An Approach for the Exploration of Unknown Environments: Multi-platform Validation and Integration on a Dog-like Robot

Abstract

[eng] In the field of robotics, exploration persists as an arduous problem yet to be satisfactorily solved. In this thesis, we propose a new frontier-based exploration algorithm to tackle this issue in indoor and confined environments. Our approach consists in deploying two different trees, namely the local tree and the global tree, which are generated by expanding an RRT that considers the robot constraints. The local tree is periodically regenerated, while the global tree accumulates environment data as the exploration progresses. Both trees are used to find points in the environment that are relevant for the exploration task, i.e. frontiers. Then, these frontiers are validated and clustered to subsequently be chosen as the next platform goal, for which the algorithm provides a path computed by means of the global tree. To demonstrate the effectiveness of our exploration approach, we compare its performance with that of GBPlanner2 using a simulated aerial robot. Besides, we show the adaptability of our solution for different platforms, as well as its operability in real environments, integrating the algorithm on a dog-like robot.