Research On Localization And Mapping Technology For Indoor Mobile Robot

Autonomous mobile robots are widely used in military projects,industries and daily domains,localization and mapping technology for which is the key step to realize intelligence and autonomy.A robot realizes localization according to the built environment map,meanwhile,its position information is used to update the map.So the localization and mapping technology has been a research focus.Our study focus on realization of this technology on the Pioneer2 DX mobile robot in indoor environment with laser ranger finder.To meet the demand of localization for mobile robots in indoor environments,a general plan of a prior maps based localization system is proposed,consisting of scan matching,graph optimization,global map building,online localization and map updating.Also,the function of each part and the relationship between them is introduced.The ordinary system,environment model and sensor model are then built,followed by the analysis of pose estimation performance for odometer.Aiming at the optimization of mobile robot pose,the principles,advantages and disadvantages of standard Iterative Closest Point,Iterative Dual Correspondence and Cox algorithm are analyzed.A scan matching method combining Iterative Dual Correspondence and Cox algorithm is proposed based on our requirements.The model of laser based graph optimization is built,the principle and characteristics of least square and stochastic gradient descent is analyzed.Considering our practical situation,a Lerbvenberg-Marquardt based graph optimization method is used.Line map observed in each pose is built,then the global line map is also built using the optimal pose estimation above.On this basis,these lines are merged by direction clustering and space clustering based on mean-shift algorithm to build the resultant off-line global map,and the online localization and map updating strategy is then proposed.Experiment validations of proposed localization and mapping method are conducted,including tests on scan matching,graph optimization,local map building,global map building,online localization and map updating.Results perform well using two real sensors' data sets...