The "Options" button on the main form produces a dialog box containing several options for the algorithm. The "Randomization Depth" and "Randomization Iter" options specify the depth and number of iterations for the randomization procedure used while building the initial solution. Randomization works as follows: In the first iteration, the best pairing of routes is chosen at each step. In the second iteration, one of the 2 best pairings is chosen at random at each step, and 'iter' solutions are built in this way. In the third iteration, one of the 3 best pairings is chosen at random at each step, and 'iter' solutions are built in this way. This continues until solutions are built choosing one of the 'depth' best pairings at each step. In total the procedure produces 'depth' x 'iter' solutions (possibly containing duplicates). The best solution is selected and improvement heuristics are run on it. Note that setting depth = iter = 1 corresponds to turning off the randomization process and running the standard Clarke-Wright algorithm to find an initial solution. In general, the smaller these parameters are, the quicker the algorithm will run but the poorer the solutions will be.
The next set of options allows you to choose which improvement heuristics to run (see Section 2). The final set of options concerns how Or-opt is performed. If the box labeled "Search all routes during Or-opt [3]" is checked, the algorithm performs a variation on the standard Or-opt (which considers moving a group of 3 cities to another position on the same route) in which the group of cities may be moved to a different route if it's cheaper to do so than to re-insert them on the same route. If the box is unchecked, the standard Or-opt procedure is performed. The next two options are the same, but applying to the 2- and 1-city Or-opt operations. In general, checking more options on this form will result in better solutions but longer run-times.