Introduction
The R Optimization library provides several optimization-oriented tools.
Main classes
The main types of classes are:
- A complete implementation of the PROMETHEE multi-criteria decision method.
- A generic Genetic Algorithm (GA).
- A complete implementation of the Grouping Genetic Algorithms (GGA).
- A complete implementation of the Similarity-based Grouping Genetic Algorithm (SGGA).
- A complete implementation of the Nearest Neighbors Grouping Genetic Algorithm (NNGGA).
- A complete implementation of the Hierarchical Genetic Algorithms (HGA).
- A complete implementation of the 2D Genetic Algorithms (2DGA).
PROMETHEE
PROMETHEE is multi-criteria decision method which allows to classify of solutions depending of different criteria, each criterion having a given weight. One of the advantages of the PROMETHEE method is that it allows to compare criteria with different scales.
The RPromKernel class provides a kernel that implements the PROMETHEE method. It manages a set of solutions that instances of the RPromSol class. The different criteria are represented by classes inheriting from RPromCriterion. Actually, the only existing type of criteria is implemented in the RPromLinearCriterion class. Once the solutions and the criteria defined, the methods RPromKernel::Assign allows to associate a value for a given pair of solution and criteria.
Genetic Algorithms (GA)
A set of classes provides a representation for a generic algorithm. These classes are implemented as templates since the genetic algorithms are a generic approach to solve optimization problem.
To implement a particular genetic algorithm, at least, three classes must be created. A first one inheriting from RChromo and representing the chromosome (i.e. the coding used by the particular GA). A second one inheriting from RInst. The third class must inherit from RThreadData and represent a set of data used by the GA and thread-dependent. In particular, when implementing a given operator (such as the crossover), it is sometimes necessary to used specific structures of data. By using the mechanism of RThreadData, these structures will be duplicated of the GA uses multiple threads (which is currently not implemented).
2D-Placement Genetic Algorithms (GA2D)
A set of classes provides a implementation of a generic 2D-placement genetic algorithm. Its aim is to place a set of objects on an area in order to minimize the total area occupied and the total distances of the corresponding connections.
Each object is represented as an instance of the RObj2D class. Its defines several connectors (RObj2DConnector), each one having multiple pins (RObj2DPin). Moreover, each object may have several configurations (RObj2DConfig), each one defining the possible pins (RObj2DConfigPin) of each connector (RObj2DConfigConnector). Finally, a RGeoInfo class represent a particular configuration (RObj2DConfig) of an object (RObj2D) placed at a given position.
A connection is represented by an instance of the RConnection class. It is represented by a container of RObj2DConnector. The RGeoInfoConnection class represents the instance of that connection. It is a container of RGeoInfoPin representing association of a pin and a geometric information.
Each chromosome (RChromo2D) implements a particular layout (RLayout) of the objects. All heuristics inherit from RPlacementHeuristic which place a list of unplaced object on a given layout.