Abstract

In this paper, we report on our early results on deploying unsupervised learning technique for solving a multi-goal path planning problem to determine a shortest path to visit a given set of 3D regions. The addressed problem is motivated by data collection missions in which a robotic vehicle is requested to visit a set of locations to perform particular measurements. Instead of precise visitation of the specified locations, it is allowed to take the measurements at the respective distance from the locations, and thus save the travel cost by exploiting non-zero sensing radius of the vehicle. In particular, the problem is formulated as a 3D variant of the Close-Enough Traveling Salesman Problem (CETSP), and the proposed approach is based on the recently introduced technique called the Growing Self-Organizing Array (GSOA). The GSOA is a neural network for routing problems that is accompanied with unsupervised learning procedure to determine a solution of the TSP-like problems in a finite number of learning epochs. Based on the reported results, the proposed GSOA-based approach provides competitive or better results than existing combinatorial heuristics based on the so-called Steiner zones, while the computational requirements are significantly lower.