Abstract: Routing problems are naturally formulated by enumerating possible routes and combining them to derive an improved solution. This combination is done by solving a set partitioning problem that assures the set of chosen routes to visit all routing customers exactly once. Column generation techniques allow implicitly considering a huge number of routes and, when searching for an optimal solution, to consider all possible ones. This talk presents a real world application of column generation to the crew transportation from and to offshore oil rigs. The resulting algorithm solves an initial mixed integer program, that is followed by a column generation phase and finally executes a post processing. Details such as the allocation of the helicopter pilot lunch time are considered. This logistic deals daily with over 2000 passengers and more 40 helicopters of different types. The main goal is to increase the global safety by minimizing the number of offshore landings. Next, the talk proceeds to showing the key elements of state-of-the-art algorithms to optimally solve the classical capacitated vehicle routing problem (CVRP). In particular, the effect of the addition of bound improving polyhedral cuts to the set partitioning formulation on the subproblem, known as pricing problem, of finding improving routes/columns. An efficient resolution of the pricing all along the resulting branch-cut-and-price algorithm run is key to its overall performance. The careful choice of the columns to price and of the cuts to consider may allow solving to optimality CVRP instances of up to 360 customers.
Novo Seminário CIO