The first part of this thesis is devoted to the general problem of stochastic shortest path problem. It is about searching for the shortest path in a graph in which arc lengths are uncertain and specified by continuous random variables. This problem is at the core of various applications, especially in robust transportation planning where paths correspond to aircraft, train, or bus rotations, crew duties or rosters, etc. We propose a novel solution method based on a discretisation of random variables which is applicable to any class of continuous random variables. We also give bounds on the approximation error of the discretised path lengths compared to the continuous path lengths. In addition, we provide theoretical results for the computational complexity of this method. In the second part we apply this method to a real world airline transportation problem: the so-called tail assignment problem. The goal of the tail assignment problem is to construct aircraft rotations, routes consisting of flight segments, for a set of individual aircraft in order to cover a set of flight segments (legs) while considering operational constraints of each individual aircraft as well as short- to long-term individual maintenance requirements. We state a stochastic programming formulation of this problem and we show how to solve it efficiently by using our method within a column generation framework. We show the gain of our stochastic approach in comparison to standard KPI in terms of less propagated delay and thus less operational costs without growth of computational complexity. A key point of our complex approach to robust optimisation problem is the fit of the underlying stochastic model with reality. We propose a delay propagation model that is realistic, not overfitted, and can therefore be used for forecasting purposes. We benchmark our results using extensive simulation. We show a significant decrease of arrival delays and thus monetary savings on average as well as in the majority of our disruption scenarios. We confirm these benefits in even more life-like benchmarks as simulation where recovery actions are taken and in scenarios which use historical delays directly instead of the stochastic model.
We propose a new coarse-to-fine approach to solve certain linear programs by column generation. The problems that we address contain layers corresponding to different levels of detail, i.e., coarse layers as well as fine layers. These layers are utilized to design
efficient pricing rules. In a nutshell, the method shifts the pricing of a fine linear program to a coarse counterpart. In this way, major decisions are taken in the coarse layer, while minor
details are tackled within the fine layer. We elucidate our methodology by an application to a complex railway rolling stock rotation problem. We provide comprehensive computational results that demonstrate the benefit of this new technique for the solution of large scale problems.
We consider reoptimization (i.e. the solution of a problem based on information available from solving a similar problem) for branch-and-bound algorithms and propose a generic framework to construct a reoptimizing branch-and-bound algorithm.
We apply this to an elevator scheduling algorithm solving similar subproblems to generate columns using branch-and-bound. Our results indicate that reoptimization techniques can substantially reduce the running times of the overall algorithm.
This paper provides a generic formulation for rolling stock planning
problems in the context of intercity passenger traffic. The main contributions
are a graph theoretical model and a Mixed-Integer-Programming
formulation that integrate all main requirements of the considered
Vehicle-Rotation-Planning problem (VRPP). We show that it is
possible to solve this model for real-world instances provided by our
industrial partner DB Fernverkehr AG using modern algorithms and
computers.
Rapid Branching
(2012)
We propose rapid branching (RB) as a general branch-and-bound heuristic
for solving large scale optimization problems in traffic and transport.
The key idea is to combine a special branching rule and a greedy node selection
strategy in order to produce solutions of controlled quality rapidly
and efficiently. We report on three successful applications of the method
for integrated vehicle and crew scheduling, railway track allocation, and
railway vehicle rotation planning.
Providing realistic performance indicators of online algorithms for a given online optimization problem is a difficult task in general. Due to significant drawbacks of other concepts like competitive analysis, Markov decision problems (MDPs) may yield an attractive alternative whenever reasonable stochastic information about future requests is available. However, the number of states in MDPs emerging from real applications is usually exponential in the original input parameters. Therefore, the standard methods for analyzing policies, i.e., online algorithms in our context, are infeasible.
In this thesis we propose a new computational tool to evaluate the behavior of policies for discounted MDPs locally, i.e., depending on a
particular initial state. The method is based on a column generation algorithm for approximating the total expected discounted cost of an
unknown optimal policy, a concrete policy, or a single action (which assumes actions at other states to be made according to an optimal policy). The algorithm determines an $\varepsilon$-approximation by
inspecting only relatively small local parts of the total state space. We prove that the number of states required for providing the approximation is independent of the total number of states, which underlines the practicability of the algorithm. The approximations
obtained by our algorithm are typically much better than the
theoretical bounds obtained by other approaches.
We investigate the pricing problem and the structure of the linear programs encountered in the column generation. Moreover, we propose and analyze different extensions of the basic algorithm in order to achieve good approximations fast.
The potential of our analysis tool is exemplified for discounted MDPs emerging from different online optimization problems, namely online bin coloring, online target date assignment, and online elevator control. The results of the experiments are quite encouraging: our method is mostly capable to provide performance indicators for online algorithms that much better reflect observations made in simulations than competitive analysis does. Moreover, the analysis allows to reveal weaknesses of the considered online algorithms. This way, we developed a new online algorithm for the online bin coloring problem that outperforms existing ones in our analyses and simulations.
This survey concerns optimization problems arising in the design of survivable communication networks. It turns out that such problems can be modeled in a natural way as non-compact linear programming formulations based on multicommodity flow network models. These non-compact formulations involve an exponential number of path flow variables, and therefore require column generation to be solved to optimality. We consider several path-based survivability mechanisms and present results, both known and new, on the complexity of the corresponding column generation problems (called the pricing problems). We discuss results for the case of the single link (or node) failures scenarios, and extend the considerations to multiple link failures. Further, we classify the design problems corresponding to different survivability mechanisms according to the structure of their pricing problem. Finally, we show that almost all encountered pricing problems are hard to solve for scenarios admitting multiple failures.
The \emph{optimal track allocation problem} (\textsc{OPTRA}), also known as the train routing problem or the train timetabling problem, is to find, in a given railway network, a conflict-free set of train routes of maximum value. We propose a novel integer programming formulation for this problem that is based on additional configuration' variables. Its LP-relaxation can be solved in polynomial time. These results are the theoretical basis for a column generation algorithm to solve large-scale track allocation problems. Computational results for the Hanover-Kassel-Fulda area of the German long distance railway network involving up to 570 trains are reported.
The standard computational methods for computing the optimal value functions of Markov Decision Problems (MDP) require the exploration of the entire state space. This is practically infeasible for applications with huge numbers of states as they arise, e.\,g., from modeling the decisions in online optimization problems by MDPs. Exploiting column generation techniques, we propose and apply an LP-based method to determine an $\varepsilon$-approximation of the optimal value function at a given state by inspecting only states in a small neighborhood. In the context of online optimization problems, we use these methods in order to evaluate the quality of concrete policies with respect to given initial states. Moreover, the tools can also be used to obtain evidence of the impact of single decisions. This way, they can be utilized in the design of policies.