compensate declining physical and cognitive capabilities

18
Christian Mande Bernd Krieg-Brückne Bernd Gersdor Christoph Budelman Marcus-Sebastian Schröde Navigation Aid for Mobility Assistants Joint CEWIT-TZI-acatech Workshop “ICT meets Medicine and Health” ICTMH 2013

Upload: koto

Post on 23-Mar-2016

43 views

Category:

Documents


2 download

DESCRIPTION

Overview: Walker with NavigationAid. Compensate declining physical and cognitive capabilities Provide navigation assistance that considers specific needs: Precise localization Route planning respecting vehicle specific constraints User interface suitable for the elderly. - PowerPoint PPT Presentation

TRANSCRIPT

PowerPoint-Prsentation

Christian MandelBernd Krieg-BrcknerBernd GersdorfChristoph BudelmannMarcus-Sebastian SchrderNavigation Aid for Mobility AssistantsJoint CEWIT-TZI-acatech WorkshopICT meets Medicine and HealthICTMH 2013

Compensate declining physical and cognitive capabilities

Provide navigation assistance that considers specific needs:

Precise localization

Route planning respecting vehicle specific constraints

User interface suitable for the elderly

Overview: Walker with NavigationAidIntroductionOutdoor LocalizationRoute PlanningUser InterfaceResults / OutlookTwo versions of OdoWheel Inertial Measurement Unit (IMU)

Current revision comprises

3-axis acceleration sensor and gyrometer

Bluetooth [Low energy] radio link

Battery [solar] driven power supply

32 bit microcontroller

Extended Kalman Filter fuses accelerometer- and gyro-data Odometry

Additional Hardware Component: OdoWheel

IntroductionOutdoor LocalizationRoute PlanningUser InterfaceResults / OutlookOSM description of road network, land usage, buildings,

Open community project Based on user-recorded GPS track logs, or vectorization of aerial images

XML vector representation with atomic building blocks: points, ways, relations

Free tagging system for annotation of properties

Handy modeling toolssuch as the Java-OpenStreetMap-Editor (JOSM)

Environment Representation: OpenStreetMap (OSM)

IntroductionOutdoor LocalizationRoute PlanningUser InterfaceResults / OutlookEnvironment Representation: OpenStreetMap (OSM)Road network stored in PMR-Quadtree

Space partitioning data structure sorting its entries into buckets

Bucket is split into four child buckets when |entries| exceeds threshold c

Let N := |position hypotheses| and M:= |road segments| O(c*N) instead of O(M*N) distance(road segment, position) queries for finding closest road segment to given pose hypothesis when using PMR-Quadtree

[1] E.G. Hoel and H. Samet: Efficient Processing of Spatial Queries in Line Segment Databases. In: Advances in Spatial Databases; Vol.: 525 of Lecture Notes in Computer Science, pages 237-256. Springer Verlag, 1991. IntroductionOutdoor LocalizationRoute PlanningUser InterfaceResults / OutlookMonte Carlo Localization: Motivation

[2] GPS Essentials of Satellite Navigation Compendium. uBlox, 2009. Online: http://www.u-blox.ch/images/downloads/Product_Docs/GPS_Compendium%28GPS-X-02007%29.pdf Sources of GPS errors

Multipath signals reflected from buildings, trees, mountains,

IntroductionOutdoor LocalizationRoute PlanningUser InterfaceResults / OutlookMonte Carlo Localization: OverviewMotion UpdateSensor UpdateResamplingModel estimate of current position by set of samples

Move each pose hypothesis according to:

Odometry measurements

Translational, and rotational noise

IntroductionOutdoor LocalizationRoute PlanningUser InterfaceResults / OutlookMonte Carlo Localization: OverviewMotion UpdateSensor UpdateResamplingScore each pose hypothesis according to:

Distance to GPS measurement

Distance to closest OSM path

Type of closest OSM path, kind of entity passed over during last motion update

IntroductionOutdoor LocalizationRoute PlanningUser InterfaceResults / OutlookMonte Carlo Localization: OverviewMotion UpdateSensor UpdateResamplingRebuild set of samples for next frame

Samples score determines probability to occur in the new set

IntroductionOutdoor LocalizationRoute PlanningUser InterfaceResults / OutlookEstimated state is a pose in 2-D Particle implementation:

Motion model:State transition based on traveled distance and rotation

Update of sample position

Monte Carlo Localization: Motion Update

IntroductionOutdoor LocalizationRoute PlanningUser InterfaceResults / OutlookMonte Carlo Localization: Sensor UpdateSensor model:

position measurement from a connected GPS device

virtual path distance measurement (always zero)

virtual measurement describing expected behavior

Computation of weighting:

IntroductionOutdoor LocalizationPath PlanningUser InterfaceResults / OutlookIntroductionOutdoor LocalizationRoute PlanningUser InterfaceResults / OutlookOSM Based Route PlanningUses 22 different path typesincluding oneway paths

Platform/user-sepcific weighting

Uses A-star algorithm

Computation of turn advices

Map View of User Interfacedetailed representation of surroundingsimmediate walking directionabstract path network with walking directionplanned pathcurrentpositionIntroductionOutdoor LocalizationRoute PlanningUser InterfaceResults / Outlook

Compass View of User Interfaceabstract path network with walking directionimmediate walking directionIntroductionOutdoor LocalizationRoute PlanningUser InterfaceResults / Outlook

Selecting (special) Targets in User Interfacepush to speak target locationtype in target locationpush to select special targetIntroductionOutdoor LocalizationRoute PlanningUser InterfaceResults / Outlook

Localization ExampleEstimated trajectory (red) vs. GPS trajectory (green)IntroductionOutdoor LocalizationRoute PlanningUser InterfaceResults / OutlookFuture WorkOutdoor Localizer

Route Planning

Evaluation

Hardware Integration

Vehicle Platforms

Barthel IndexNASA Task Load Index

IntroductionOutdoor LocalizationRoute PlanningUser InterfaceResults / Outlook

Navigation Aid for Mobility AssistantsJoint CEWIT-TZI-acatech WorkshopICT meets Medicine and HealthICTMH 2013Thank you for your attention! Questions?