Search

23rd EURO Working Group on Transportation Meeting, EWGT 2020, online, 16 Sep 2020 - 18 Sep 2020; Transportation research procedia 52, 123-130 (2021). doi:10.1016/j.trpro.2021.01.014 special issue: "23rd EURO Working Group on Transportation Meeting, EWGT 2020, 16-18 September 2020, Paphos, Cyprus / Edited by Stelios Timotheou, Christos G. Panayiotou", Published by Elsevier, Amsterdam [u.a.]

Two important characteristics encountered in many real-world scheduling problems are heterogeneous processors and a certain degree of uncertainty about the processing times of jobs. In this paper we address both, and study for the first time a scheduling problem that combines the classical unrelated machine scheduling model with stochastic processing times of jobs. By means of a novel time-indexed linear programming relaxation, we show how to compute in polynomial time a scheduling policy with provable performance guarantee for the stochastic version of the unrelated parallel machine scheduling problem with the weighted sum of completion times objective. Our performance guarantee depends on the squared coefficient of variation of the processing times and we show that this dependence is tight. Currently best-known bounds for deterministic scheduling problems are contained as special cases.

Scheduling a set of n jobs on m identical parallel machines so as to minimize the makespan or maximize the minimum machine load are two of the most important and fundamental scheduling problems studied in the literature. We consider the general online scenario where jobs are consecutively added to and/or deleted from an instance. The goal is to maintain a near-optimal assignment of the current set of jobs to the m machines. This goal is essentially doomed to failure unless, upon arrival or departure of a job, we allow reassigning some other jobs. Considering that the reassignment of a job induces a cost proportional to its size, the total cost for reassigning jobs must preferably be bounded by a constant r times the total size of added or deleted jobs. The value r is called the reassignment factor of the solution and it is a measure of our willingness to adapt the solution over time. Our main result is that, for any ε > 0, it is possible to achieve (1 + ε)-competitive solutions with constant reassignment factor r(ε). For the minimum makespan problem this is the first improvement on the (2 + ε)-competitive algorithm by Andrews et al. (1999) [Andrews M, Goemans M, Zhang L (1999) Improved bounds for on-line load balancing. Algorithmica 23(4):278–301]. Crucial to our algorithm is a new insight into the structure of robust, almost optimal schedules.

We consider online versions of the minimum spanning tree (MST) problem and the traveling salesman problem (TSP) where recourse is allowed. The nodes of an unknown graph with metric edge cost appear one by one and must be connected in such a way that the resulting tree or tour has low cost. In the standard online setting, with irrevocable decisions, no algorithm can guarantee a constant-competitive ratio. In our model we allow recourse actions by giving a limited budget of edge rearrangements per iteration. It has been an open question for more than 20 years whether an online algorithm equipped with a constant (amortized) budget can guarantee constant-approximate solutions. As our main result, we answer this question affirmatively in an amortized setting. We introduce an algorithm that maintains a nearly optimal tree when given a constant amortized budget. Unlike in classical TSP variants, the standard double-tree and shortcutting approach does not give constant guarantees in the online setting. We propose...

We present a e/(e1)-approximation algorithm for the nonpreemptive scheduling problem to minimize the total weighted completion time of jobs on a single machine subject to release dates and precedence constraints. The previously best known approximation algorithm dates back to 1997; its performance guarantee can be made arbitrarily close to the Euler constant e (Schulz and Skutella, 1997).

Network flows over time form a fascinating area of research. They model the temporal dynamics of network flow problems occurring in a wide variety of applications. Research in this area has been pursued in two different and mainly independent directions with respect to time modeling: discrete and continuous time models. In this paper we deploy measure theory in order to introduce a general model of network flows over time combining both discrete and continuous aspects into a single model. Here, the flow on each arc is modeled as a Borel measure on the real line (time axis) which assigns to each suitable subset a real value, interpreted as the amount of flow entering the arc over the subset. We focus on the maximum flow problem formulated in a network where capacities on arcs are also given as Borel measures and storage might be allowed at the nodes of the network. We generalize the concept of cuts to the case of these Borel Flows and extend the famous MaxFlow-MinCut Theorem.

Earliest arrival flows model a central aspect of evacuation planning: In a dangerous situation, as many individuals as possible should be rescued at any point in time . Unfortunately, given a network with multiple sinks, flows over time satisfying this condition do not always exist. We analyze the special case of flows over time with zero transit times and characterize which networks always allow for earliest arrival flows.

Earliest arrival flows capture the essence of evacuation planning. Given a network with capacities and transit times on the arcs, a subset of source nodes with supplies and a sink node, the task is to send the given supplies from the sources to the sink “as quickly as possible.” The latter requirement is made more precise by the earliest arrival property, which requires that the total amount of flow that has arrived at the sink is maximal for all points in time simultaneously. It is a classical result from the 1970s that, for the special case of a single source node, earliest arrival flows do exist and can be computed by essentially applying the successive shortest-path algorithm for min-cost flow computations. Although it has previously been observed that an earliest arrival flow still exists for multiple sources, the problem of computing one efficiently has been open for many years. We present an exact algorithm for this problem whose running time is strongly polynomial in the input plus output size of the problem.

Consider the classical online scheduling problem where jobs that arrive one by one are assigned to identical parallel machines with the objective of minimizing the makespan. We generalize this problem by allowing the current assignment to be changed whenever a new job arrives, subject to the constraint that the total size of moved jobs is bounded by~$\beta$ times the size of the arriving job. Our main result is a linear time `online approximation scheme', that is, a family of online algorithms with competitive ratio~$1+\epsilon$ and constant migration factor~$\beta(\epsilon)$, for any fixed~$\epsilon>0$. This result is of particular importance if considered in the context of sensitivity analysis: While a newly arriving job may force a complete change of the entire structure of an optimal schedule, only very limited `local' changes suffice to preserve near-optimal solutions. We believe that this concept will find wide application in its own right. We also present simple deterministic online algorithms with migration factors~$\beta=2$ and~$\beta=4/3$, respectively. Their competitive ratio~$3/2$ beats the lower bound on the performance of any online algorithm in the classical setting without migration. We also present improved algorithms and similar results for closely related problems. In particular, there is a short discussion of corresponding results for the objective to maximize the minimum load of a machine. The latter problem has an application for configuring storage servers that was the original motivation for this work.

We analyze teh performance fo evolutionary algorithms on various matroid optimization problems that encompass a vast number of efficiently solvable as well as NP-hard combinatorial optimization problems (including many well-known examples such as minimum spanning tree and maximum bipartite matching). We obtain very promising bounds on the expected running time and quality of the computed solution. Our results establish a better theoretical understanding of why randomized search heuristics yield empirically good results for many real-world problems., Reihe CI; 225-07

Due to the importance of robustness in many real-world optimization problems, the field of robust optimization has gained a lot of attention over the past decade. We concentrate on maximum flow problems and introduce a novel robust optimization model which, compared to known models from the literature, features several advantageous properties: (i) We consider a general class of path-based flow problems which can be used to model a large variety of network routing problems (and other packing problems). (ii) We aim at solutions that are robust against targeted attacks by a potent adversary who may attack any flow path of his choice on any edge of the network. (iii) In contrast to previous robust maximum flow models, for which no efficient algorithms are known, optimal robust flows for the most important basic variants of our model can be found in polynomial time. We also consider generalizations where the flow player can spend a budget to protect the network against the interdictor. Here, we show that the problem can be solved efficiently when the interdiction costs are determined by the flow player from scratch. However, the problem becomes hard to approximate when the flow player has to improve an initial protection infrastructure.

Flows over time (also called dynamic flows) generalize standard network flows by introducing an element of time. They naturally model problems where travel and transmission are not instantaneous. Traditionally, flows over time are solved in time-expanded networks that contain one copy of the original network for each discrete time step. While this method makes available the whole algorithmic toolbox developed for static flows, its main and often fatal drawback is the enormous size of the time-expanded network. We present several approaches for coping with this difficulty. First, inspired by the work of Ford and Fulkerson on maximal $s$-$t$-flows over time (or “maximal dynamic $s$-$t$-flows”), we show that static length-bounded flows lead to provably good multicommodity flows over time. Second, we investigate “condensed” time-expanded networks which rely on a rougher discretization of time. We prove that a solution of arbitrary precision can be computed in polynomial time through an appropriate discretization leading to a condensed time-expanded network of polynomial size. In particular, our approach yields fully polynomial-time approximation schemes for the NP-hard quickest min-cost and multicommodity flow problems. For single commodity problems, we show that storage of flow at intermediate nodes is unnecessary, and our approximation schemes do not use any.

We study the incremental facility location problem, wherein we are given an instance of the uncapacitated facility location problem UFLP and seek an incremental sequence of opening facilities and an incremental sequence of serving customers along with their fixed assignments to facilities open in the partial sequence. We say that a sequence has a competitive ratio of k, if the cost of serving the first i¾? customers in the sequence is at most k times the optimal solution for serving any i¾? customers for all possible values of i¾?. We provide an incremental framework that computes a sequence with a competitive ratio of at most eight and a worst-case instance that provides a lower bound of three for any incremental sequence. We also present the results of our computational experiments carried out on a set of benchmark instances for the UFLP. The problem has applications in multistage network planning. © 2015 Wiley Periodicals, Inc. NETWORKS, Vol. 654, 306-311 2015

In traditional multi-commodity flow theory, the task is to send a certain amount of each commodity from its start to its target node, subject to capacity constraints on the edges. However, no restriction is imposed on the number of paths used for delivering each commodity; it is thus feasible to spread the flow over a large number of different paths. Motivated by routing problems arising in real-life applications, e.g., telecommunication, unsplittable flows have moved into the focus of research. Here, the demand of each commodity may not be split but has to be sent along a single path. In this paper a generalization of this problem is studied. In the considered flow model, a commodity can be split into a bounded number of chunks which can then be routed on different paths. In contrast to classical (splittable) flows and unsplittable flows, the single-commodity case of this problem is already NP-hard and even hard to approximate. We present approximation algorithms for the single- and multi-commodity case and point out strong connections to unsplittable flows. Moreover, results on the hardness of approximation are presented. In particular, we show that some of our approximation results are in fact best possible, unless P = NP.

In many scheduling applications it is required that the processing of some job be postponed until some other job, which can be chosen from a pregiven set of alternatives, has been completed. The traditional concept of precedence constraints fails to model such restrictions. Therefore, the concept has been generalized to so-called AND/OR precedence constraints which can cope with this kind of requirement. In the context of traditional precedence constraints, feasibility, transitivity, and the computation of earliest start times for jobs are fundamental, well-studied problems. The purpose of this paper is to provide efficient algorithms for these tasks for the more general model of AND/OR precedence constraints. We show that feasibility as well as many questions related to transitivity can be solved by applying essentially the same linear-time algorithm. In order to compute earliest start times we propose two polynomial-time algorithms to cope with different classes of time distances between jobs.

We consider the problem of preemptively scheduling a set of n jobs on m (identical, uniformly related, or unrelated) parallel machines. The scheduler may reject a subset of the jobs and thereby incur job-dependent penalties for each rejected job, and he must construct a schedule for the remaining jobs so as to optimize the preemptive makespan on the m machines plus the sum of the penalties of the jobs rejected. We provide a complete classification of these scheduling problems with respect to complexity and approximability. Our main results are on the variant with an arbitrary number of unrelated machines. This variant is APX-hard, and we design a 1.58-approximation algorithm for it. All other considered variants are weakly NP-hard, and we provide fully polynomial time approximation schemes for them. Finally, we argue that our results for unrelated machines can be carried over to the corresponding preemptive open shop scheduling problem with rejection.

We consider the optimization problem of scheduling a given set of jobs on unrelated parallel machines with total weighted completion time objective. This is a classical scheduling problem known to be NP-hard since the 1970s. We give a new and simplified version of the currently best-known approximation algorithm, which dates back to 1998. It achieves performance ratio 3 / 2, and is based on an optimal solution to a convex quadratic program.

We consider the scheduling problem of minimizing the average weighted completion time of n jobs with release dates on a single machine. We first study two linear programming relaxations of the problem, one based on a time-indexed formulation, the other on a completion-time formulation. We show their equivalence by proving that a O(n log n) greedy algorithm leads to optimal solutions to both relaxations. The proof relies on the notion of mean busy times of jobs, a concept which enhances our understanding of these LP relaxations. Based on the greedy solution, we describe two simple randomized approximation algorithms, which are guaranteed to deliver feasible schedules with expected objective function value within factors of 1.7451 and 1.6853, respectively, of the optimum. They are based on the concept of common and independent $\alpha$-points, respectively. The analysis implies in particular that the worst-case relative error of the LP relaxations is at most 1.6853, and we provide instances showing that it is at least $e/(e-1) \approx 1.5819$. Both algorithms may be derandomized; their deterministic versions run in O(n2) time. The randomized algorithms also apply to the on-line setting, in which jobs arrive dynamically over time and one must decide which job to process without knowledge of jobs that will be released afterwards.

We present a new class of randomized approximation algorithms for unrelated parallel machine scheduling problems with the average weighted completion time objective. The key idea is to assign jobs randomly to machines with probabilities derived from an optimal solution to a linear programming (LP) relaxation in time-indexed variables. Our main results are a $(2+\varepsilon)$-approximation algorithm for the model with individual job release dates and a $(3/2+\varepsilon)$-approximation algorithm if all jobs are released simultaneously. We obtain corresponding bounds on the quality of the LP relaxation. It is an interesting implication for identical parallel machine scheduling that jobs are randomly assigned to machines, in which each machine is equally likely. In addition, in this case the algorithm has running time O(n log n) and performance guarantee 2. Moreover, the approximation result for identical parallel machine scheduling applies to the on-line setting in which jobs arrive over time as well, with no difference in performance guarantee.

Flows over time are used to model many real-world logistic and routing problems. The networks underlying such problems-streets, tracks, etc.-are inherently undirected and directions are only imposed on them to reduce the danger of colliding vehicles and similar problems. Thus, the question arises, what influence the orientation of the network has on the network flow over time problem that is being solved on the oriented network. In the literature, this is also referred to as the contraflow or lane reversal problem. We introduce and analyze the price of orientation: How much flow is lost in any orientation of the network if the time horizon remains fixed? We prove that there is always an orientation where we can still send one-third of the flow and this bound is tight. For the special case of networks with a single source or sink, this fraction is half, which is again tight. We present more results of similar flavor and also show nonapproximability results for finding the best orientation for single and multicommodity maximum flows over time. © 2015 Wiley Periodicals, Inc. NETWORKS, Vol. 663, 196-209 2015

The stable allocation problem is one of the broadest extensions of the well-known stable marriage problem. In an allocation problem, edges of a bipartite graph have capacities and vertices have quotas to fill. Here we investigate the case of uncoordinated processes in stable allocation instances. In this setting, a feasible allocation is given and the aim is to reach a stable allocation by raising the value of the allocation along blocking edges and reducing it on worse edges if needed. Do such myopic changes lead to a stable solution? In our present work, we analyze both better and best response dynamics from an algorithmic point of view. With the help of two deterministic algorithms we show that random procedures reach a stable solution with probability one for all rational input data in both cases. Surprisingly, while there is a polynomial path to stability when better response strategies are played (even for irrational input data), the more intuitive best response steps may require exponential time. We also study the special case of correlated markets. There, random best response strategies lead to a stable allocation in expected polynomial time.

Since the early days of combinatorial optimization, algorithms and techniques from the closely related area of mathematical programming have played a pivotal role in solving combinatorial optimization problems. This holds both for ‘easy’ problems that can be solved efficiently in polynomial time, such as, e. g., the weighted matching problem [3], as well as for NP-hard problems whose solution might take exponential time in the worst case, such as, e. g., the traveling salesperson problem [1].

We consider the online MST and TSP problems with recourse. The nodes of an unknown graph with metric edge cost appear one by one and must be connected in such a way that the resulting tree or tour has low cost. In the standard online setting, with irrevocable decisions, no algorithm can guarantee a constant competitive ratio. In our model we allow recourse actions by giving a limited budget of edge rearrangements per iteration. It has been an open question for more than 20 years if an online algorithm equipped with a constant (amortized) budget can guarantee constant-approximate solutions [7]. As our main result, we answer this question affirmatively in an amortized setting. We introduce an algorithm that maintains a nearly optimal tree when given constant amortized budget. In the non-amortized setting, we specify a promising proof technique and conjecture a structural property of optimal solutions that would prove a constant competitive ratio with a single recourse action. It might seem rather optimistic that such a small budget should be sufficient for a significant cost improvement. However, we can prove such power of recourse in the offline setting in which the sequence of node arrivals is known. Even this problem prohibits constant approximations if there is no recourse allowed. Surprisingly, already a smallest recourse budget significantly improves the performance guarantee from non-constant to constant. Unlike in classical TSP variants, the standard double-tree and shortcutting approach does not give constant guarantees in the online setting. However, a non-trivial robust shortcutting technique allows to translate online MST results into TSP results at the loss of small factors.

Flows over time and generalized flows are two advanced network flow models of utmost importance, as they incorporate two crucial features occurring in numerous real-life networks. Flows over time feature time as a problem dimension and allow to realistically model the fact that commodities (goods, information, etc.) are routed through a network over time. Generalized flows allow for gain/loss factors on the arcs that model physical transformations of a commodity due to leakage, evaporation, breeding, theft, or interest rates. Although the latter effects are usually time-bound, generalized flow models featuring a temporal dimension have never been studied in the literature. In this paper we introduce the problem of computing a generalized maximum flow over time in networks with both gain factors and transit times on the arcs. While generalized maximum flows and maximum flows over time can be computed efficiently, our combined problem turns out to be NP-hard and even completely non-approximable. A natural special case is given by lossy networks where the loss rate per time unit is identical on all arcs. For this case we present a (practically efficient) FPTAS.

In this paper, we consider the generalized min-sum set cover problem, introduced by Azar, Gamzu, and Yin (STOC 2009). Bansal, Gupta, and Krishnaswamy (SODA 2010) give a 485-approximation algorithm for the problem. We are able to alter their algorithm and analysis to obtain a 28-approximation algorithm, improving the performance guarantee by an order of magnitude. We use concepts from α -point scheduling to obtain our improvements.

We study flows over time in networks with transit times on the arcs. Transit times describe how long it takes to traverse an arc. A flow over time specifies for each arc a time-dependent flow rate that must always be bounded by the arc's capacity. Only recently, Melkonian introduced an alternative model where so-called bridge capacities bound the total amount of flow traveling along an arc, at any point of time. The contribution of this paper is twofold. Firstly, we introduce a common generalization of both the classical flow over time model and Melkonian's model. Secondly, we present a non-trivial extension of an FPTAS by Fleischer and Skutella to our new flow model. Prior to this, no approximation algorithm was known for Melkonian's model.

In welding cells a certain number of robots perform spot welding tasks on a workpiece. The tours of the welding robots are planned in such a way that all weld points on the component are visited and processed within the cycle time of the production line. During this operation, the robot arms must not collide with each other and safety clearances have to be kept. On the basis of these specifications, we concentrate on the Discrete Optimization aspects of the stated problem. This leads to a Vehicle Routing based problem with additional scheduling and timing aspects induced by the necessary collision avoidance. This problem can then be solved as an Integer Linear Program by Column Generation techniques. In this context, we adapt the Resource Constrained Shortest Path Problem, so that it can be used to solve the pricing problem with collision avoidance. Using this approach, we can solve generated test instances based on real world welding cells of reasonable size.

We consider scheduling on an unreliable machine that may experience unexpected changes in processing speed or even full breakdowns. We aim for a universal solution that performs well without adaptation for any possible machine behavior. For the objective of minimizing the total weighted completion time, we design a polynomial time deterministic algorithm that finds a universal scheduling sequence with a solution value within 4 times the value of an optimal clairvoyant algorithm that knows the disruptions in advance. A randomized version of this algorithm attains in expectation a ratio of e. We also show that both results are best possible among all universal solutions. As a direct consequence of our results, we answer affirmatively the question of whether a constant approximation algorithm exists for the offline version of the problem when machine unavailability periods are known in advance. When jobs have individual release dates, the situation changes drastically. Even if all weights are equal, there are instances for which any universal solution is a factor of Ω(log n/ log log n) worse than an optimal sequence. Motivated by this hardness, we study the special case when the processing time of each job is proportional to its weight. We present a non-trivial algorithm with a small constant performance guarantee. © 2010 Springer-Verlag.

We report on the solution of a real-time scheduling problem that arises in the design of software-based operation control of aircraft. A set of tasks has to be distributed on a minimum number of machines and offsets of the tasks have to be computed. The tasks emit jobs periodically starting at their offset and then need to be executed on the machines without any delay. Also, further constraints in terms of memory usage and redundancy requirements have to be met. Approaches based on standard integer programming formulations fail to solve our real-world instances. By exploiting structural insights of the problem we obtain an IP-formulation and primal heuristics that together solve the real-world instances to optimality and outperform text-book approaches by several orders of magnitude. Our methods lead, for the first time, to an industry strength tool to optimally schedule aircraft sized problems.

Minimizing the makespan or maximizing the minimum machine load are two of the most important and fundamental parallel machine scheduling problems. In an online scenario, jobs are consecutively added and/or deleted and the goal is to always maintain a (close to) optimal assignment of jobs to machines. The reassignment of a job induces a cost proportional to its size and the total cost for reassigning jobs must preferably be bounded by a constant r times the total size of added or deleted jobs. Our main result is that, for any e > 0, one can always maintain a (1 + e)-competitive solution for some constant reassignment factor r(e). For the minimum makespan problem this is the first improvement of the (2+e)-competitive algorithm with constant reassignment factor published in 1996 by Andrews, Goemans, and Zhang.

Runtime analyses of randomized search heuristics for combinatorial optimization problems often depend on the size of the largest weight. We consider replacing the given set of weights with smaller weights such that the behavior of the randomized search heuristic does not change. Upper bounds on the size of the new, equivalent weights allow us to obtain upper bounds on the expected runtime of such randomized search heuristics independent of the size of the actual weights. Furthermore we give lower bounds on the largest weights for worst-case instances. Finally we present some experimental results, including examples for worst-case instances., Reihe CI; 253-08

Scheduling jobs on unrelated parallel machines so as to minimize the makespan is one of the basic, well-studied problems in the area of machine scheduling. In the first part of the paper we prove that the power of preemption, i.e., the ratio between the makespan of an optimal nonpreemptive and an optimal preemptive schedule, is exactly 4. This result is a definite answer to an important basic open problem in scheduling. The proof of the lower bound is based on a clever iterative construction while the rounding technique we use to prove the upper bound is an adaptation of Shmoys and Tardos' rounding for the generalized assignment problem. In the second part of the paper we apply this adaptation to the more general setting in which orders, consisting of several jobs, have to be processed on unrelated parallel machines so as to minimize the sum of weighted completion times of the orders. We obtain the first constant factor approximation algorithms for the preemptive and nonpreemptive case, improving and extending a recent result by Leung et. al.

We study the minimum s-t-cut problem in graphs with costs on the edges in the context of evolutionary algorithms. Minimum cut problems belong to the class of basic network optimization problems that occur as crucial subproblems in many real-world optimization problems and have a variety of applications in several different areas. We prove that there exist instances of the minimum s-t-cut problem that cannot be solved by standard single-objective evolutionary algorithms in reasonable time. On the other hand, we develop a bicriteria approach based on the famous MaxFlow-MinCut Theorem that enables evolutionary algorithms to find an optimum solution in expected polynomial time., Reihe CI; 242-08

Algorithmica, 47 (3), ISSN:0178-4617, ISSN:1432-0541

Earliest arrival flows capture the essence of evacuation planning. Given a network with capacities and transit times on the arcs, a subset of source nodes with supplies and a sink node, the task is to send the given supplies from the sources to the sink "as quickly as possible". The latter requirement is made more precise by the earliest arrival property which requires that the total amount of flow that has arrived at the sink is maximal for all points in time simultaneously. It is a classical result from the 1970s that, for the special case of a single source node, earliest arrival flows do exist and can be computed by essentially applying the successive shortest path algorithm for min-cost flow computations. While it has previously been observed that an earliest arrival flow still exists for multiple sources, the problem of computing one efficiently has been open for many years. We present an exact algorithm for this problem whose running time is strongly polynomial in the input plus output size of the problem

Many approximation results for single machine scheduling problems rely on the conversion of preemptive schedules into (preemptive or non-preemptive) solutions. The initial preemptive schedule is usually an optimal solution to a combinatorial relaxation or a linear programming relaxation of the scheduling problem under consideration. It therefore provides a lower bound on the optimal objective function value. However, it also contains structural information which is useful for the construction of provably good feasible schedules. In this context, list scheduling in order of so-called α-points has evolved as an important and successful tool. We give a survey and a uniform presentation of several approximation results for single machine scheduling with total weighted completion time objective from the last years which rely on the concept of α-points.

We consider parallel, identical machine scheduling problems where the jobs are subject to precedence constraints, release dates, and the processing times of jobs are governed by independent probability distributions. The objective is to minimize the expected value of the total weighted completion time. Building upon a linear programming relaxation by Möhring, Schulz and Uetz (Journal of the ACM 46, 1999, 924-942) and an idle time charging scheme by Chekuri, Motwani, Natarajan, and Stein (SIAM Journal on Computing 31, 2001, 146-166) we derive the first constant-factor approximation algorithms for this model.

Consider the classical online scheduling problem where jobs that arrive one by one are assigned to identical parallel machines with the objective of minimizing the makespan. We generalize this problem by allowing the current assignment to be changed whenever a new job arrives, subject to the constraint that the total size of moved jobs is bounded by β times the size of the arriving job. Our main result is a linear time 'online approximation scheme', that is, a family of online algorithms with competitive ratio 1 + ∈ and constant migration factor β(∈), for any fixed e > 0. This result is of particular importance if considered in the context of sensitivity analysis: While a newly arriving job may force a complete change of the entire structure of an optimal schedule, only very limited 'local' changes suffice to preserve near-optimal solutions. We believe that this concept will find wide application in its own right. We also present simple deterministic online algorithms with migration factors β = 2 and β = 4/3, respectively. Their competitive ratio 3/2 beats the lower bound on the performance of any online algorithm in the classical setting without migration. We also present improved algorithms and similar results for closely related problems. In particular, there is a short discussion of corresponding results for the objective to maximize the minimum load of a machine. The latter problem has an application for configuring storage servers that was the original motivation for this work.

Flow variation over time is an important feature in network flow problems arising in various applications such as road or air traffic control, production systems, communication networks (e.g., the Internet), and financial flows. The common characteristic are networks with capacities and transit times on the arcs which specify the amount of time it takes for flow to travel through a particular arc. Moreover, in contrast to static flow problems, flow values on arcs may change with time in these networks. While the 'maximum s-t-flow over time' problem can be solved efficiently and 'min-cost flows over time' are known to be NP-hard, the complexity of (fractional) 'multicommodity flows over time' has been open for many years. We prove that this problem is NP-hard, even for series-parallel networks, and present new and efficient algorithms under certain assumptions on the transit times or on the network topology. As a result, we can draw a complete picture of the complexity landscape for flow over time problems.

We consider the problem of scheduling unrelated parallel machines so as to minimize the total weighted completion time of jobs. Whereas the best previously known approximation algorithms for this problem are based on LP relaxations, we give a 3/2-approximation algorithm that relies on a convex quadratic programming relaxation. For the special case of two machines we present a further improvement to a 1.2752-approximation; we introduce a more sophisticated semidefinite programming relaxation and apply the random hyperplane technique introduced by M.X. Goemans and D.P. Williamson (1995) for the MAXCUT problem and its refined version of U. Feige and M.X. Goemans (1995). To the best of our knowledge, this is the first time that convex and semidefinite programming techniques (apart from LPs) are used in the area of scheduling.

In the single source unsplittable min-cost flow problem, commodities must be routed simultaneously from a common source vertex to certain destination vertices in a given graph with edge capacities and costs; the demand of each commodity must be routed along a single path so that the total flow through any edge is at most its capacity. Moreover, the total cost must not exceed a given budget. This problem has been introduced by Kleinberg [7] and generalizes several NP-complete problems from various areas in combinatorial optimization such as packing, partitioning, scheduling, load balancing, and virtual-circuit routing. Kolliopoulos and Stein [9] and Dinitz, Garg, and Goemans [4] developed algorithms improving the first approximation results of Kleinberg for the problem of minimizing the violation of edge capacities and for other variants. However, known techniques do not seem to be capable of providing solutions without also violating the cost constraint. We give the first approximation results with hard cost constraints. Moreover, all our results dominate the best known bicriteria approximations. Finally, we provide results on the hardness of approximation for several variants of the problem.

In traditional multi-commodity flow theory, the task is to send a certain amount of each commodity from its start to its target node, subject to capacity constraints on the edges. However, no restriction is imposed on the number of paths used for delivering each commodity; it is thus feasible to spread the flow over a large number of different paths. Motivated by routing problems arising in real-life applications, such as, e. g., telecommunication, unsplittable flows have moved into the focus of research. Here, the demand of each commodity may not be split but has to be sent along a single path.In this paper, a generalization of this problem is studied. In the considered flow model, a commodity can be split into a bounded number of chunks which can then be routed on different paths. In contrast to classical (splittable) flows and unsplittable flows, already the single-commodity case of this problem is NP-hard and even hard to approximate. We present approximation algorithms for the single- and multicommodity case and point out strong connections to unsplittable flows. Moreover, results on the hardness of approximation are presented. It particular, we show that some of our approximation results are in fact best possible, unless P=NP.

Traditionally, flows over time are solved in time-expanded networks which contain one copy of the original network for each discrete time step. While this method makes available the whole algorithmic toolbox developed for static flows, its main and often fatal drawback is the enormous size of the time-expanded network. In particular, this approach usually does not lead to efficient algorithms with running time polynomial in the input size since the size of the time-expanded network is only pseudo-polynomial.We present two different approaches for coping with this difficulty. Firstly, inspired by the work of Ford and Fulkerson on maximal s-t-flows over time (or 'maximal dynamic s-t-flows'), we show that static, lengthbounded flows lead to provably good multicommodity flows over time.These solutions not only feature a simple structure but can also be computed very efficiently in polynomial time.Secondly, we investigate 'condensed' time-expanded networks which rely on a rougher discretization of time. Unfortunately, there is a natural tradeoff between the roughness of the discretization and the quality of the achievable solutions. However, we prove that a solution of arbitrary precision can be computed in polynomial time through an appropriate discretization leading to a condensed time expanded network of polynomial size. In particular, this approach yields a fully polynomial time approximation scheme for the quickest multicommodity flow problem and also for more general problems.

We consider the problem of scheduling unrelated parallel machines subject to release dates so as to minimize the total weighted completion time of jobs. The main contribution of this paper is a provably good convex quadratic programming relaxation of strongly polynomial size for this problem. The best previously known approximation algorithms are based on LP relaxations in time- or interval-indexed variables. Those LP relaxations, however, suffer from a huge number of variables. As a result of the convex quadratic programming approach we can give a very simple and easy to analyze 2-approximation algorithm which can be further improved to performance guarantee 3/2 in the absence of release dates. We also consider preemptive scheduling problems and derive approximation algorithms and results on the power of preemption which improve upon the best previously known results for these settings. Finally, for the special case of two machines we introduce a more sophisticated semidefinite programming relaxation and apply the random hyperplane technique introduced by Goemans and Williamson for the MaxCut problem; this leads to an improved 1.2752-approximation.

In network scheduling a set of jobs must be scheduled on unrelated parallel processors or machines which are connected by a network. Initially, each job is located on some machine in the network and cannot be started on another machine until sufficient time elapses to allow the job to be transmitted there. This setting has applications, e. g., in distributed multi-processor computing environments and also in operations research; it can be modeled by a standard parallel machine environment with machine-dependent release dates.We consider the objective of minimizing the total weighted completion time.The main contribution of this paper is a provably good convex quadratic programming relaxation of strongly polynomial size for this problem. Until now, only linear programming relaxations in time- or interval-indexed variables have been studied. Those LP relaxations, however, suffer from a huge number of variables. In particular, the best previously known relaxation is of exponential size and can therefore not be solved exactly in polynomial time. As a result of the convex quadratic programming approach we can give a very simple and easy to analyze randomized 2-approximation algorithm which slightly improves upon the best previously known approximation result. Furthermore, we consider preemptive variants of network scheduling and derive approximation results and results on the power of preemption which improve upon the best previously known results for these settings.

In this paper, we provide a new class of randomized approximation algorithms for scheduling problems by directly interpreting solutions to so-called time-indexed LPs as probabilities. The most general model we consider is scheduling unrelated parallel machines with release dates (or even network scheduling) so as to minimize the average weighted completion time. The crucial idea for these multiple machine problems is not to use standard list scheduling but rather to assign jobs randomly to machines (with probabilities taken from an optimal LP solution) and to perform list scheduling on each of them.

Three characteristics encountered frequently in real-world machine scheduling are jobs released over time, precedence constraints between jobs, and average performance optimization. The general constrained one-machine scheduling problem to minimize the average weighted completion time not only captures these features, but also is an important building block for more complex problems involving multiple machines.

Due to its obvious practical relevance, the Time-Cost Tradeoff Problem has attracted the attention of many researchers over the last forty years. While the Linear Time-Cost Tradeoff Problem can be solved in polynomial time, its discrete variant is known to be NP-hard. We present the first approximation algorithms for the Discrete Time-Cost Tradeoff Problem. Specifically, given a fixed budget we consider the problem of finding a shortest schedule for a project. We give an approximation algorithm with performance ratio 3/2 for the class of projects where all feasible durations of activities are either 0, 1, or 2. We extend our result by giving approximation algorithms with performance guarantee O(log l), where l is the ratio of the maximum duration of any activity to the minimum nonzero duration of any activity. Finally, we discuss bicriteria approximation algorithms which compute schedules for a given deadline or budget such that both project duration and cost are within a constant factor of the duration and cost of an optimum schedule for the given deadline or budget.