The initial solution is constructed using an extended version of the cheapest insertion heuristic I1 in [36]. The constructor
searches over several values for the most critical parameters to get a higher quality initial solution. Planner also has several
extensions that deal with constraints and objectives of complex real-world problems, but these do not apply in our case.
To be more precise about the VND intensification procedure, SPIDER Planner applies a set of well-known local search
neighbourhood operators in sequence. These have been extended to cater for the extended, rich VRP model. In our
investigation, the following subset of available, extended operators is used: Relocate [34], 2-opt* [30], CROSS-exchange,
an opportunistic variant of CROSS-exchange [38], 2-opt [10], 3-opt [14], and Tour Depletion. The operators were applied
with the best accept strategy, i.e., the best improving move is first identified and then accepted.