Abstract
This paper presents new concepts to apply Self-Organizing Maps (SOM) to vehicle path problems. It is reported SOM is capable of solving a traveling salesman problem (TSP), one of the vehicle path problems. However, it is not investigated how SOM is applied to other types of the vehicle path problems, such as the shortest path problem (SPP) and n traveling salesmen problem (n-TSP). Numerical experiments prove that SOM cannot lead to good solution when it is applied to SPP and n-TSP. To improve SOM, two multi-neuron concepts are introduced to solve these problems. Numerical experiments verify that SOM using multi-neuron leads to better solutions than the conventional SOM does. Furthermore, SOM with variable number neurons is proposed to make SOM's solution converge efficiently.
