Let w min = min i ∈ I 1 S { w i }

Let w min = min i ∈ I 1 S { w i } and w max = max i ∈ I 1 S { w i } . In addition, let T = w min + α(w max − w min ) be a threshold value, where α ∈ (0, 1) is a parameter that controls the degree of greediness or randomness of the greedy procedure. We randomly select a facility i ∗from a restricted candidate list RCL that contains candidate facilities whose w i ≥ T . Facility i ? is added to the set of open facilities S . At the end of each iteration, the Covered set is updated and customers j ∈ J ( i ∗) are allocated to i ∗if they were not previously covered or if i ∗is more preferred than its previous assignment A ( j ). In addition, A ( j ) is updated for all j ∈ Covered . In other words, Covered is set to ? i ∈ (S∪ I 2 ) J(i ) , which is the set of all customers within the coverage radius of some open facility. For all j ∈ Covered, A ( j ) is set to arg max i ∈ (S∪ I 2 ) ∩ I(j) { g ij } , which means that each covered customer is allocated to their most preferred location among all lo- cations in S ∪ I 2 . The constructive phase ends when p facilities are selected. The randomized greedy procedure of the proposed GRASP heuristic is given in Algorithm 3.1 .