Solving multiple travelling salesman problem through deep convolutional neural network

Abstract The multiple travelling salesman problem (mTSP) is a classical optimisation that widely applied in various fields. Although the mTSP was solved using both algorithms and artificial neural networks, reiteration inevitable for these methods when presented with new samples. To meet online high‐speed logistics requirements deploying information technology, iterative algorithm may not be reliable timely. In this study, deep convolutional network (DCNN)‐based solution method proposed, which can establish mapping between parameters optimal solutions directly avoid use of iterations. facilitate DCNN establishing mapping, an image representation transfer from into computer vision presented. While maintaining excellent quality results, efficiency achieved by proposed much higher than traditional after training. Meanwhile, to solve under different constraints learning.