tailieunhanh - Robotics 2010 Current and future challenges Part 13

Tham khảo tài liệu 'robotics 2010 current and future challenges part 13', kỹ thuật - công nghệ, cơ khí - chế tạo máy phục vụ nhu cầu học tập, nghiên cứu và làm việc hiệu quả | 410 Robotics 2010 Current and Future Challenges Fig. 7. Obtained robot position uncertainty when using the Information filter for only distributed only onboard or both sensors and when CI is used for onboard sensors. Fig. 8. Representation of the space and mobile robots with occupancy grids. The idea of occupancy grids is to divide the space in a grid usually evenly spaced. Each cell in the grid is given a measure representing the probability that the cell is occupied by an object. In order to make the combination easier we model both the robot s shape and the geometry of the space using occupancy grids . instead of using landmarks like in the previous section here we use an occupancy grid representation of a map of all static objects in the space . An example illustrating the overall model is given in Fig. 8. In addition to the occupancy grid representing the map of the space there is also an occupancy grid that outlines the Robot Localization Using Distributed and Onboard Sensors 411 shape of the robot. The space map is fixed to the world coordinate system whereas the robot model changes its position with the robot. Estimation of the robot position First we assume that the occupancy grids representing the robot and space models are known. A method how to build them will be explained in the next subsection. Occupancy grids do not allow a straightforward usage of the Kalman filter or similar methods like in the previous section. A common approach is to use a particle filter based approach Arulampalam et al. 2002 and one standard particle filter based method used in robot localization with onboard sensors is Monte Carlo localization MCL Dellaert et al. 1999 . The idea of MCL is to represent the belief on the robot s position by a set of particles each representing one hypothesis on the current pose - . the robots x 1 position and orientation 0. At every step the algorithm goes through several functions to update the set of particles namely motion model .