An Alternative and Smarter Route Planner for Wheelchair Users

Exploring Open Data

adia P. Kozievitch, Leonelo D. A. Almeida, Ricardo Dutra da Silva and Rodrigo Minetto

UTFPR, Department of Informatics, Curitiba, Paran

a, Brazil

Keywords:

Route Planner, Wheelchair Users, GIS, Open Data.

Abstract:

In this paper we describe a bottom-up approach to integrate GIS maps (endorsed by discrete features, such as

points, lines, polygons), in order to develop a route planner for wheelchair users. We integrate public available

data with a novel model for route planning, based on sidewalks, crosswalks and curb ramps, as opposed to

traditional street-based approaches. We show that our sidewalk-based model is more suitable than available

planning routes under mobility constraints, using a case study in Curitiba, Brazil.

1 INTRODUCTION

Efforts have been made to develop route plan-

ning services for people with disabilities (Kulakov

et al., 2015; Menkens et al., 2011; Sumida et al.,

2012). This is a challenging problem that involves

the processing of a huge amount of data, such as

maps, images, detailed public transport information

and collaborative user feedback, to properly deﬁne

wheelchair accessible paths. Nevertheless, there is

a lack of route planning services (such as the route

planner of Google maps

) for wheelchair users.

The available solutions for wheelchair users are

street-based. Nevertheless, a myriad of complex spa-

tial factors could be considered in order to locate the

best routes, such as sidewalks, crosswalks and curb

ramps. In many applications, such as geographic in-

formation systems (GIS), data can set the stage by dis-

playing individual maps of decision criteria, in order

to provide the detailed information needed to locate

the best route.

In this paper we are concerned with the routing

problem for wheelchair users. The input data for this

problem is a set of base map layers (streets, sidewalks

and city blocks). The output data is a cost weighted

map, which we also present as a graph. The novelty is,

taking advantage of open data, to include additional

factors (such as distinct sidewalks in a street or miss-

ing curb ramps), in order to propose a best route to a

wheelchair user. We explore our method using open

data from Curitiba, Brazil. The city belongs to the

https://maps.google.com Last visited on 01/07/2015.

C40 cities

, a group which set ambitious targets to im-

prove urban life quality and protect their environment.

The remainder of this paper is organized as follows.

Section 2 contains a description of related work. Sec-

tion 3 presents an overview of our method. Section 4

presents the experiments and, ﬁnally, section 5 state

the conclusions and future work.

2 RELATED WORK

Several online libraries and services are already avail-

able for route planning (such as Google Directions

API

, the JavaScript API Yandex.Map

, or other on-

line services

). In particular, fewer of them are

toward speciﬁc functionality, such as wheelchairs

(as OpenRouteService

, Routino

,and OpenTripPlan-

ner

). Nevertheless, some of the online libraries also

present acessibility problems (Medina et al., 2015).

There are norms for urban environments that de-

ﬁne accessibility speciﬁcations as DIN 18024-1 (Ger-

http://www.c40.org Last visited on 30/05/2015.

https://developers.google.com/maps/documentation/

directions/ Last visited on 23/06/2015.

https://tech.yandex.ru/maps/ Last visited on

23/06/2015.

http://wiki.openstreetmap.org/wiki/Routing/

online routers Last visited on 23/06/2015.

http://www.rollstuhlrouting.de/ Last visited on

23/06/2015.

http://www.routino.org/ Last visited on 23/06/2015.

http://www.opentripplanner.org/ Last visited on

23/06/2015.

Kozievitch, N., Almeida, L., Silva, R. and Minetto, R.

An Alternative and Smarter Route Planner for Wheelchair Users - Exploring Open Data.

In Proceedings of the 5th International Conference on Smart Cities and Green ICT Systems (SMARTGREENS 2016), pages 94-99

ISBN: 978-989-758-184-7

man Institute for Standardization , 1998) from Ger-

many and ABNT NBR 9050:2004 (ABNT (Brazil-

ian Association of Technical Standards), 2004) from

Brazil. Other studies as Kasemsuppakorn and

Karimi (Kasemsuppakorn and Karimi, 2009) investi-

gated the most relevant aspects regarding wheelchair

accessibility, for each segment of the route network

(i.e. length, width, slope, sidewalk surface, steps,

sidewalk conditions and sidewalk trafﬁc). More-

over, some studies consider not only barriers but also

provide information about relevant locations (called

Points of Interest - POI) as restaurants, bus stops, ac-

cessible toilets, and police departments (e.g. Menkens

et al. (Menkens et al., 2011), and Wheelmap

Sumida et al. (Sumida et al., 2012) argues that bar-

riers information should be collected through actual

measurement data and adapted an electric wheelchair

for collecting data as the force necessary to move the

wheelchair and the passage width. Other studies use

data provided by the routing service users (e.g. Open-

RouteService, and Menkens et al. (Menkens et al.,

2011)), volunteers (e.g. OpenRouteService; Menkens

et al. (Menkens et al., 2011); Kulakov et al. (Kulakov

et al., 2015)), authorities (e.g. Kulakov et al. (Ku-

lakov et al., 2015)). Consequently, the collected data

is usually limited to a city or a district. An exception

is the Wheelmap project

that provides resources for

crowd sourcing information based on OpenStreetMap

data.

The lack of public available information regarding

accessibility of urban spaces (e.g., streets, sidewalks,

curbs, and type of surface) contribute to the fact that

most of studies on accessible routing are constrained

to cities or districts. Some initiatives as Wheelmap

explore the aspects of crowds for collecting relevant

information regarding POI. Currently, to our knowl-

edge, none of such initiatives are widely adopted in

Curitiba (e.g. there are only about 10 POI registered

in Wheelmap, most of them are accessible bus stops).

Routing estimation is another challenging task. In

order to improve the results of routing estimation,

some services provide additional parameters (e.g.,

maximum inclination, type of surface, maximum curb

height

), personalized estimation according to the

users’ proﬁle (e.g. Menkens et al. (Menkens et al.,

2011)). As already noted by A. M. Bishop within

the Routino application, routing planning can use

graph algorithms (e.g. Dijkstra and A*) for calculat-

ing the shortest/least-cost path. Some of those stud-

ies adapt these algorithms aiming at improving per-

formance in terms of execution time (e.g. the con-

cepts of super-segments and super-nodes from A. M.

Bishop

). Others claim that there is no clear answer

http://wheelmap.org/en. Last Visited 24/08/2015.

as to shortest path algorithm which runs fastest on

real road networks (due to real time computation, the

large network size, and the resulting intensive com-

puting) (Zhan, 1997). Among other critical factors

for route planning we can mention: (1) types of bar-

riers considered for collecting and estimating routes,

(2) data sources for maps information and barriers,

and (3) approaches for route planning.

3 OUR METHOD

Problem Formulation. For GIS applications, the

shortest path based on a road is a basic operation.

In practice, however, users are always interested in

several constraints (such as the combination of spatial

and textual information). From GIS and map analysis

perspective, the routing problem can be described as

a three steps process: the calculation of discrete cost,

accumulated cost and steepest path (Berry, 1993).

The idea is, using base maps (such as roads), create

other derived maps (to calculate information that is

too difﬁcult to collect, such as curb ramps) in order

to ﬁnally create cost/avoidance maps which translate

this information into decision criteria. Within this per-

spective, map layers are thematic representations of

geographic information, as shown in Figure 1. In par-

ticular, base maps can be represented by streets, street

blocks (Figure 5) and sidewalks, among others.

The derived maps, composed by large polygon

subdivisions, are often simpliﬁed in order to reduce

the total number of vertices which deﬁnes it (known

as the map simpliﬁcation problem (Estkowski and

Mitchell, 2001)). The calibration of the individual

cost maps is an important and sensitive step in the sit-

ing process. Since the computer has no idea of the rel-

ative preferences this step requires human judgment.

As an example, you might be interested in identify-

ing the most preferred route for a wheelchair user that

minimizes its visual exposure to huge avenues, and

maximizes the visual exposure to bus stops.

From the geometry perspective, streets are one or

more single lines (which can also be represented by

one ore more edges in a graph), which thereby are

composed by points (which can also be represented

by a vertice in a graph). Blocks (Figure 5) can be

decomposed by the respective lines, and lines can be

decomposed by respective points (Figure 6).

Formally, a GIS database contains a set of geome-

tries which can be resumed as a set P of points on a

network G = (V, E), where V represents vertices and

E represents edges. The network is a directed con-

nected graph. A point p ∈ P locates on an edge e ∈ E.

The distance between any two points (or vertices) can

An Alternative and Smarter Route Planner for Wheelchair Users - Exploring Open Data

Figure 1: Map Layers

Figure 2: Overview of the proposed system.

be denoted as d(p

, p

), as the length of the shortest

path connecting them.

Our Approach. Our method comprised the follow-

ing steps (shown in Figure 2) toward a sidewalk-based

model: (1) the base map (streets, sidewalks, blocks,

etc.) characterization; (2) the creation of derived map

for sidewalk polygon simpliﬁcation; (3) the creation

of derived map for location of intersection areas; (4)

the creation of derived map for hypothetical cross-

roads; (5) the establishment of a weighting criterion

to edges (in particular, considering wheelchair users);

and (6) the generation of the graph.

For the characterization of the base maps (iden-

tiﬁed by Figures 5 and 6), external open data were

integrated with data from Open Street Map with a spa-

tial database.

The second step started with the simpliﬁca-

tion of sidewalk polygons and their derived points

(performed with a spatial database function named

st simplify).

The next derived map comprised the road inter-

section areas, basically discovering the average dis-

tance among streets and sidewalks. Circles in Figure

7 represent this phase, performed with spatial func-

http://webhelp.esri.com/ Last access on 17/09/2015.

tions, such as st buffer.

The last derived map comprised the hypothetical

crossroads, identiﬁed by lines, triangles and rectan-

gles inside the road intersection area (circles). Ba-

sically this map was derived using spatial database

functions such as st convexhull (shown in Figure 3).

The weighting criterion (step 5) was initially set

as the distance between points if the ”edge” has no

barriers, or inﬁnite if it does. Lately these weights

could be modiﬁed to register temporal issues, such as

constructions which might impact the normal ﬂow of

pedestrians. In practice, these distances can be ob-

tained with spatial functions, such as st distance.

Figure 3: Example of SQL query to compute crosswalks.

From the GIS perspective, the shortest path would

not be based on the street map, but on the compo-

sition of the sidewalk map and on the derived map

for the crossroads. Note that: (i) the same sidewalks

could have temporally distinct costs (if for example,

one sidewalk is next to a construction during the be-

ginning of an year) and (ii) the distinct crossroads ge-

ometries could help the identiﬁcation of most danger-

ous locations for pedestrians.

4 EXPERIMENTS

We concentrated our efforts in preliminary experi-

mental results, using an ordinary linux server, and ex-

ploring spatial database functions. In order to sim-

plify our tests, we explore the shortest path as a

graph (step 6), but indeed, there are already ongoing

projects (such as pgRouting

) which could be used

within the spatial database.

Datasets. In particular, we are exploring a case

study in Curitiba. Curitiba has 1.8 million people in-

side a total area of 430,9 km

, IDH of 0,823, accord-

ing to the Brazilian Institute of Geography and Statis-

tics (IBGE)

. This area encompasses 75 neighbor-

hood districts. According to the same census, Curitiba

had in the urban region approximately 95 thousand

people with some degree of motor disability. Among

them, more than 31 thousand informed to experience,

http://pgrouting.org/ Last visited on 23/09/2015.

http://www.ibge.gov.br. Last visited on 14/05/2015.

SMARTGREENS 2016 - 5th International Conference on Smart Cities and Green ICT Systems

at least, great difﬁculty for locomotion. Despite of not

providing more detailed information regarding barri-

ers for motor impaired people, the Census revealed

that only 12.6% of the urban area of Curitiba pre-

sented lowered curbs. Thus, currently, providing ef-

fective and efﬁcient route plannings for people with

motor disabilities, especially wheelchair users, is po-

tentially relevant for removing barriers to people’s

daily lives.

For the acquisition/characterization of the base

maps, the dataset from the Institute of Research and

Urban Planning of Curitiba (IPPUC

), along with

data from Open Street Map

were used. Figure 4

shows the DDL example for the creation of street

block table. The complete data set initially included

streets (39,948 rows), sidewalks (9,614 rows) and

blocks (13,459 rows) from Curitiba. The complete

data set was inserted in a PostGIS

database. Later,

speciﬁc tablespaces and indexes were created in or-

der to optimize the access. Nevertheless, several se-

mantic errors were present (such as different street ge-

ometries and names - more details in (Barczyszyn,

2015)).The data was visualized with QGIS

Figure 4: Example of SQL for street block creation.

Figure 5: Example of Street Blocks.

Tests. For our tests, the three base maps toward the

selected area of the Batel district resulted in 51 streets,

http://www.ippuc.com.br. Last visited on 14/05/2015.

http://www.openstreetmap.org Last visited on

14/05/2015.

http://www.postgis.net Last visited on 15/05/2014.

http://www.qgis.org Last visited on 15/05/2015.

Figure 6: Example of Sidewalk Points.

Figure 7: Example of Road Intersection Areas.

Figure 8: Hypothethical Crossroad.

94 sidewalks and 71 blocks. The derived maps re-

sulted in 309 road intersection areas (within a zoom

in Figure 7) and 185 number of possible crossroads

(within a zoom in Figure 8). Note that crossroads

within Figure 8 can belong to geometric groups: lines

and polygons (with three or four edges).

Consider the shortest walking path (highlighted in

Figure 9) from point A on Benjamin Lins Street to

point B on Vicente Machado Avenue. The path is

computed by considering the possible combination of

streets that can be followed in order to get from one

point to another. A subset of the combinations can

be modeled by a graph as the one in Figure 9. The

vertices are a set of intersections and the edges are

segments of streets deﬁned by two intersections.

The graph reﬂects the input data used for comput-

An Alternative and Smarter Route Planner for Wheelchair Users - Exploring Open Data

134

205

131

199

105

178

119

111

179

187

135

183

128

102

125

176

Figure 9: Shortest walking path between points A and B

computed using a street-based graph.

134

205

131

199

105

178

119

111

179

187

135

183

128

102

125

176

Figure 10: Ordinary route planners do not consider accessi-

bility constraints.

ing shortest paths on common route planning. This is

a street-based graph which takes into account a weight

factor for each edge according to known properties of

street segments, such as the distance between two in-

tersections, the slope of a street segment and the con-

dition of a sidewalk. The graph in Figure 9 considers

distances between intersections.

People with disabilities, such as wheelchair users,

may face some problems due to conditions not cap-

tured in route planners considering such a model. As

highlighted in Figure 10, there is a crossing without

curb ramps that would make it difﬁcult to proceed

along the route. The model must be able to encode

that kind of constraints perhaps by using a weight on

the vertices of the graph or by making it reﬂect on

the weight of the streets. Anyway, those can also be

ﬂawed solutions: it is possible one street has side-

walks on both sides and one side has problems while

the second one is in perfect conditions to go along.

The sidewalk-based model proposed in this paper

allows to make all of these conditions explicitly avail-

able for a route planner, making it more suitable for

searching paths that meet the needs of people with

disabilities. The model for the same region of the

previous discussion is summarized by the graph in

Figure 11. The hypothetical curb ramps are vertices

of the graph and its edges are sidewalks or adjoining

curb ramps on streets. Thereafter, every sidewalk can

have its own weight and the lack of curb ramps be-

tween adjoining streets also be encoded on the edges

connecting them. It is possible, therefore, to avoid

streets without curb ramps and sidewalks on poor con-

ditions. The shortest path in our model is highlighted

in Figure 11, depicting not only which sidewalk to fol-

low along a street but also which curb ramps to use.

125

218

193

184

283

172

170

173

105

173

181

177

129

177

129

224

128

199

119

244

125

170

Figure 11: A sidewalk-based model and a route considering

accessibility.

125

218

193

85+50

184

283

172

170

173

113

131

105

173

181

177

129

177

129

224

128

199

119

125

170

Figure 12: Modiﬁed walkway-based model: a sidewalk was

given a bigger weight (see region A) and a curb ramp, along

with a crosswalk, was added (see region B).

The model may have its attributes modiﬁed by

users: (i) an edge (segment of a sidewalk) of the

graph could have its weigh increased if the condition

of the sidewalk is degraded due to holes in the pave-

ment, narrow width, obstruction, among others; and

SMARTGREENS 2016 - 5th International Conference on Smart Cities and Green ICT Systems

(ii) previously nonexistent curb ramps and their re-

lated crosswalks could be added to improve the route

planning. The two types of modiﬁcations are illus-

trated in Figure 12. The weight of one edge was mod-

iﬁed due to a possible poor condition of the sidewalk.

In the second modiﬁcation a vertex and an edge were

added to the model to illustrate that a new curb ramp

has been built.

Discussion. Among the parameters which impact

the planning of a route for wheelchair users, we can

mention: Database Issues: lacks of standard within

the data sources (the different data semantics, geome-

tries, etc.), the domain understanding of how GIS ge-

ometries might be mapped and how they can be ex-

plored in other domains (such as graphs), the theoretic

abstraction of how different layers of points, lines, ge-

ometries can be mapped to multigraphs, among oth-

ers; Domain Issues: how to manage different areas

(such as databases, human computer interaction, the-

ory, urbanism, etc.) toward working in an uniﬁed

solution, regarding terms and technologies; Theory:

how to theoretically add temporal events which might

change in a graph (such as edge weights which change

along the interactions); how to develop a mathemati-

cal approach and elaborate algorithms of route build-

ing on the graph on condition of various types of ob-

stacles; how to better calculate the edge weight es-

timation. Although being preliminary, the tests state

that the implementation of the method is suitable for

wheelchair users. In summary, we understand that

better approaches to the routing problem depends on a

sinergic combination of data, knowledge domain, and

user feedback.

5 CONCLUSIONS

Research in accessibility and GIS is not recent, but the

exploration through different domains using real data

is still an ongoing effort. The possibility of imple-

menting models within GIS and integrate them with

different sources provides planners with a powerful

and ﬂexible tool for analyzing applications, and de-

ciding on new business permits. This paper presents

the concepts, application, and challenges of exploring

a route planner for wheelchair users. Later these def-

initions are explored in a practical case study, within

the Curitiba metropolitan region, Brazil.Future work

includes the personalization of weights for different

users, and integration of a off-line mode, among oth-

ers.

ACKNOWLEDGEMENTS

We would like to thank the Curitiba City Municipal-

ity, IPPUC, and RNP.

REFERENCES

ABNT (Brazilian Association of Technical Standards)

(2004). Abnt nbr 9050:2004: Accessibility to build-

ings, equipment and the urban environment.

Barczyszyn, G. L. (2015). Integration of geographic data

for the urban planning of the city of Curitiba (in

portuguese). Universidade Tecnol

ogica Federal do

Paran

Berry, J. K. (1993). Beyound Mapping - Concepts, Algo-

rithms and Issues in GIS, volume 1. John Wiley Pub-

lishers, 1 edition.

Estkowski, R. and Mitchell, J. S. B. (2001). Simplifying

a polygonal subdivision while keeping it simple. In

SCG ’01, pages 40–49, New York, NY, USA. ACM.

German Institute for Standardization (1998). Barrier-free

design - part 1: Streets, places, roads and recreational

areas; planning basics. Standard Speciﬁcation 18024-

1, German Institute for Standardization), Berlin.

Kasemsuppakorn, P. and Karimi, H. A. (2009). Person-

alised routing for wheelchair navigation. J. Locat.

Based Serv., 3(1):24–54.

Kulakov, K., Shabaev, A., and Shabalina, I. (2015). The

route planning services approach for people with dis-

ability. In FRUCT ’15, pages 89–95.

Medina, J. L., Cagnin, M. I., and Paiva, D. M. B. (2015).

Evaluation of web accessibility on the maps domain.

SAC ’15, pages 157–162, New York, NY, USA. ACM.

Menkens, C., Sussmann, J., Al-Ali, M., Breitsameter, E.,

Frtunik, J., Nendel, T., and Schneiderbauer, T. (2011).

Easywheel - a mobile social navigation and support

system for wheelchair users. In ITNG ’11, pages 859–

866.

Sumida, Y., Hayashi, M., Goshi, K., and Matsunaga, K.

(2012). Development of a route ﬁnding system for

manual wheelchair users based on actual measure-

ment data. In UIC/ATC ’2012, pages 17–23.

Zhan, F. B. (1997). Three fastest shortest path algorithms

on real road networks: Data structures and procedures.

Journal of geographic information and decision anal-

ysis, 1(1):69–82.

An Alternative and Smarter Route Planner for Wheelchair Users - Exploring Open Data