R/applyLocalSearch.R

In jakobbossek/dynvrp: EMOA for a dynamic Orienteering Problem

#' Given an individual and an instance, this function performs
#' local search (focused on tour length minimization) accounting
#' for partial tours which are already driven.
#'
#' @param ind [\code{VRPIndividual}]\cr
#'   Individual.
#' @param instance [\code{Network}]\cr
#'   Netgen network.
#' @param more.args [\code{list}]\cr
#'   Further parameter passed down to local search procedure.
#' @param ... [any]\cr
#'   Not used at the moment.
#' @return Modified individual.
applyLocalSearch = function(ind, instance, more.args, ...) {
  for (v in seq_len(ind$n.vehicles)) {
    active.nodes = which(ind$b == 1L & ind$it != 1 & ind$v == v)
    idx.tour = sort(which(ind$t %in% active.nodes))

    # we need to consider the initially driven tour,
    # i.e., search for Hamiltionian path from last customer of
    # initial tour to end depot
    n.init = length(ind$init.tours[[v]])
    start.id = if (n.init == 0) 1L else (ind$init.tours[[v]][n.init] + 2L) # FIXME: do we need to add 2 to the latter!?

    ls.res = findHamiltonianPath(
      instance,
      active.nodes = sort(active.nodes) + 2L,
      start.id = start.id, dest.id = 2L,
      more.args)

    # nodes 1 and 2 are depots above. Hence, we need to recode to 1-based representation.
    ind$t[idx.tour] = ls.res - 2L
  }
  return(ind)
}