GRASP (greedy randomized adaptive search procedure) is a
metaheuristic in which each iteration is composed of two phases
(Feo & Resende, 1995): construction of a trial solution using a greedy
randomized heuristic, and then improvement of this solution by
local search