Research On Indoor Path Planning Based On Mobile Crowdsensing

With the popularity of wireless and mobile devices,the traditional business model and ecology is changing due to the combination of indoor location based services(LBS)and mobile devices.The major Internet giants have deployed some LBS applications,so that indoor LBS have been developing rapidly and have entered people's life to some extent.The goal of indoor LBS is to propose various indoor path planning approaches.Indoor path planning is still a challenging task and is becoming fundamental research topic in industry and academia.Motivated by these challenges,this thesis presents theoretic,technical,and application investigation on indoor path planning.First,this thesis presents a semantic mapping method via mobile crowdsensing.This method uses both the vision and the motion information.The vision information can provide rich semantics and accurate geometric relationships between two images.Motion signals can provide the rough absolute location and orientation of images in the world coordinate system to reduce computing overhead.To efficiently and correctly localize semantics in a map,a combinatorial optimization and probabilistic occupancy technique is proposed using images acquired from multiple viewpoints.Second,since indoor environments are usually dynamics,these semantics in the initial indoor semantic floorplan would be out-of-date,resulting in a deteriorated and even break down performance of LBS systems.Therefore,new semantics should be labeled and out-of-date semantics should be removed.This general problem is formalized as indoor semantic floorplan updating.This thesis presents a mobile crowdsourcing system to automatically and continuously update semantics of general entities indoor environments.The performance of the proposed methods has been tested on several datasets.Third,conventional designs of indoor navigation systems depend on either infrastructures or indoor floor maps.This thesis presents a ubiquitous indoor navigation solution,which relies on the point clouds acquired by 3D camera embedded in a mobile device.Compared to other indoor signals(e.g.,radio frequency signals and geomagnetic data),features extracted from point clouds are not affected by variations in scale,rotation and illumination.The proposed method presents the walking trace of a user inferring algorithm,3D path-map generating algorithm,particle filter based tracking algorithm and deviation detecting algorithm.Extensive experiments are conducted on office building.Experimental results indicate that the proposed method exhibits outstanding navigation performance.Finally,people have a strong need for navigation from a large open indoor environment to an outdoor destination in real life.However,conventional designs of navigation systems mainly focus on either indoor or outdoor navigation.This thesis presents a joint architecture of indoor-outdoor navigation and a crowdsourcing solution by sharing the traces of different users.Extensive experiments are conducted and the experimental results indicate that the proposed method exhibits outstanding navigation performance from an indoor location to an outdoor destination.