Self-localization Algorithm of Mobile Robot Based on Unscented Particle Filter

Based on Competition of RoboCup Standard Platform League, this paper will do some researches about how to use the data of odometer and the images from camera of Nao robot to realize self-localization. First, this paper defines the world coordinate system and the robot coordinate system. Based on the coordinate systems, this paper defines the state variables, presents the state equations and observation equations of the dynamic system, and describes how to calculate the observation information of the robot pose through recognition information of the camera. Then taking real scene of Robocup into account, it introduces Unscented Kalman Filter which is put into the particle filter framework to get Unscented Particle Filter(UPF). The UPF algorithm is used to realize self-localization. Finally, this localization algorithm is implemented on Nao robot through a series of simulation experiments. The experiment shows that the efficiency, accuracy, stability of UPF algorithm is much higher than the Particle Filter(PF) algorithm, which proves the superiority of unscented particle filter algorithm in self-localization of mobile robot.