A Genetic Programming Hyper-Heuristic Approach for Evolving

We present a genetic programming hyper-heuristic system to evolve a ‘disposable’ heuristic for each of a wide range of benchmark instances of the two-dimensional strip packing problem. The evolved heuristics are constructive, and decide both which piece to pack next and where to place that piece, given the current partial solution. Usually, there is a trade-off between the generality of a packing heuristic and its performance on a specific instance. This hyper-heuristic approach is particularly beneficial because it produces a heuristic tuned to the instance, and it is general enough that it can do so for any strip packing instance, with no change of parameters. The contribution of this paper is to show that the genetic programming hyper-heuristic can be employed to automatically generate heuristics which are better than the human-designed state of the art constructive heuristics, in a very well studied area.