A user's guide to tabu search

We describe the main features of tabu search, emphasizing a perspective for guiding a user to understand basic implementation principles for solving combinatorial or nonlinear problems. We also identify recent developments and extensions that have contributed to increasing the efficiency of the method. One of the useful aspects of tabu search is the ability to adapt a rudimentary prototype implementation to encompass additional model elements, such as new types of constraints and objective functions. Similarly, the method itself can be evolved to varying levels of sophistication. We provide several examples of discrete optimization problems to illustrate the strategic concerns of tabu search, and to show how they may be exploited in various contexts. Our presentation is motivated by the emergence of an extensive literature of computational results, which demonstrates that a well-tuned implementation makes it possible to obtain solutions of high quality for difficult problems, yielding outcomes in some settings that have not been matched by other known techniques.The research of this paper has been supported in part by the joint Air Force Office of Scientific Research and Office of Naval Research Contract No. F49620-90-C-0033, at the University of Colorado and by the Swiss National Research Council Grant No. 20-27926.89.     

Categorized Assignment Scheduling: a Tabu Search Approach

The assignment problem (AP) and bottleneck assignment problem (BAP) are well studied in operational research literature. In this paper we consider two related problems which simultaneously generalize both AP and BAP. Unlike AP and BAP, these generalizations are strongly NP-complete. We propose two heuristics to solve these generalized problems: one based on a greedy principle and the other based on tabu search. Computational results are presented which show that these heuristics, when used together, produce good quality solutions. Our adaptation of tabu search also gives some new insight into the application of tabu search on permutation problems.     

An evolutionary and constructive approach to a crew scheduling problem in underground passenger transport

Operations management of subway systems is associated with combinatorial optimization problems (i.e. crew and train scheduling and rostering) which belong to the np-hard class of problems. Therefore, they are generally solved heuristically in real situations. This paper considers the problem of duty generation, i.e. identifying an optimal trips set that the conductors should complete in one workday. With regard to operational and labor conditions, the problem is to use the lowest possible number of conductors and minimize total idle time between trips. The problem is modeled and solved using a constructive hybrid approach, which has the advantage of visualizing a solution construction similar to the manual approach typically used. Our approach takes advantage of the benefits offered by evolutionary methods, which store a candidate solutions population in each stage, thus controlling the combinatorial explosion of possible solutions. The results thus obtained for problems similar to those that are solved manually in the Santiago Metro System were compared with two alternative approaches, based on tabu search and a greedy method. The hybrid method produced solutions with the minimum number of duties in six of the ten problems solved. However, the tabu search method provided better results in terms of idle time than either the hybrid method or the greedy method.     

Path assignment for call routing: An application of tabu search

We describe an implementation of tabu search that solves the path assignment problem, which is the problem of routing video data through an undercapacitated telecommunications network. Two versions of the tabu search were studied. Our results compare very favourably with those from other methods.This research was supported by a grant from the Colorado Advanced Software Institute. The work of Jennifer Ryan was also partially supported by the Air Force Office of Scientific Research and the Office of Naval Research Contract #F49620-90-C-0033.     

Evolutionary Approaches to DNA Sequencing with Errors

In the paper, two evolutionary approaches to the general DNA sequencing problem, assuming both negative and positive errors in the spectrum, are compared. The older of them is based on the idea of genetic approach and is enhanced by a greedy algorithm. The newly proposed algorithm combines the tabu search and the scatter search methods. After conducting experiments with random and coding DNA sequences, our results suggest that the tabu and scatter search algorithm finds solutions of higher quality and more reliably than the genetic algorithm.     

A tabu search procedure for multicommodity location/allocation with balancing requirements

We propose a tabu search heuristic for the location/allocation problem with balancing requirements. This problem typically arises in the context of the medium term management of a fleet of containers of multiple types, where container depots have to be selected, the assignment of customers to depots has to be established for each type of container, and the interdepot container traffic has to be planned to account for differences in supplies and demands in various zones of the geographical territory served by a container shipping company. It is modeled as a mixed integer program, which combines zero-one location variables and a multicommodity network flow structure. Extensive computational results on a set of benchmark problems and comparisons with an efficient dual ascent procedure are reported. These show that tabu search is a competitive approach for this class of problems.     

A tabu search approach for solving a difficult forest harvesting machine location problem

This paper deals with two main problems in forest harvesting. The first is that of selecting the locations for the machinery to haul logs from the points where they are felled to the roadside. The second consists in designing the access road network connecting the existing road network with the points where machinery is installed. Their combination induces a very important and difficult problem to solve in forest harvesting. It can be formulated as a combination of two difficult optimization problems: a plant location problem and a fixed charge network flow problem. In this paper, we propose a solution approach based on tabu search. The proposed heuristic includes several enhancements of the basic tabu search framework. The main difficulty lies in evaluating neighboring solutions, which involves decisions related to location of machinery and to road network arcs. Hence, the neighborhood is more complex than in typical applications of metaheuristics. Minimum spanning tree algorithms and Steiner tree heuristics are used to deal with this problem. Numerical results indicate that the heuristic approach is very attractive and leads to better solutions than those provided by state-of-the-art integer programming codes in limited computation times, with solution times significantly smaller. The numerical results do not vary too much when typical parameters such as the tabu tenure are modified, except for the dimension of neighborhood.     

Computational experience with a group theoretic integer programming algorithm

This paper gives specific computational details and experience with a group theoretic integer programming algorithm. Included among the subroutines are a matrix reduction scheme for obtaining group representations, network algorithms for solving group optimization problems, and a branch and bound search for finding optimal integer programming solutions. The innovative subroutines are shown to be efficient to compute and effective in finding good integer programming solutions and providing strong lower bounds for the branch and bound search.This research was supported in part by the U.S. Army Research Office (Durham) under contract no. DAHC04-70-C-0058. This paper is not an official National Bureau of Economic Research publication.     

A polynomial oracle-time algorithm for convex integer minimization

In this paper we consider the solution of certain convex integer minimization problems via greedy augmentation procedures. We show that a greedy augmentation procedure that employs only directions from certain Graver bases needs only polynomially many augmentation steps to solve the given problem. We extend these results to convex N-fold integer minimization problems and to convex 2-stage stochastic integer minimization problems. Finally, we present some applications of convex N-fold integer minimization problems for which our approach provides polynomial time solution algorithms.     

A multi-class multi-level capacitated lot sizing model

When demand loading is higher than available capacity, it takes a great deal of effort for a traditional MRP system to obtain a capacity-feasible production plan. Also, the separation of lot sizing decisions and capacity requirement planning makes the setup decisions more difficult. In a practical application, a production planning system should prioritize demands when allocating manufacturing resources. This study proposes a planning model that integrates all MRP computation modules. The model not only includes multi-level capacitated lot sizing problems but also considers multiple demand classes. Each demand class corresponds to a mixed integer programming (MIP) problem. By sequentially solving the MIP problems according to their demand class priorities, this proposed approach allocates finite manufacturing resources and generates feasible production plans. In this paper we experiment with three heuristic search algorithms: (1) tabu search; (2) simulated annealing, and (3) genetic algorithm, to solve the MIP problems. Experimental designs and statistical methods are used to evaluate and analyse the performance of these three algorithms. The results show that tabu search and simulated annealing perform best in the confirmed order demand class and forecast demand class, respectively.     

Block truncated-Newton methods for parallel optimization

Truncated-Newton methods are a class of optimization methods suitable for large scale problems. At each iteration, a search direction is obtained by approximately solving the Newton equations using an iterative method. In this way, matrix costs and second-derivative calculations are avoided, hence removing the major drawbacks of Newton's method. In this form, the algorithms are well-suited for vectorization. Further improvements in performance are sought by using block iterative methods for computing the search direction. In particular, conjugate-gradient-type methods are considered. Computational experience on a hypercube computer is reported, indicating that on some problems the improvements in performance can be better than that attributable to parallelism alone.Partially supported by Air Force Office of Scientific Research grant AFOSR-85-0222.Partially supported by National Science Foundation grant ECS-8709795, co-funded by the U.S. Air Force Office of Scientific Research.     

Applying tabu search to the job-shop scheduling problem

In this paper, we apply the tabu-search technique to the job-shop scheduling problem, a notoriously difficult problem in combinatorial optimization. We show that our implementation of this method dominates both a previous approach with tabu search and the other heuristics based on iterative improvements.Partially supported by research contracts MPI 40% and 60% of the Italian Ministry of University and Scientific Research.     

一种改进的禁忌搜索算法及其在选址问题中的应用

本文研究了选址问题中无容量限制的p-中值问题,在Rolland等人提出的有效禁忌搜索算法基础上,提出了一种以目标函数变化量作为评价函数的改进禁忌搜索算法,并进行了理论分析,然后将其与有效禁忌搜索算法作了性能比较.通过比较三个公共测试数据集的计算结果,验证了本文提出的禁忌搜索算法的可行性和有效性.     

Incomplete Dynamic Backtracking for Linear Pseudo-Boolean Problems

Many combinatorial problems can be modeled as 0/1 integer linear programs. Problems expressed in this form are usually solved by Operations Research algorithms, but good results have also been obtained using generalised SAT algorithms based on backtracking or local search, after transformation to pseudo-Boolean form. A third class of SAT algorithm uses non-systematic backtracking to combine constraint propagation with local search-like scalability, at the cost of completeness. This paper describes such an algorithm for pseudo-Boolean models. Experimental results on a variety of problems are encouraging, in some cases yielding improved solutions or performance compared to previous algorithms.     

Serial and parallel simulated annealing and tabu search algorithms for the traveling salesman problem

This paper describes serial and parallel implementations of two different search techniques applied to the traveling salesman problem. A novel approach has been taken to parallelize simulated annealing and the results are compared with the traditional annealing algorithm. This approach uses abbreviated cooling schedule and achieves a superlinear speedup. Also a new search technique, called tabu search, has been adapted to execute in a parallel computing environment. Comparison between simulated annealing and tabu search indicate that tabu search consistently outperforms simulated annealing with respect to computation time while giving comparable solutions. Examples include 25, 33, 42, 50, 57, 75 and 100 city problems.     

Scheduling ambulance crews for maximum coverage

This paper addresses the problem of scheduling ambulance crews in order to maximize the coverage throughout a planning horizon. The problem includes the subproblem of locating ambulances to maximize expected coverage with probabilistic response times, for which a tabu search algorithm is developed. The proposed tabu search algorithm is empirically shown to outperform previous approaches for this subproblem. Two integer programming models that use the output of the tabu search algorithm are constructed for the main problem. Computational experiments with real data are conducted. A comparison of the results of the models is presented.     

A primogenitary linked quad tree approach for solution storage and retrieval in heuristic binary optimization

A data structure, called the primogenitary linked quad tree (PLQT), is used to store and retrieve solutions in heuristic solution procedures for binary optimization problems. Two ways are proposed to use integer vectors to represent solutions represented by binary vectors. One way is to encode binary vectors into integer vectors in a much lower dimension and the other is to use the sorted indices of binary variables with values equal to 0 or equal to 1. The integer vectors are used as composite keys to store and retrieve solutions in the PLQT. An algorithm processing trial solutions for insertion into or retrieval from the PLQT is developed. Examples are provided to demonstrate the way the algorithm works. Another algorithm traversing the PLQT is also developed. Computational results show that the PLQT approach takes only a very tiny portion of the CPU time taken by a linear list approach for the same purpose for any reasonable application. The CPU time taken by the PLQT managing trial solutions is negligible as compared to that taken by a heuristic procedure for any reasonably hard to solve binary optimization problem, as shown in a tabu search heuristic procedure for the capacitated facility location problem. Compared to the hashing approach, the PLQT approach takes the same or less amount of CPU time but much less memory space while completely eliminating collision.     

Multi-coloring and job-scheduling with assignment and incompatibility costs

Consider a scheduling problem (P) which consists of a set of jobs to be performed within a limited number of time periods. For each job, we know its duration as an integer number of time periods, and preemptions are allowed. The goal is to assign the required number of time periods to each job while minimizing the assignment and incompatibility costs. When a job is performed within a time period, an assignment cost is encountered, which depends on the involved job and on the considered time period. In addition, for some pairs of jobs, incompatibility costs are encountered if they are performed within common time periods. (P) can be seen as an extension of the multi-coloring problem. We propose various solution methods for (P) (namely a greedy algorithm, a descent method, a tabu search and a genetic local search), as well as an exact approach. All these methods are compared on different types of instances.     

Bus driver duty optimization using an integer programming and evolutionary hybrid algorithm

This paper describes the details of a successful application where an integer programming and evolutionary hybrid algorithm was used to solve a bus driver duty optimization problem. The task is NP-hard, therefore theoretically optimal solutions can only be calculated for very small problem instances. Our aim is to obtain solutions of good quality within reasonable time limits. We first applied an integer programming approach to a set partitioning problem. The model was solved with a column generation algorithm in a branch and bound scheme. In order to solve larger real-life problems, we have combined the integer programming method with a greedy 1+1 steady state evolutionary algorithm. The resulting hybrid algorithm was capable of providing near-optimal solutions within reasonable timescales to larger instances of the bus driver scheduling problem. We present the results and running times of our algorithm in detail, as well as possible directions of future improvements.     

Multi-objective tabu search using a multinomial probability mass function

A tabu search approach to solve multi-objective combinatorial optimization problems is developed in this paper. This procedure selects an objective to become active for a given iteration with a multinomial probability mass function. The selection step eliminates two major problems of simple multi-objective methods, a priori weighting and scaling of objectives. Comparison of results on an NP-hard combinatorial problem with a previously published multi-objective tabu search approach and with a deterministic version of this approach shows that the multinomial approach is effective, tractable and flexible.