6. Conclusions
This paper has presented a new iterative MILP-based algorithm
for coping with large-scale ship routing and scheduling problems.
The mathematical model, which is based on the notion of general
precedence, utilizes a continuous time domain representation
and is able to optimize multiple objectives when triangle inequality
violations are introduced in distance matrixes. However, such
exact optimization approach remains computationally efficiently
only for small-to-medium size problems. In order to overcome this
limitation, an iterative procedure was derived by embedding the
rigorous formulation within heuristic rules to effectively find feasible
and near-optimal solutions for large-scale instance of the
problem within a short computational time. The procedure is
based on a systematic decomposition strategy that, by solving
highly constrained versions of the model on every iteration, allows
that the number of decisions be maintained at a reasonable level
by fixing a set of binary variables.
The MILP model was first validated by solving a series of small
instances deriving from a real-world case study faced by a
multi-national shipping company operating a fleet of
multi-parcel chemical tankers. The results obtained were compared
with others presented by two authors from the literature.
Comparison between results reveals that the general precedence
based model has a better computational performance than the
time-slots based model proposed to solve the same problem
instances. Despite this, the exact approach has not converged and
the MIP solver terminated because the memory capacity was
exceeded when the full problem, involving 10 ships, 36 ports,
and 79 potential cargos, is considered. After that, the iterative
algorithm was applied to solve the same full problem instance. A
convergence to a near-optimal solution was achieved in only
764 s of CPU time. Such computational performance significantly
overcomes these ones achieved by other algorithms presented in
the literature. Moreover, the new schedule improves profits by
approximately 40% with regards to actually used by the company