Cloud Based Real Time Localization For Ground Mobile Robot In Outdoor Environment

Localization is an important problem for the navigation of outdoor mobile robots.Robot localization especially for large-scale exploration is mainly based on inertial navigation technology and Global Positioning System(GPS).The performance has been limited by the sensor measurement accuracy,GPS signal and other inherent computation and storage constraints of the onboard processor.With the development of the Internet communication technology,the concept of cloud robotics is proposed.The large amount of experiment data and computation load can be distributed to the cloud to lower the cost and increase the portability of the robot.The main purpose of this project is to propose a cloud based real time localization architecture to allow a ground mobile robot to identify its location in large scale outdoor environment without the constraints of onboard inertial navigation sensors and the GPS.The proposed cloud based architecture consists of two phases: offline and online.The extraction of road network information and the construction of the reference images constitute the offline phase.The former is to extract new road networks and update the existing road network map stored in the cloud.To further improve localization accuracy for mobile robot,a reference image database is constructed in the proposed framework.Image of a specific location in the road network is taken as a reference for localization.The vision based supplement is used to increase the accuracy of localization.The online phase is for mobile robot localization based on the cloud.The update of the road network map and the computation of the robot terrain inclination model as well as reference image matching are executed in the cloud.A particle filter with network delay compensation localization algorithm is executed on the mobile robot based on the local RTI model and recognized location sent from the cloud.The proposed methods are tested on three different challenging outdoor scenarios with a ground mobile robot equipped with minimal onboard hardware where the longest trajectory was 13.1km.The three scenarios consist of different road types including service way,urban canyon and tunnel etc which are challenging for localization process.Experimental results show that this method could be applicable to large-scale outdoor environments for autonomous robots in real-time.