Maritime inventory routing (MIR) problem is an academic name for a practical logistic problem that represents the routing or scheduling of vessels to carry product(s) between ports. Meanwhile, the product(s) inventory levels in these ports must remain between operational bounds during the entire planning horizon. This thesis focus on how to support decision on a real-life MIR problem faced by a Brazilian petroleum company. To do so, we structure a set of tests to compare different formulation from literature and identify which is more adherent to real problem. Due to computational complexity of the problem, we present an heuristic approach that provides reasonably good solutions when compared to deterministic mixed integer linear programming (MILP) formulations and reduces considerably the computational time of solving real-life instances. However, uncertainty events have great impact in the ship scheduling planning. Therefore, we propose a robust optimization approach that considers uncertainty in the time spent at ports in each ship visit. Our approach is able to determine the probability of infeasibility and the impact in the objective function for each level of robustness, helping to measure the uncertain aversion of the decision maker. Our experiments identified that, for a certain instance, varying the level of robustness one may reduce the probability of infeasibility from 87 per cent (of deterministic solution) to 2 per cent and it represents an increase in the transportation costs of about 13 per cent.