Online heuristics for unloading boxes off a gravity conveyor

This paper addresses the problem of minimising the number of moves to unload a set of boxes off a gravity conveyor by a forklift. If the input data are known in advance, the problem is efficiently solvable with a dynamic programming approach. However, this method is rarely applicable in practice for two reasons. First, the problem generally occurs in a real-time environment where the input data are revealed over time. Second, computing devices are in most cases not available in forklifts or gravity conveyors for decision-making. Online approaches that can easily be applied by human operators are therefore sought in practice. With this in mind, we first propose some intuitive approaches and analyse their performance through an extensive experimental study. The results show that these approaches are quite inefficient as they average between 14.7 and 59.3% above the optimum. A less intuitive but still simple approach is then designed that consistently produces good results with an average gap of 6.1% to the optimum.