A hybrid method for transportation with stochastic demand

In industrial transportation, the forecast demand at each destination may be affected by a number of factors. Consequently, a conventional transport plan often fails to match the reality, and the planned transport capacity is either insufficient to meet the demand or wastefully excessive. In this paper, we introduce a new algorithm to generate a minimal cost transport plan that meets a given level of reliability. The reliability of a candidate solution is measured through simulating each candidate solution against a large number of scenarios. To search for reliable solutions, a genetic algorithm method is applied as an external loop. The minimal transport cost is achieved through a deterministic optimisation algorithm. We show that this problem decomposition in principle enables the optimal solution of the original non-deterministic problem to be found. Experimental results establish the practical usefulness of the proposed algorithm.