Exploration of Unknown Multiply Connected Environments Using Minimal Sensory Data

Abstract

In robotics, Bug/Gap algorithms have shown good results as an alternative for traditional roadmap techniques, with a promising future, these results were locally optimal and sufficient to navigate and achieve goals. However, such algorithms have not been applied, or tested, on all types of environments. This work is aiming at improving and adding to this category of algorithms using minimal sensory data. To achieve this objective, we adapt a dynamic data structure called Gap Navigation Trees (GNT) that represents the depth discontinuities (gaps). The final GNT characterizes a roadmap that robots can follow. The basic GNT data structure is reported to model simple environments. In this paper, we extend GNT to unknown multiply connected environments. In addition, we add landmarks to eliminate infinite cycles. The proposed algorithm can be used in a variety of solid applications such as exploration, target finding, and search and rescue operations. The solution is cost effective, which enables the production of affordable robots in order to replace expensive ones in such applications. The simulation results had validated the algorithm and confirmed its potential.