An efficient indoor large map global path planning for robot navigation

Meysami, A., Kelouwani, S., Cuillière, J.-C., François, V., Amamou, A. et Allani, B. (2024). An efficient indoor large map global path planning for robot navigation. Expert Systems with Applications, 248 . p. 123388. ISSN 0957-4174 DOI 10.1016/j.eswa.2024.123388

PDF Sous embargo jusqu'au 15 août 2026. Disponible sous licence Creative Commons Attribution Non-commercial No Derivatives. Télécharger (3MB)

Résumé

Large indoor cluttered environment representation is still a challenging task when non-uniform triangle cell-based or quadrangle cell-based decomposition is used to build the map. This paper aims at proposing a new method to represent a large indoor environment for efficient and global robot path planning using a trade-off among three criteria: path length, distance to obstacles, and path search complexity. In this regard, three steps are involved: (i) the design of a tiled map that represents several regular sub-regions of the given large environment; (ii) the selection of the best representation between non-uniform triangle cell-based submap and regular quadrangle cell-based submap for each tile using the predefined efficient path planning criteria. Hence, the entire large environment is represented by a hybrid cell-based map; (iii) the path search using the well-known A* algorithm. Moreover, to find the best representation, firstly, a generative method based on cellular automata is used to build a large synthetic database of maps of the same size as the tiles. Each map of the database is associated with the representation which provides the most efficient path planning. Given a tile, the corresponding image is used to find the closed tile image from the database, and the associated representation is selected as the best representation. Extensive simulations, as well as experiments, suggest that for a given large, cluttered environment, the hybrid representation with a mix of triangle-based cells and quadrangles-based cells can provide a more efficient global path compared to the traditional regular quadrangle or triangle representations.

Type de document:	Article
Mots-clés libres:	Robotics mapping Hybrid representations Convolutional neural network Mesh conformity
Date de dépôt:	28 oct. 2025 13:49
Dernière modification:	28 oct. 2025 17:23
Version du document déposé:	Post-print (version corrigée et acceptée)
URI:	https://depot-e.uqtr.ca/id/eprint/12280

Actions (administrateurs uniquement)

Éditer la notice