“…As the Skill VRP does not involve time window and capacity constraints, we use only the geographical information of Solomon's instances to generate three types of geographical data for Skill VRP instances, namely, R, C and RC, which represent the random, clustered and a mixed of random and clustered geographical setting, respectively. Similar to Cappanera et al (2011) and Schwarze and Voß (2012), we consider a skill set with three levels 1, 2 and 3, where skill 1 denotes the lowest level, and skill 3 the highest. Each task i ∈ C is associated with a skill requirement s i ∈ {1, 2, 3}, which must be fulfilled by a technician k ∈ K having a skill levelŝ k ≥ s i , whereŝ k ∈ {1, 2, 3}.…”