Sie sind auf Seite 1von 2

1.

1 Motivation A combinatorial problem is the problem of finding an object with some desired property among an infinite set of possible alternatives. Many problems from industry exhibit a combinatorial nature. An example is the optimal routing of trucks to deliver goods from a depot to customers. There are many alternatives to distribute the goods among the trucks, and for each such distribution there are many alternative routes for each individual truck. Moreover, often we are restricted to deliver goods only within a certain time frame for each customer. This makes the search for an optimal solution even harder, because there may only be a few optimal solutions, respecting the time frames, among a huge set of possible alternatives. To solve combinatorial problems, we cannot simply consider all exponentially many possible alternatives. ome combinatorial problems are solvable by an algorithm whose running time is bounded by a polynomial in the si!e of the representation of the problem. These problems are considered to be efficiently solvable, and are said to belong to the class ". #or other problems such method is not known to exist and they are classified as follows. $f we can determine in polynomial time whether or not a particular alternative is a solution to a certain problem, the problem is said to be in the class %". %ote that all problems in " are also in %". $f a problem is in %" and moreover every other problem in %" can be transformed to this problem in polynomial time, the problem is said to be %"&complete. %"&complete problems are the hardest problems in %". $n this thesis we focus on solution methods for %"&complete combinatorial problems. everal solution methods have been proposed to solve combinatorial problems faster. 'owever, many real&life problems are still (yet) unsolvable, even when such techni*ues are applied, so further research is necessary. $n this thesis we consider techni*ues from operations research and constraint programming to model and solve combinatorial problems.

+perations ,esearch The field of operations research, sometimes also called management science, became an independent discipline in the late 1-./0s, although its foundations were laid earlier. The emergence of this discipline was motivated by the need to efficiently plan military operations during world war $$. 'ence the name 1operations2 research. A proper definition for this field, including all of its still growing branches, is hard to provide. According to the $nstitute for +perations ,esearch and the Management ciences13 +perations ,esearch and the Management ciences are the professional disciplines that deal with the application of information technology for informed decision&making.2 'owever, this definition may not be sufficiently specific to distinguish it from related fields. $n particular, constraint programming could also be viewed as a part of operations research, while we wish to distinguish the two in this research.

$nstead of following a definition, we will use the term 1operations research2 to specify a particular set of methods and solution techni*ues for combinatorial problems. This set includes for example techni*ues from graph theory, (integer) linear programming and semi definite programming. Many of these methods can be characteri!ed by the fact that they provide efficient solution techni*ues for specific problems. Another common feature of these methods is that they are particularly useful for optimi!ation problems. 4onsider for example linear programming. $n linear programming, a problem needs to be represented by linear e*uations (or ine*ualities) on continuous variables. 5iven such representation, linear programming can solve the problem in polynomial time. $f the problem cannot be expressed by linear e*uations or ine*ualities on continuous variables, we cannot apply the efficient linear programming algorithms. This is often the case, for example because some of the problem variables should be integral. Many operations research techni*ues have proven to be very useful in practice. #or many real&life problems, operations research provides the method of choice, in particular those problems that match the above characteri!ations. As described above, 6e propose to exploit operations research techni*ues in both components.

Das könnte Ihnen auch gefallen