Locating a mobile robot in a non_accessible area

Localization component is fundamental to autonomous robotics along with obstacle avoidance path planning and control strategies. Due to the imperfections of sensors, a navigating robot should localize itself using information from different sensors. Localization is nontrivial problem to solve due to uncertainties in both the driving actuators process and the sensing measurements. In non accessible area such as shielded buildings, space or underground the localization problem becomes more difficult. Several researchers addressed the localization problem. However, the development of robust and efficient robot localization algorithm remains unsolved specially, when the robot is moving for long time and the initial position is unknown. Particle Filters (PFs) were recently introduced as a powerful estimation tool that provides a general framework for estimation of nonlinear/ non-Gaussian dynamic systems including localization problem. In this work an attempt to improve the particle filters by investigating the effects of several re sampling approaches in the behavior of Particle Filter (PF) based robot localization and selects the best approach. Also, another improvement to particle filter is introduced by using excitation strategy. This strategy is based on re-exciting the particles if their weights fall below a certain weight value. An algorithm of particle filter is built in Matlab environment to host these ideas implementation. The performances of PF by using different re sampling schemes are evaluated in terms of computational time and error from ground truth and the results are reported. Also, this study presents a proposed Enhanced Particle Kalman Filter (EPKF) algorithm which combines the advantage of Particle Filter (PF) and Extended Kalman Filter (EKF) in S/W and H/W. The analysis and implementation of several localization algorithms including the proposed EPKF algorithm for evaluation purposes are conducted