Mobile Robot Metric-Topological Map Building And Localization

The research of mobile robot involves a lot of knowledge such as intelligent control, computer science, pattern recognition and artificial intelligence. With the fast development of new technology and the wide use of advanced sensors, mobile robot has become the focus in the field of robotics and automation. Map building and localization are two essential tasks for an autonomous mobile robot's navigation. This paper researches on mobile robot's metric-topological map building and hybrid localization.The paper introduces the model and system error of SmartROB-2 mobile robot's odometer and laser range finder firstly. It points out the limitation in dead-reckoning and gives out the state covariance matrix. Then the paper analyses various environmental infections on laser range finder. It describes the error distribution of distance and angle and builds uncertainty model in range measurement of LMS 200 laser range finder.For other map building methods, a novel metric map building approach based on dual laser ranger finder is proposed. The clustering method has advantages of simple principle, little calculation and precise result and the experimental results show the method's practicability and validity. The paper realizes precise localization using extended Kalman filter based on real-time metric map. High-level features including corners, doors and openings are identified and extracted which is the base for metric-topological map building. This paper present a hybrid localization approach to complete both extended Kalman filter based pose tracking in local metric map and re-localization in topological nodes using the binding high-level features. Experiment results implemented in SmartROB-2 mobile robot platform and further data analysis show the method can solve the failure of EKF localization.This paper researches topological navigation based on POMDP exploringly and gives the state estimator, set of states, set of actions, observation function and transition model. The research establishes a foundation for future work.