TAP-Net

Abstract

We introduce the transport-and-pack (TAP) problem, a frequently encountered instance of real-world packing, and develop a neural optimization solution based on reinforcement learning. Given an initial spatial configuration of boxes, we seek an efficient method to iteratively transport and pack the boxes compactly into a target container. Due to obstruction and accessibility constraints, our problem has to add a new search dimension, i.e., finding an optimal transport sequence , to the already immense search space for packing alone. Using a learning-based approach, a trained network can learn and encode solution patterns to guide the solution of new problem instances instead of executing an expensive online search. In our work, we represent the transport constraints using a precedence graph and train a neural network, coined TAP-Net, using reinforcement learning to reward efficient and stable packing. The network is built on an encoder-decoder architecture, where the encoder employs convolution layers to encode the box geometry and precedence graph and the decoder is a recurrent neural network (RNN) which inputs the current encoder output, as well as the current box packing state of the target container, and outputs the next box to pack, as well as its orientation. We train our network on randomly generated initial box configurations, without supervision , via policy gradients to learn optimal TAP policies to maximize packing efficiency and stability. We demonstrate the performance of TAP-Net on a variety of examples, evaluating the network through ablation studies and comparisons to baselines and alternative network designs. We also show that our network generalizes well to larger problem instances, when trained on small-sized inputs.

A review on learning to solve combinatorial optimisation problems in manufacturing

“…Except the standard constraints in BPP, for example, capacity/ weight constraints, overlap constraint, it could be a valuable direction to cope with realistic constraints in learning, so that the trained models could be used in real-world settings. The additional constraints are different in varying domains, for example, physical stability with robots [65], customer requirements in delivery [49], and preference constraints in the style of picking and packing [67]. Regarding the criteria in BPP, existing methods mostly focus on minimising the utilisation rate and bin number.…”

“…They use prioritised oversampling to improve the training and evaluate their method for offline/online 3D-BPPs [104]. Hu et al resort to an encoder-decoder neural architecture to learn the selection of the pair of an item and its orientation [67].…”

“…Hu et al. resort to an encoder‐decoder neural architecture to learn the selection of the pair of an item and its orientation [67]. They consider a special case where the items are initialised by a spatial configuration, which leads to a preference constraint between items, for example, an item can only be selected to pack when all items on top of it have been packed.…”

“…Other variants of packing problems and their corresponding MDPs (e.g. packing with online settings, multiple bins or continuous positions) can be found in literature [18,49,67,68].…”

“…Regarding the criteria in BPP, existing methods mostly focus on minimising the utilisation rate and bin number. Some other objectives are also critical to consider in the optimisation, for example, stability, compactness, pyramidality [67]. Thus it could be worthwhile to extend multi-objective learning techniques for solving BPPs to gain more reliable performance.…”