Probabilistic Self-Localization and Mapping: An Asynchronous Multirate Approach
[EN] In this paper, we present a set of robust and efficient algorithms with O(N) cost for the solution of the Simultaneous Localization And Mapping (SLAM) problem of a mobile robot. First, we introduce a novel object detection method, which is mainly based on multiple line fitting method for landma...
| Autores: | , , , |
|---|---|
| Tipo de recurso: | artículo |
| Fecha de publicación: | 2008 |
| País: | España |
| Institución: | Universitat Politècnica de València (UPV) |
| Repositorio: | RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia |
| Idioma: | inglés |
| OAI Identifier: | oai:riunet.upv.es:10251/152474 |
| Acceso en línea: | https://riunet.upv.es/handle/10251/152474 |
| Access Level: | acceso abierto |
| Palabra clave: | Multirate fusion Probabilistic Localization Mapping FastSLAM INGENIERIA DE SISTEMAS Y AUTOMATICA |
| Sumario: | [EN] In this paper, we present a set of robust and efficient algorithms with O(N) cost for the solution of the Simultaneous Localization And Mapping (SLAM) problem of a mobile robot. First, we introduce a novel object detection method, which is mainly based on multiple line fitting method for landmark detection with regular constrained angles. Second, a line-based pose estimation method is proposed, based on LeastSquares (LS). This method performs the matching of lines, providing the global pose estimation under assumption of known Data-Association. Finally, we extend the FastSLAM (FActored Solution To SLAM) algorithm for mobile robot self-localisation and mapping by considering the asynchronous sampling of sensors and actuators. In this sense, multi-rate asynchronous holds are used to interface signals with different sampling rates. Moreover, an asynchronous fusion method to predict and update mobile robot pose and map is also presented. In addition to this, FastSLAM 1.0 has been also improved by considering the estimated pose with the LS-approach to re-allocate each particle of the posterior distribution of the robot pose. This approach has a lower computational cost than the original Extended Kalman Filtering (EKF) approach in FastSLAM 2.0. All these methods have been combined in order to perform an efficient and robust self-localization and map building process. Additionally, these methods have been validated with experimental real data, in mobile robot moving on an unknown environment for solving the SLAM problem. |
|---|