Map Representation and Path Planning for Indoor Navigation on a Smartphone Chet Gnegy*, Balajee Kannan † Department of Electrical Engineering, University of Pittsburgh, Pi#sburgh, PA 15261* Robotics Institute, Carnegie Mellon University, Pittsburgh, PA 15289† The NavPal Applica.on Computa8onally intensive localiza8on, planning, and predic8on processes on the Android Nexus S Smartphone Localiza8on to pinpoint user’s loca8on in a public building Path Planning to help user navigate to desired des8na8on Capability to interpret floor plans that have not been hand prepared A lightweight representa8on of a hierarchical map network that allows for landmarks or obstacles to be tagged by user Des8na8on predic8on based on user’s current path and movement histories. Localiza8on Map Representa8on Path Planner Des8na8on Predic8on WiFi Signal Mapping Onboard Gyroscope User Interface Campus Building Floor Room Ver8ces Edges Grid Hierarchical Map Network Map hierarchy is represented by a connected graph of ver8ces and edges Each level has connec8ons (i.e. roads, elevators, or doorways) that link the ver8ces The user interacts beneath the lowest hierarchy on the grid The grid supports annota8ons, as well as more fine grain path planning User’s movement between rooms is sent to the des8na8on predictor and the most likely des8na8on is returned to the path planner Path Planning Recursively searches through the network of ver8ces and edges un8l a connec8on is found If a common ancestor cannot be found between the user’s room and the des8na8on’s room, no path exists Coarse grain planning is done on ver8ces to determine the closest elevator or staircase en route to the des8na8on Planner narrows in on user posi8on and only uses fine grain grid planning to lead them to that nearest elevator or stairwell Path planning will lead user to specified landmark or predicted des8na8on Interpre.ng the Map Black and white color data on map gives good es8mate of boundaries and free space Standard PDF files can be converted to vector data supplying the shape of the room PDF obtained from Carnegie Mellon website Color data and vector data are combined to give an accurate representa8on of the walls on the floor plan Acknowledgements I would like to thank my colleagues Evan Glasgow and Piotr Yornadov for their important contribu8ons to the NavPal applica8on this summer. Thanks to Google for sponsoring the development of this applica8on. PROVIDED PDF Exit