Abstract
This paper proposes a novel localization algorithm (Pseudo Image Localization: PIL) using a grid able to solve global localization and robot kidnapping issues. The proposed approach has 2 steps. The off-line step is a learning phase during which a simulated robot completely learns its environment at attitude θ = 0. Several orientation-free level functions are defined. These functions are the different filters on which the on-line phase will rely to select the position candidates. During the on-line phase the robot smartly uses the massive data produced during the off-line phase to recover first its position and second its orientation. The paper proposes the theoretical development of this localization approach, some simulation examples and an experimental validation carried out with a Pioneer 3DX robot moving in a predefined environment.