1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154

#[cfg(test)]
#[path = "../../tests/unit/solver/population_test.rs"]
mod population_test;

use crate::algorithms::nsga2::{select_and_rank, MultiObjective, Objective};
use crate::models::Problem;
use crate::solver::{Individual, Population, SOLUTION_ORDER_KEY};
use crate::utils::compare_floats;
use std::cmp::Ordering;
use std::sync::Arc;

/// A simple evolution aware implementation of [`Population`] trait with the the following
/// characteristics:
///
/// - sorting of individuals in population according their objective fitness using [`NSGA-II`] algorithm
/// - maintaining diversity of population based on their crowding distance
///
/// [`Population`]: ./trait.Population.html
/// [`NSGA-II`]: ../algorithms/nsga2/index.html
///
pub struct DominancePopulation {
    problem: Arc<Problem>,
    max_population_size: usize,
    individuals: Vec<Individual>,
}

/// Contains ordering information about individual in population.
#[derive(Clone, Debug)]
struct DominanceOrder {
    orig_index: usize,
    seq_index: usize,
    rank: usize,
    crowding_distance: f64,
}

impl DominancePopulation {
    /// Creates a new instance of `DominancePopulation`.
    ///
    /// * `problem` - a Vehicle Routing Problem definition.
    /// * `max_population_size` - a max size of population size.
    pub fn new(problem: Arc<Problem>, max_population_size: usize) -> Self {
        assert!(max_population_size > 0);

        Self { problem, max_population_size, individuals: vec![] }
    }
}

impl Population for DominancePopulation {
    fn add_all(&mut self, individuals: Vec<Individual>) -> bool {
        let was_empty = self.size() == 0;

        individuals.into_iter().for_each(|individual| {
            self.individuals.push(individual);
        });

        self.sort();
        self.ensure_max_population_size();
        self.is_improved(was_empty)
    }

    fn add(&mut self, individual: Individual) -> bool {
        let was_empty = self.size() == 0;

        self.individuals.push(individual);

        self.sort();
        self.ensure_max_population_size();
        self.is_improved(was_empty)
    }

    fn nth(&self, idx: usize) -> Option<&Individual> {
        self.individuals.get(idx)
    }

    fn cmp(&self, a: &Individual, b: &Individual) -> Ordering {
        self.problem.objective.total_order(a, b)
    }

    fn ranked<'a>(&'a self) -> Box<dyn Iterator<Item = (&Individual, usize)> + 'a> {
        Box::new(self.individuals.iter().map(|individual| (individual, Self::gen_dominance_order(individual).rank)))
    }

    fn size(&self) -> usize {
        self.individuals.len()
    }
}

impl DominancePopulation {
    fn sort(&mut self) {
        let objective = self.problem.objective.clone();

        // get best order
        let best_order = select_and_rank(self.individuals.as_slice(), self.individuals.len(), objective.as_ref())
            .into_iter()
            .zip(0..)
            .map(|(acc, idx)| DominanceOrder {
                orig_index: acc.index,
                seq_index: idx,
                rank: acc.rank,
                crowding_distance: acc.crowding_distance,
            })
            .collect::<Vec<_>>();

        assert_eq!(self.individuals.len(), best_order.len());

        // remember dominance order
        best_order.into_iter().for_each(|order| {
            self.individuals[order.orig_index].solution.state.insert(SOLUTION_ORDER_KEY, Arc::new(order));
        });

        // sort by best order
        self.individuals.sort_by(|a, b| {
            let a = Self::gen_dominance_order(a);
            let b = Self::gen_dominance_order(b);

            a.seq_index.cmp(&b.seq_index)
        });

        // deduplicate population
        self.individuals.dedup_by(|a, b| {
            let order_a = Self::gen_dominance_order(a);
            let order_b = Self::gen_dominance_order(b);

            if order_a.rank == order_b.rank {
                // TODO investigate why just using crowding distance here does not work
                let fitness_a = objective.objectives().map(|o| o.fitness(a));
                let fitness_b = objective.objectives().map(|o| o.fitness(b));

                fitness_a.zip(fitness_b).all(|(a, b)| compare_floats(a, b) == Ordering::Equal)
            } else {
                false
            }
        });
    }

    fn ensure_max_population_size(&mut self) {
        if self.individuals.len() > self.max_population_size {
            self.individuals.truncate(self.max_population_size);
        }
    }

    fn is_improved(&self, was_empty: bool) -> bool {
        was_empty
            || self
                .individuals
                .first()
                .map(Self::gen_dominance_order)
                .map_or(false, |dominance_order| dominance_order.orig_index != dominance_order.seq_index)
    }

    fn gen_dominance_order(individual: &Individual) -> &DominanceOrder {
        individual.solution.state.get(&SOLUTION_ORDER_KEY).and_then(|s| s.downcast_ref::<DominanceOrder>()).unwrap()
    }
}