lemon/csp.h
author deba
Wed, 08 Oct 2008 09:17:01 +0000
changeset 2624 dc4dd5fc0e25
parent 2487 568ff3572a96
permissions -rw-r--r--
Bug fixes is HaoOrlin and MinCostArborescence

MinCostArborescence
- proper deallocation
HaoOrlin
- the target needn't to be the last in its bucket
- proper size of container (if each node starts in different buckets initially)
     1 /* -*- C++ -*-
     2  *
     3  * This file is a part of LEMON, a generic C++ optimization library
     4  *
     5  * Copyright (C) 2003-2008
     6  * Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport
     7  * (Egervary Research Group on Combinatorial Optimization, EGRES).
     8  *
     9  * Permission to use, modify and distribute this software is granted
    10  * provided that this copyright notice appears in all copies. For
    11  * precise terms see the accompanying LICENSE file.
    12  *
    13  * This software is provided "AS IS" with no warranty of any kind,
    14  * express or implied, and with no claim as to its suitability for any
    15  * purpose.
    16  *
    17  */
    18 
    19 #ifndef LEMON_CSP_H
    20 #define LEMON_CSP_H
    21 
    22 ///\ingroup approx
    23 ///\file
    24 ///\brief Algorithm for the Resource Constrained Shortest Path problem.
    25 ///
    26 
    27 #include <lemon/list_graph.h>
    28 #include <lemon/graph_utils.h>
    29 #include <lemon/error.h>
    30 #include <lemon/maps.h>
    31 #include <lemon/tolerance.h>
    32 #include <lemon/dijkstra.h>
    33 #include <lemon/path.h>
    34 #include <lemon/counter.h>
    35 namespace lemon {
    36 
    37   ///\ingroup approx
    38   ///
    39   ///\brief Algorithms for the Resource Constrained Shortest Path Problem
    40   ///
    41   ///The Resource Constrained Shortest (Least Cost) Path problem is the
    42   ///following. We are given a directed graph with two additive weightings
    43   ///on the edges, referred as \e cost and \e delay. In addition,
    44   ///a source and a destination node \e s and \e t and a delay
    45   ///constraint \e D is given. A path \e p is called \e feasible
    46   ///if <em>delay(p)\<=D</em>. Then, the task is to find the least cost
    47   ///feasible path.
    48   ///
    49   template<class Graph,
    50 	   class CM=typename Graph:: template EdgeMap<double>,
    51 	   class DM=CM>
    52   class ConstrainedShortestPath 
    53   {
    54   public:
    55     
    56     GRAPH_TYPEDEFS(typename Graph);
    57     
    58     typedef SimplePath<Graph> Path;
    59     
    60   private:
    61     
    62     const Graph &_g;
    63     Tolerance<double> tol;
    64 
    65     const CM &_cost;
    66     const DM &_delay;
    67 
    68     class CoMap 
    69     {
    70       const CM &_cost;
    71       const DM &_delay;
    72       double _lambda;
    73     public:
    74       typedef typename CM::Key Key;
    75       typedef double Value;
    76       CoMap(const CM &c, const DM &d) :_cost(c), _delay(d), _lambda(0) {}
    77       double lambda() const { return _lambda; }
    78       void lambda(double l)  { _lambda=l; }
    79       Value operator[](Key &e) const 
    80       {
    81 	return _cost[e]+_lambda*_delay[e];
    82       }
    83     };
    84 
    85     CoMap _co_map;
    86     
    87     
    88     Dijkstra<Graph, CoMap> _dij;
    89 
    90   public:
    91     
    92     /// \brief Constructor
    93     
    94     ///Constructor
    95     ///
    96     ConstrainedShortestPath(const Graph &g, const CM &ct, const DM &dl)
    97       : _g(g), _cost(ct), _delay(dl),
    98 	_co_map(ct, dl), _dij(_g,_co_map) {}
    99     
   100 
   101     ///Compute the cost of a path
   102     double cost(const Path &p) const
   103     {
   104       double s=0;
   105       //      Path r;  
   106       for(typename Path::EdgeIt e(p);e!=INVALID;++e) s+=_cost[e];
   107       return s;
   108     }
   109 
   110     ///Compute the delay of a path
   111     double delay(const Path &p) const
   112     {
   113       double s=0;
   114       for(typename Path::EdgeIt e(p);e!=INVALID;++e) s+=_delay[e];
   115       return s;
   116     }
   117     
   118     ///Runs the LARAC algorithm
   119     
   120     ///This function runs a Lagrange relaxation based heuristic to find
   121     ///a delay constrained least cost path.
   122     ///\param s source node
   123     ///\param t target node
   124     ///\param delta upper bound on the delta
   125     ///\retval lo_bo a lower bound on the optimal solution
   126     ///\return the found path or an empty 
   127     Path larac(Node s, Node t, double delta, double &lo_bo) 
   128     {
   129       double lambda=0;
   130       double cp,cq,dp,dq,cr,dr;
   131       Path p;
   132       Path q;
   133       Path r;
   134       {
   135 	Dijkstra<Graph,CM> dij(_g,_cost);
   136 	dij.run(s,t);
   137 	if(!dij.reached(t)) return Path();
   138 	p=dij.path(t);
   139 	cp=cost(p);
   140 	dp=delay(p);
   141       }
   142       if(delay(p)<=delta) return p;
   143       {
   144 	Dijkstra<Graph,DM> dij(_g,_delay);
   145 	dij.run(s,t);
   146 	q=dij.path(t);
   147 	cq=cost(q);
   148 	dq=delay(q);
   149       }
   150       if(delay(q)>delta) return Path();
   151       while (true) {
   152 	lambda=(cp-cq)/(dq-dp);
   153 	_co_map.lambda(lambda);
   154 	_dij.run(s,t);
   155 	r=_dij.path(t);
   156 	cr=cost(r);
   157 	dr=delay(r);
   158 	if(!tol.less(cr+lambda*dr,cp+lambda*dp)) {
   159 	  lo_bo=cq+lambda*(dq-delta);
   160 	  return q;
   161 	}
   162 	else if(tol.less(dr,delta)) 
   163 	  {
   164 	    q=r;
   165 	    cq=cr;
   166 	    dq=dr;
   167 	  }
   168 	else if(tol.less(delta,dr))
   169 	  {
   170 	    p=r;
   171 	    cp=cr;
   172 	    dp=dr;
   173 	  }
   174 	else return r;
   175       }
   176     }
   177   };
   178   
   179 
   180 } //END OF NAMESPACE LEMON
   181 
   182 #endif