The present research work deals with the implementation of heuristics and genetic algo- rithms to solve various bin packing problems (BPP). Bin packing problems are a class of optimization problems that have numerous applications in the industrial world, ranging from efficient cutting of material to packing various items in a larger container. Bin packing problems are known to be non-deterministic polynomial-time hard (NP-hard), and hence it is impossible to solve them exactly in polynomial time. Thus heuristics are very important to design practical algorithms for such problems. In this research we avoid the use of linear programming because we consider it to be a very cumbersome approach for analysing these types of problems and instead we proposed a simple and very efficient algorithm which is a combination of the fi fi heuristic algorithm in combination with the genetic algorithm, to solve the two and three – dimensional bin packing problems. The packing was carried out in two phases, wherein the fi phase the bins are packed by means of the fi fi heuristic algorithm with the help of other auxiliary techniques, and in the second phase the genetic algorithm is implemented. The purpose of the second phase is to improve the initial arrangements by performing combinatorial optimization for either a limited number of bins or the whole set at one time without destroying the original pattern (elitist strategy). The programming code developed can be used to write high-speed and capable software, which can be used in real-time applications. To conclude, the developed optimization ap- proach signifi tly helps to handle the bin packing problem. Numerical results obtained by optimizing existing industrial problems demonstrated that in many cases it was possible to achieve the optimum solution within only a few seconds, whereas for large-scale complex problems the result was near optimum efficiency over 90% within the same period of time.
Two and three-dimensional bin packing problems : an efficient implementation of evolutionary algorithms