Algorithm for blind navigation along a GPS track

Transcription

1 Algorithm for blind navigation along a GPS track Rosen Ivanov Abstract: Most of existing navigation systems for the blind require a precise GPS maps. This makes them unusable in regions where there are no GPS maps, or they are not sufficiently accurate. lgorithm for GPS navigation for the visually impaired along a GPS track, which describes the path as a sequence of waypoints, is proposed. The main advantages of the algorithm are: natural voice navigation, adaptive to the velocity and accuracy of the GPS data, start of the navigation from any waypoint; correction of the direction of movement if it is necessary, return the user to the route if deviation is detected, work with and without electronic compass, detection of the movement of the user in the opposite direction. Key words: blind navigation, navigation along a track, Java mobile applications INTRODUCTION At present the most of outdoor navigation systems are based on Global Positioning System (GPS). The accuracy in determining the position of the users using GPS is in the range meters and it is not sufficient for navigation of the blind. Possible ways to improve the accuracy of GPS positioning are: reduction of random errors in GPS data (canyon effects in urban environment, delays of the GPS signals in the ionosphere and troposphere, and etc.) by using statistical techniques or by using Differential GPS (DGPS) or Satellite Based Augmentation Systems (SBAS) in which accuracy is less than 5 meters. The idea for using a GPS with the aim of developing navigation systems for the blind was proposed in 1985 by Collins [2] and Loomis [6]. The commercial applications in this area have high prices. They are based mainly on the use of smartphones, PDA or specialized hardware, GPS maps, and screen readers, for example: Sendero GPS (Sendero Group LLC, based on the BrailleNote platform, price around $1600), MobileGeo (Code- Factory, runs on smartphones equipped with the Mobile Speack screen reader and Pocket PC devices, price $1150), and Trekker (Humanware, runs on Maestro PDA, price $1600). In the absence of GPS maps for certain regions or insufficient precision of GPS maps one of the alternatives is to use the navigation alogn a track. The track is obtained as a sequence of GPS coordinates of the waypoints that are significant for the navigation process. The waypoints that form the track most often are obtained after specified period of time or distance. Part of the resulting waypoints do not describe exactly the route, mainly because of random errors in calculating the position. To minimize these errors, positions filtering is used. The navigation algorithm involves using only those waypoints (called critical points) where there is a real change of direction of movement. This implies the use of track simplification algorithms, for example Douglas-Peucker algorithm [3]. After track filtering and simplification navigation of type "from waypoint to waypoint can be used. There are few applications that allow blind navigation along a track. For example, application Loadstone GPS [10], which is developed by the blind for the blind, allows the recording of waypoints and navigation from one waypoint to another. This application, written in C for Symbian OS, has many disadvantages, for example: requires a screen reader, navigation is not possible to continue after the deviation from the route, can be installed only on the Nokia smartphones with platform S60, and etc. Java programming language is the most common used on modern mobile platforms. This guarantee a lower cost and greater availability of navigation systems, realized using Java mobile platforms [5]. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. CompSysTech'11, June 16 17, 2011, Vienna, Austria. Copyright 2011 ACM /11/06...$

2 DESCRIPTION OF THE ALGORITHM Navigation is a complex process for the people with visual disabilities. It includes the following stages: route planning (result is track that consists waypoints that describes the path), orientation, following a path, and overcoming the obstacles [1]. The proposed algorithm for the navigating along a track is part of an Java (J2ME) application for mobile phones which realize stages "route planning", "orientation and navigation along a track. Although the user can record voice messages (voice waypoints), describing obstacles or risky sections of the path, it is assumed that users use the white cane and/or a guide dog to avoid hazards and obstacles on the path. 1. GPS positions filtering To reduce random errors in the GPS data an algorithm for filtering of GPS positions is used. It combines the following techniques: knowledge-based GPS data pre-processing, a fuzzy estimation of noise level in GPS data and an adaptive Kalman filtering. Fig. 1 shows an example (one lap of the stadium on foot) for operation of the algorithm used to filter and simplify the GPS tracks. raw data: filtered data: track simplification, track simplification, 493 points, 444m 485 points, 406m filter off: 17 points, 414m filter on: 12 points, 401m Fig.1. GPS track filtering and simplification example 2. Select the time interval for checking the user s position GPS receivers generate information about the position of users during an interval of one second. When navigating along a track it is not necessary to check the position of the the user after every second, because he/she will not traveled significant for the navigation process distance. It is proposed distance D th, after which should to check the position of the user, to depend on velocity by the following equation: D th [ ] vf, vf 0, 100 km h =, (1) 40, vf > 100km h where v f is the filtered velocity: and i is the discrete time. v f i ( 1 ) ν i = αv f + α, (2) i 1 The check for the new position of the user is realized if the following conditions are fulfilled: 1) GPS receiver generates valid data (number of the visible satellites is greater than 3), 2) The estination for the horizontal accuracy - horizontal dilution of precision (hdop) - must be less than 3.5, 3) Filtered velocity v f must be greater than 1km/h, and 4) Distance traveled must be greater than or equal to D th. 373

3 3. Start of the navigation Before activate the mode "navigation" the user should load the desired track. When creating a new track, the user must give voice names of the first and last waypoints. When loading a track, the application inform the user by voice for the selected track. Let assume that one of the waypoints that describe the track is first, and another - last. There are two possible ways for the navigation along a track: from first to the last waypoint and vice versa - from last to the first waypoint. The following variables are introduced: N is the number of points which describe the track; firstpointindex is the index of first waypoint; lastpointindex is the index of last waypoint; currentpointindex is the index of current waypoint; nextpointindex is the index of next waypoint; direction encodes the direction of movement: 1, if movement isfrom firstpoint to lastpoint, firstpointindex = 0, lastpointindex = N 1 direction = (3) 1,if movement is from lastpoint to firstpoint, lastpointindex = 0, firstpointindex = N 1 If the movement is from currentpoint to nextpoint, then index of each next waypoint is calculated by the following equation: nextpointindex = nextpointindex + direction, if nextpointindex destinationpointindex (4) 4. Select the starting point The algorithm allows navigation along a track, regardless of the starting position of the user. Let the current position of the user is described with waypoint named currposition (see Fig. 2). p index p l d l d p r d r currposition p index+1 p index-1 Fig.2. Find the coordinates of the start point The calculation of the position of the startpoint is realized as follows: 1. Minimum distance d from the current position of the user (currposition) to each of the points p i, i=0 N-1 is calculated: d = min startposition, p, (5) i i where operator means GPS distance. In this case (see Fig. 2) the nearest point is p index. 374

4 2. If the condition d>min_dist_to_track is fulfilled it is assumed that the user is too far away from the track and navigation is impossible at this time. 3. We introduce the variable pos, which is used to find the starting point. For this purpose we obtain the following distances: d currposition, = p index, d l currposition, pl =, d r = currposition, pr, (6) where: d is the distance from current position to a waypoint p index ; d l is the minimum distance from the current position to the segment formed by points p index and p index-1 ; d r is the minimum distance from the current position to the segment formed by points p index and p index+1 ; The minimum distance d min =min(d, d l d r ) must be calculated. The start point of the track is that for which minimum distance d min was obtained: pindex, if d = d min, pos = 0 startpoint = pl, if dl = d min, pos = 1 pr, if dr = d min, pos = 1 (7) The index of the the next waypoint is given by the following equation: index,if pos = direction nextpointi ndex = (8) index + direction,otherwise 5. Navigate to the next waypoint There are several ways for voice navigation of the blind users: 1) Use the cardinal expression. The following words are used: north, south, east, and west, for example: west 25m to home. This way for navigation is not suitable for the blind. 2) Use the clock expression [9]. It is assumed that the user must follow the direction pointed by the hour hand of the the clock, for example: "home 25m at 11 o'clock". 3) Use the relative expression [8] (left, right, straight, and back), for example: turn left. The last method is suitable for the navigation in urban environment, but not outside the city. Experiments show [7] that the most appropriate and fast way for navigation of the blind is to use relative expressions with information for direction of movement in degrees (bearing mode), eg: turn 70 to the left. In the proposed algorithm is used bearing mode in which the angle is rounded to 10, since greater accuracy is difficult to be interpreted from blind. The users must move from the current waypoint (currentpointindex) to the next waypoint (nextpointindex), as shown in Fig. 3. North nextpointindex α 2 α=α 2 -α 1 α 1 d pc d next D currpointindex Fig.3. Navigation to next waypoint 375

5 Let the current position of the user is waypoint p c. Let assume that there is no deviation from the track - the distance d is less than a predetermined threshold value d th. If there is no deviation from the course and the user is close to the next waypoint (d next <D), he/she must be informed that after D meters must change its direction of movement. The value of the parameter D depends on the velocity and the current value of hdop (a low hdop value indicates better relative geometry and higher corresponding accuracy), D = f v, f hdop : ( ) ( 10 hdop + 5) + 0. v f D = 65. (9) When reaching nextpoint the user must change its direction of movement. Let α = α 2 α 1, where α 1 is the current direction of movement in degrees and α 2 is the new direction that must be followed. The blind user must change its course with α degrees. If α>0 the correction of the course is to the right to the current direction of movement, and if α<0 - the correction is to the left. If nextpointindex=lastpointindex navigation is stopped since the destination waypoint is reached. Otherwise the index to the next waypoint is calculated: nextpointi ndex = nextpointindex + direction. (10) 6. Navigation if deviation from the track is detected Deviation from an intended path is one of main problem for the blind [11]. The navigation algorithm must inform user when the route is left or lost. When moving from currpoint to nextpoint is possible the user to deviate from the track. Research in this area [4] shows that blind users deviate from the intended course by 1.3 on each step. Therefore, the navigation system must check the new position of the user in about ten meters since the error accumulates. For quantitative evaluation of deviation from the course is used perpendicular distance (d) from the current position of the blind (waypoint p c ) to the segments that form the points currpoint and nextpoint (see Fig. 4). North β 1 p c d β 2 β nextpointindex currpointindex Fig.4. Navigation if deviation from the track is detected If d d th it is assumed that there is significant deviation from the track and the user must be informed. The value of threshold d th in meters is calculated adaptively by the following equation: ( ) d th = hdop. (11) In case of deviation the blind is informed for the distance d and how many degrees (β=β 1 -β 2 ) left or right of course is the deviation. If β>0, the deviation is on the right, and if β<0 - the deviation is on the left. 376

6 7. Return to the track after the deviation Let assume that the user has deviated from the route at point A and is informed for the the deviation at point B (see Fig. 5). p B d th nextpointindex p A currpointindex p C γ Fig.5. Return to the track after the deviation Let assume that after being notified for the deviation, the user changes its course and returns to the track at point C. The user must be informed before it reaches point C how to change course to follow the correct course to nextpoint. For this purpose we calculate the difference between the course to point C and the course which the blind user must follows - γ. If γ>0 the correction of the course must be γ degrees to the left, and if γ<0 - γ degrees to the right. If the blind does not change its course after the notification for the deviation (point B), recording of new track at point A is started. Subsequently, through this new track will be possible to guide the blind to point A in order to continue navigation along the old track. 8. Detection of movement in the opposite direction It is possible the blind to move in the opposite direction. For this purpose calculates the absolute value of difference between the desired course α 1 and direction α 2 in which the the user actually moves. If the mobile phone has a built-in compass, the movement in the opposite direction is detected easily and quickly. If there is no compass, the algorithm takes time (the user must travel about 10 meters) to determine if the blind moves in the opposite direction. 9. Deviation from the nextpoint Let assume that the the user is approached to the nextpoint (d next <D) and the system had informed it that after D meters must change its course. If in this moment the user changes its course, at the next step of the algorithm the navigation should be back to nextpointindex, rather than nextpointindex+direction. This event is detected by comparing i i 1 the distances d next for the last two steps of the algorithm. If d next > 0.8d next it is assumed that the blind has deviated from the nextpoint. 10. Skipping a waypoints Let D p is the distance between waypoints currpoint and nextpoint. If condition D p <D is fulfilled the accuracy of the GPS receiver is not sufficient to realize a reliable navigation to the waypoint nextpoint. The waypoints nextpoint are skipped until meet the condition D p D. EXPERIMENTAL RESULTS Java mobile application for blind navigation along a GPS track is developed. The requirements for the successful installation of the application are as follows: profile MIDP 2.0/2.1 (JSR-118), configuration CLDC 1.1 (JSR-139), File Connection and PIM API (JSR- 75), Java APIs for Bluetooth (JSR-82), and Location API (JSR-179) optional. 377

7 All experiments were conducted with mobile phones Nokia 6630, Nokia N95, and Nokia C7-00. The application has been tested in urban (10 tracks) and in non-urban environment (20 tracks). The tests were performed with 5 users with visual impairment (myopia from 4 to 6 dioptres) and 10 users without visual impairments, but blindfolded. For the hdop values below 1.5 all tests are successfully completed. At lower accuracy (1.5<hdop<3.5) the time to reach the destination waypoint is increasing by 21% due to deviations from the track. When the value of the hdop is greater than 3.5 navigation is risky for the blind. Fig. 6 shows screenshots that describe various moments of the navigation process: the user is to far away from the track (42m), navigation without deviation from the track (go streight 59m), navigation with deviation from the track (deviation 89m, 30 to the right), and next waypoint has been reached (turn 46 to the right). Fig.6. Test of the algorithm In the future, the algorithm will be tested by the blind users. CONCLUSIONS An algorithm for the blind navigation along a track is developed and implemented. The main advantages of the algorithm are: Voice-based navigation. Start of the navigation from any waypoint of the track. The algorithm is adaptive to the transportation mode (pedestrian or vehicle) and accuracy of the GPS data. Detection of the movement in the opposite direction. Detection of the deviation, and the ability for the navigation back to the track. Possibility to pass consecutive waypoints if accuracy of GPS data is not sufficient. Navigation at presence and absence of an electronic compass. Change the direction of the movement (from first to last waypoint and vice versa) at any time. The users are notified via voice messages for the following: How many meters remain to the next waypoint. How much time and how many meters remain to the destination waypoint. Change of the status of the GPS data (transition from valid to invalid data and vice versa, insufficient accuracy, and etc.). Current accuracy of GPS receiver in meters. When there is a deviation from the track how many meters is the deviation and how the blind user should correct its course. How to change the course when reaching the next waypoint. 378

8 The destination waypoint has been reached. The voice waypoint has been reached. Since algorithm is adaptive to the velocity, it can be used not only for developing application for navigation of visually impaired and blind but also in GPS navigators in offroad and free modes. REFERENCES [1] Adams, C., An investigation of navigation process in human locomotion behaviour, Master thesis, Virginia Polytechnic Institute and State University, [2] Collins, C.C., On mobility aids for the blind, Electronic Spatial Sensing for the Blind (Warren, D. H. and Strelow, E. R., Eds.), Matinus Nijhoff, Dordrecht, The Netherlands, pp , [3] Hershberger, J., J. Snoeyink, Speeding up the Douglas-Peucker line simplification algorithm, In Proceedings of the 5th Int. Symposium on Spatial Data Handling, vol. 1, pp , [4] Kallie, C.S., P.R. Schrater, and G.E. Legge, Variability in stepping direction explains the veering behaviour of blind walkers, Journal of Experimental Psychology: Human Perception and Performance, 33(1): , [5] Kamiski, Ł., A. Stepnowski, J. Demkowicz, Wearable system supporting navigation of the blind, Int. Journal of Geology, 5(2), [6] Loomis, J., Digital Map and Navigation System for the Visually Impaired, University of California, Available: [7] Marston, J.R., J.M. Loomis, R.L. Klatzky, R.G. Golledge, and E.L. Smith, Evaluation of spatial displays for navigation without sight, ACM Transactions on Applied Perception, 3(2), pp , [8] Pressl, B., M. Wieser, A computer-based navigation system tailored to the needs of blind people, in K. Miesenberger et al. (eds.), ICCHP 2006, LNCS 4061, Springer Verlag Berlin Heidelberg, pp , [9] Sanchez, J.H., F.A. Aguayo, and T.M. Hassler, Independent outdoor mobility for the blind, Virtual Rehabilitation, pp , [10] The Loadstone GPS team, Loadstone GPS project, Available: [11] Tóth, B., G. Németh, Speech Enabled GPS Based Navigation System for Blind People on Symbian Based Mobile devices in Hungarian, Regional Conference on Embedded and Ambient Systems, pp , Hungary, ABOUT THE AUTHOR Assoc.Prof. Rosen Ivanov, PhD, Department of Computer Systems and Technologies, Technical University - Gabrovo, -mail: rs-soft@ieee.org. 379