ParticleSwarmOptimization.h

Go to the documentation of this file.

/*
 GRT MIT License
 Copyright (c) <2012> <Nicholas Gillian, Media Lab, MIT>
 
 Permission is hereby granted, free of charge, to any person obtaining a copy of this software
 and associated documentation files (the "Software"), to deal in the Software without restriction,
 including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense,
 and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so,
 subject to the following conditions:
 
 The above copyright notice and this permission notice shall be included in all copies or substantial
 portions of the Software.
 
 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT
 LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
 IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
 WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
 SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 */

#ifndef GRT_PARTICLE_SWARM_OPTIMIZATION_HEADER
#define GRT_PARTICLE_SWARM_OPTIMIZATION_HEADER

#include "../../Util/GRTCommon.h"
#include "PSOParticle.h"

GRT_BEGIN_NAMESPACE

template<class PARTICLE_TYPE,class OBSERVATION_TYPE>
class ParticleSwarmOptimization{
    
public:
    ParticleSwarmOptimization() : infoLog("[ParticleSwarmOptimization]"), errorLog("[ERROR ParticleSwarmOptimization]") {
        initialized = false;
        minImprovement = 1.0e-10;
        maxIter = 500;
        maxNumIterNoChange = 10;
    }
    
    virtual ~ParticleSwarmOptimization(){
        globalBestX.clear();
        particles.clear();
    }
    
    PARTICLE_TYPE& operator[](const unsigned int &index){
        return particles[ index ];
    }
    
    virtual bool init(const unsigned int numParticles,const unsigned int K,const VectorFloat &xMin,const VectorFloat &xMax,const VectorFloat &propagationModel){
        
        initialized = false;
        
        if( K != xMin.size() || K != xMax.size() ) return false;
        
        this->K = K;
        this->xMin = xMin;
        this->xMax = xMax;
        this->propagationModel = propagationModel;
        
        //Clear any previous searches
        particles.clear();
        iterHistory.clear();
        globalBestXHistory.clear();
        
        //Reset the global variables
        globalBestCost = 0;
        globalBestX.clear();
        globalBestX.resize(K,0);
        
        //Initialize the new particles with random positions and velocities
        Random random;
        particles.resize(numParticles);
        typename Vector< PARTICLE_TYPE >::iterator pIter = particles.begin();
        for(pIter = particles.begin(); pIter != particles.end(); pIter++){
            pIter->setRandomSeed( random.getRandomNumberInt(0, 100000) );
            pIter->init(K,xMin,xMax);
        }
        
        //Store the initial distrubution
        iterHistory.push_back( particles );
        globalBestXHistory.push_back( globalBestX );
        
        //Flag that everything has been initialized
        initialized = true;
        
        return true;
    }
    
    virtual bool reset(){
        
        if( !initialized ) return false;
        
        typename Vector< PARTICLE_TYPE >::iterator pIter = particles.begin();
        for(pIter = particles.begin(); pIter != particles.end(); pIter++){
            pIter->reset();
        }
        
        return true;
    }
    
    virtual bool search(OBSERVATION_TYPE &observation){
        
        if( !initialized ) return false;

        //Propagate all the particles
        typename Vector< PARTICLE_TYPE >::iterator pIter;
        for(pIter = particles.begin(); pIter != particles.end(); pIter++){
            if( !pIter->propagate( propagationModel ) ){
                errorLog << __GRT_LOG__ << " Particle propagation failed!" << std::endl;
                return false;
            }
        }
        
        unsigned int iterCounter = 0;
        unsigned int numIterNoChangeCounter = 0;
        Float currentMaxima = 0;
        Float lastMaxima = 0;
        Float delta = 0;
        bool keepSearching = true;
        
        //Decrement the global best cost
        globalBestCost = 0;
        
        while( keepSearching ){
            currentMaxima = searchIteration( observation );
            
            iterHistory.push_back( particles );
            globalBestXHistory.push_back( globalBestX );
            
            delta = fabs( currentMaxima - lastMaxima );
            lastMaxima = currentMaxima;
            
            if( delta < minImprovement ){
                if( ++numIterNoChangeCounter >= maxNumIterNoChange ){
                    keepSearching = false;
                    infoLog << __GRT_LOG__ << " Reached no change limit, stopping search at iteration: " << iterCounter << " with a change of: " << fabs( currentMaxima - lastMaxima ) << std::endl;
                }
            }else numIterNoChangeCounter = 0;
            
            if( ++iterCounter == maxIter ){
                keepSearching = false;
                infoLog << __GRT_LOG__ << " Max iteration reached, stopping search at iteration: " << iterCounter << std::endl;
            }
        }
        
        //Compute the final state estimation
        finalX = globalBestX;
        
        return true;
    }
    
    virtual Float searchIteration(OBSERVATION_TYPE &observation){
        
        if( !initialized ) return 0;
        
        unsigned int index = 0;
        unsigned int bestIndex = 0;
        Float currentBestMaxima = 0;
        Float epsilon = 0;
        typename Vector< PARTICLE_TYPE >::iterator pIter;
        
        //Compute the cost for each particle, tracking the best cost of all the particles in the swarm
        for(pIter = particles.begin(); pIter != particles.end(); pIter++){
            
            //Evaluate the current particle
            epsilon = pIter->evaluate(observation);

            //Check to see if this is the best
            if( epsilon >  currentBestMaxima ){
                currentBestMaxima = epsilon;
                bestIndex = index;
            }
            
            index++;
        }
        
        //Check to see if we need to update the global cost and position
        if( currentBestMaxima > globalBestCost ){
            globalBestCost = currentBestMaxima;
            globalBestX = particles[ bestIndex ].x;
        }

        //Update the position and velocity of all of the particles
        for(pIter = particles.begin(); pIter != particles.end(); pIter++){
            pIter->update( globalBestX );
        }
        
        return currentBestMaxima;
    }
    
    bool setPropagationModel( const VectorFloat &propagationModel ){
        
        if( !initialized ) return false;
        
        this->propagationModel = propagationModel;
        
        return true;
    }

    bool initialized;   
    unsigned int K;     
    Float minImprovement;
    unsigned int maxIter;
    unsigned int maxNumIterNoChange;
    Float globalBestCost;  
    VectorFloat finalX;  
    VectorFloat globalBestX;   
    VectorFloat xMin;  
    VectorFloat xMax;  
    VectorFloat propagationModel;  
    Vector< PARTICLE_TYPE > particles;  
    Vector< VectorFloat > globalBestXHistory;  
    Vector< Vector< PARTICLE_TYPE > > iterHistory;  
    
    InfoLog infoLog;
    ErrorLog errorLog;
};
    
GRT_END_NAMESPACE

#endif //GRT_PARTICLE_SWARM_OPTIMIZATION_HEADER