«Navi-Campus» : an orientation and navigation app for helping visually impaired people to walk independently on any university campus Jesus ZEGARRA FLORES Altran Research Medic@ Laurence RASSENEUR Université Strasbourg Marie Odile HEMERLIN Université Strasbourg Catherine PEYRANI René FARCY Association d aveugles d Alsace Université Paris Sud Laboratoire Aimé Cotton
State of the art Use of Talk Back or Voice Over accessibility layer to give the vocal instructions (phone interaction and navigation app) Human Machine Interface (HMI) for entering addresses. Guidance information heading and distance or first to the right, etc.. Examples of current GPS devices: Trekker (2004) Navigon Mobile Navigator (2009) Kapten Mobility (2012) Via Opta Nav (Information turn right, etc.) (last version, may 2016) BlindSquare (Distance & heading information, points of interests) (last version june 2017) Ariadne GPS for iphone (Distance & heading ) (july 2016)
Problems of the GPS heading in urban areas Heading GPS walking displacement Heading GPS in a car
Problems of the GPS heading in case of hesitation 90 360 45 225 270 180 225 270 6 m 90 6 mètres
Practical orientation problems with current devices If you only meet right angle intersections, if you never take the wrong way, if your starting point is not a large place etc. you will arrive at destination. In case of punctual hesitation, of an Y intersection, of a large place to cross etc you will be definitively lost.
Specific Work Merge of GPS antennas, compasses and inertial sensors in Smartphones in order to give instant updated orientation to visually impaired people in Urban environments : Naviurbain Hiking environments : Navirando Large private facilities as campuses : Navicampus We will present the outcome of instant orientation, and the specific problems of mapping on a campus.
Materials Prototype developed in a ios operative system and Android. Programming in Objective C (X code) and JAVA. Devices which come with sensors: GPS, compass, accelerometer and gyrometers.
Results: improving heading information in case of hesitation Routes Real heading ( ) Maximu m error ( ) heading GPS Standard deviatio n heading GPS Maxim um error ( ) headin g IMU Standard deviation heading IMU route A-B 0 170 128.3 11 4.39 route B-C 90 57 0 19 4.67 route C-D 180 33 0 22 4.45 route D-A 270 123 47.26 21 12.72 route A-C 45 113 27.31 23 2.06 route C-A 225 127 44.06 17 4.9 route A-E 90 163 52.45 24 10.23
Interface The app uses the «heading» and «distance» information because it takes into account the axis displacement of the person and also the change of direction. It has been used the «heading» given by the compass coupled to the gyroscope and GPS. Heading and distance : case 1: 9 o clock 150 meters ; case 2: 6 o clock 150 meters 30 meters before the current point, information to the next point
Body axis Basic guidance : Point 3 3 o clock 150 meters Law faculty Basic guidance : point 3 12 o clock 150 meters Information for buildings at less than 50 meters : Law Faculty on the right Point 3 X Advance guidance : point 3 12 o clock, 150 meters + Advance Information : Route at 3 o clock X Spatial representation and anticipation training
Problems about cartography In some GPS devices, the GPS coordinates found from a faculty building address do not correspond to the real entrance (the path does not exist). Google does not enter in private facilities. Google position Real position 90 meters
Adapted cartography In order to have a good cartography adapted to the visually impaired displacements: - All public entries of buildings are identified - All practicable segments of routes are identified - Path finding using Dijkstra algorithm, the paths generated priories the main roads which have good haptic feedbacks for the white cane.
Test with a visual impaired student Person always accompanied, the purpose is to reach autonomy in the campus Previous basic mobility course We have taken the information from the speed of the GPS, 10 meters before and 10 meter after each turn (T1 T2 T3, etc.) in order to have an average speed. The efficiency in a turn is obtained by using: average speed in the turn / average speed in a straight line. The average speed in straight line is: 3km/h.
Campus Esplanade Test Escarpe to Cafétéria Chimie (688 m): T1: 90 turn T2: 90 turn T3: 90 turn Cafétéria Chimie to Pangloss (438m) : T4: <45 turn T5: 90 turn in a small square 25 m2 Pangloss to Institue Physique de Globe (588m): T6: 90 turn T7: 90 turn T8: 90 turn T9: 90 turn
Results: 1 st path : Escarpe to Cafétéria Chimie (15min 46 seconds) 2 nd path : Cafétéria Chimie to Pangloss (17min 25 seconds) 3 rd path Pangloss to Institute de Physique de Globe (11min 25seconds) km/h Efficiency % T1 2,73 91% T2 2,11 70% T3 2,15 71% T4 1,53 51% T5 1,74 58% T6 2,27 75% T7 2,32 77% T8 2,42 80% T9 2,5 83%
51% 58% T4 : the person did not take the road to the left. T5: the person hesitated because of the small square
Conclusions Merging GPS heading with compass and gyroscope improve the heading information in case of hesitation. Instant updated information relatively to body axis can extend GPS use to complex environments. Training is important to be efficient in merging GPS information with his own sensorial information. To extend mobility and maintain security is the challenge.