Preliminar Assesment of Public Transportation Routes using free available Cloud Services and Belief Routes

Nelson Baloian1, Jonathan Frez2, José Pino1, Gustavo Zurita3 1Department of Computer Science, Universidad de Chile, Santiago, Chile

{nbaloian,jpino}@dcc.uchile.cl2Universidad Diego Portales, Santiago, Chile

 jonathan.frez@mail.udp.cl3Management Control and Information Systems Department, Universidad de Chile

gzurita@fen.uchile.cl

Abstract. Route planning for a bus line of the public transport system is a key issue in modern cities. The planning process should involve acquiring data about the estimated demand the route might have and the estimated time a bus will take between involves When planning a route for a bus or the line for a tram or subway it is necessary to consider the demand of the people for this service. In this work we presented a method to use existing crowdsourcing data (like Waze and OpenStreetMap) and cloud services (like Google Maps) to support a transportation network decision making process. The method is based the Dempster-Shafer Theory to model transportation demand and uses data from Waze to provide a congestion probability and data from OpenStreetMap to provide information about location of facilities such as shops, in order to predict where people may need to start or end their trip using public transportation means. The paper also presents an example about how to use this method with real data. The example shows how to analyze the current availability of public transportation stops in order to discover its weak points.

Keywords: Dempster-Shafer theory, transportation networks, smart cities.

1 Introduction and related work

Cities are constantly growing. Their decision making problems are increasingly complex, and better methods to evaluate solutions are needed in order to support this growing [1]. Many decision problems are spatial. A typical problem is to define an area to support a certain requirement or service, e.g., a place for a new road, set an industrial area, locate a hospital. Besides, cities are constantly changing, and they have dynamic problems, like in public transportation services: the routes can be dynamically defined to cope with new requirements and constraints. Some of the decisions may be to find a place for a new bus station, define a route, define multiple routes, or even plan a full transportation network.

According to [2], a “smart city” is a city that monitors and integrates data and information of its critical infrastructures, including roads, bridges, tunnels, rails, subways, airports, seaports to better optimize its resources, and maximize services to its citizens.

At same time, citizens are using information technologies (IT) to consume and provide data that can be used to support the decision making process of several cities requirements. Some of the IT used by citizens can be supported by Cloud Computing Services providing Software as a Service (SaaS). The SaaS model of Cloud Computing is often accessed by citizens from mobile applications and web interfaces [3]. Some of the SaaS services with spatial data properties are, e.g., Google Maps, OpenStreetMaps and Waze. These services provide geo-localized data in a graphical way, they are free, and they share a singular characteristic: they use Crowdsourcing data to provide data.

Here we report the use of services provided by Google Maps, OpenStreetMaps and Waze to develop a Spatial Decision Support System for transportation network planning, specifically for the Origin-Destination (OD) evaluation. The OD evaluation is done with the Dempster-Schafer theory[4]. This theory allows to model decisions based on uncertain and incomplete data, by studyng the extent a hypothesis can be supported by data.

In [6], authors describe a user-friendly web-based spatial decision support system (wSDSS) aimed at generating optimized vehicle routes for multiple vehicle routing problems that involve serving the demand located along arcs of a transportation network. The wSDSS incorporates Google Maps (cartography and network data), a database, a heuristic and an ant-colony meta-heuristic developed by the authors to generate routes and detailed individual vehicle route maps. It accommodates realistic system specifics, such as vehicle capacity and shift time constraints, as well as network constraints such as one-way streets and prohibited turns. The wSDSS can be used for “what-if” analysis related to possible changes to input parameters such as vehicle capacity, maximum driving shift time, seasonal variations of demand, network modifications, and imposed arc orientations. The system was tested for urban trash collection in Coimbra, Portugal.

A crowdsourcing database is the OpenStreetMap project [7]. Worldwide, several volunteers are contributing information to this “free” geodatabase. In some cases this database exceeds proprietary ones by a 27% [8], and for some authors [9] it is more complete than Google maps or Bing maps. OpenStreetmap data has been proposed to support traffic related decisions by developing traffic simulations [10], or solutions to achieve a new web-based trip optimization tool [11]. It has also been used to support transportation network planning [12]. Also, a spoken-dialogue prototype for pedestrian navigation in Stockholm by using various grounding strategies and based on OpenStreetMaps is described in [13]. Similarly, Jacob et al. [14] present a web-based campus guidance system for pedestrian navigation aimed at providing support for visitors. They developed an OpenStreetMap based system to generate short paths using both outdoor walkways and indoor corridors between various locations.

Another popular crowdsourcing geodatabase is being generated by Waze. Waze is a mobile GPS application that allows to measure and report traffic conditions and events, e.g., it automatically detects traffic jams, users can report accidents, weather effects on the roads, and other alerts. In the literature we did not find a decision support system using

Waze data, maybe because it is hard to obtain. However, we found traffic condition analysis systems [15] based on real time data obtained from Waze.com using a WebCrawler, and an accidents data mining analysis proposal [16] based on the same real time data from Waze.com. In our work the data is obtained using the same technique: we developed a WebCrawler to reconstruct an historical database based on published data on waze.com.

We focus on spatial DSS using belief functions [17], in particular Dempster-Shafer theory (DST). The DST proposes to use sets of hypotheses regarding a variable (e.g. the temperatures at a location are between t1 and t2) associated with a probability of being correct. Using belief functions we can provide a “hypotheses support value” called belief. The belief can be assigned to a certain geographical area satisfying a hypotheses set.

2 The OD route problem

Regional transportation networks are composed of various transportation lines designed to cooperate and complement an urban scale transportation solution [18]. The planning of the paths or routes of a new transportation method is usually based on existing network data, volume predictions and the distribution in the network [19].

When a decision maker chooses a route, the travel time and time reliability are important factors under demand and supply uncertainty. When designing an urban route for a new transportation service, the choices must consider the behavior and reliability of the transportation network. Another factor is the OD traffic demand [21]. The OD describes the traffic demand between a particular OD during a time period.

A public transportation system is typically a complex network. In Santiago it includes bus stations and routes, subway stations and routes, shared taxi stations and routes. Each OD is composed of a start station and an ending station. A single OD can have multiple sub-OD in a single route. The design of a public transportation system is a task requiring analysis of the transportation demand, the traffic conditions of the possible routes for each OD, and the reliability of the OD. In order to define an OD route based on uncertain demand information, we propose to adapt belief maps [17]. This concept is based on the DST. A belief map allows to evaluate a geographical area generating a suitability evaluation on a set of hypotheses supporting a possible solution. For example, a map can show the belief degree of find people in each evaluated area. The hypotheses supporting this map can be “People go to commercial areas”, or “People can be in schools”.

3 Determining the OD route

We propose a new concept called belief routes (BR). A BR can be used to evaluate demand hypotheses of an OD. A BR is composed of 3 basic elements: A set of hypotheses

that defines a possible transportation demand of an OD; The Origin and Destination; The “polyline” or path of the route. Using this definition of BR, the decision maker can compare various paths satisfying the demands in each OD. Also, the decision maker can adjust the Origin or Destination. Another factor in the OD evaluation is the transportation and reliability time. In order to support both indicators we propose to combine the traffic information from Waze creating a belief value based on historical data. We call the result of this combination a Belief Congestion Route (BCR).

In order to explain the use of the BR and BCR in the decision making processes we are going to use a basic example: A single OD with two possible paths. In Figures 1 and 2, two options are shown (A and B respectively). In this example route A is shorter than B, and the travel time is also shorter according to Google Directions API.

Fig 1. Route A Fig 2 Route B

In Figs. 3 and 4, the transportation demand is represented by a BR, according to OpenStreetMap and the proposed hypothesis, route B has more demand than route A.

Fig 4 Belief route A Fig 5 Belief route B

In Figs. 5 and 6 the BCR of both routes are shown. According to the Waze information for both paths, route A has more belief of having congestions or traffic jams which implies less reliability.

Fig 6 Congestion route A Fig 7 Congestion route B

From this example we can note that route B has less congestion and more demand than route A. However route A is shorter and the decision will depend of what kind of OD the decision maker is looking for. In order to support the decision, the visual evaluation of the BR and BCR is not enough. An evaluation metrics framework is needed and it will be part of our future work.

4 An Application for developing Belief Routes

We developed a prototype that allows its users to define an OD pair and it automatically provides the shortest route using the Google Route Service. It also allows the user to specify hypotheses for transportation demand modelling, after which it can generate two types of visualization: The demand Belief Route and the Congestion route (see Fig 8). The platform provides Average Belief of the generated 3D visualization.

Fig. 8. Evaluation of an OD using the developed application

The application allows setting a transportation demand hypotheses set compatible with the Dempster-Shaffer Theory. It also allows including some model restrictions, for

example: avoid schools. After the hypotheses are included, the application allows choosing the type of 3d map which will be generated and shown: BR or BCR.

4 Belief Routes in Real World

In order to test the proposed concept, we are going to use real data from a public transportation system. The testing method is simple: we use real data to evaluate an hypothesis set (used to build the BR), if the prediction generated by the hypothesis set “matches” with the real data we assume the hypotheses hold and thusthe generated BRs are valid.

For the test we selected two different areas with high transportation activities in the city of Santiago de Chile. These areas were selected because one is representative for having many shops in the city center and the other is representative for a residential area with high population. Both areas have an important number of subway and bus stations. For these transportation methods we have the information about the time and location were people starts using each service. This was possible because the integrated public transportation system of Santiago (called “Transantiago”) uses exclusively plastic cards with magnetic bands which should be pre-loaded with money as payment method. In this way, the system registers the point where every passenger started her trip. However, it does not register the point where it ends. Data were obtained from the Ministry of Transport.

Let’s call “A” to the area in the city center, and B to an area that leads to residential areas with a high population. We first used the data for generating a heatmap coloring with intensity from 0 to 1 the areas where bus and subway stops are concentrated. The generated heatmaps showing the concentration of public transportation stops are shown of Figure 9.

Fig. 9. Transportation network spatial distribution. Left: "A”. Right: “B”.

In a similar way, we used the data for generating a heatmap according to the number of people starting their trips at a certain bus or subway stop. The generated maps for both areas are shown in figure 10.

Fig. 10. Colored areas show spots where people start their trip using public transportation. Left: "A”. Right: “B”.

From figures 9 and 10 we can conclude that people concentrate their starting point at fewer stations than the available, which is something we would expect. This does not mean they are not using the other stations since they can be used to get off at them, but we cannot collet this data.

In order to generate scenarios which allow us to calculate belief routes which are usable for the travelers we have to propose hypotheses that generate scenarios according to the reality. This means, we should construct hypotheses that predict at least where

people are going to start their trips with public transport. For this, we will propose two hypotheses which relate elements in the city with transportation needs for the people:

1. People start their trips using public transport means near “shops”.2. People start their trips using public transport means near “amenities”.

Both terms, “shops” and “amenities”, are standard tags for labelling geographical objects in OpenStreetMap. The definition of “amenity” covers “an assortment of community facilities including toilets, telephones, banks, pharmacies and schools.”

The results of applying of both hypotheses to area “B” can be seen in Figure 11. Comparing these scenarios with Figure 10 right, we can note that shop places coincide in both maps with minimal differences. On the contrary, for amenities there is almost no match.

Fig. 11 Belief map applying the hypotheses for "B" area separately. Left: Amenity places. Right: Shop Places.

From these results, we can conclude that the hypothesis that shop places generate trips starting in their near environment is a better predictor. (see figure 12).

Fig. 12 Comparison at "B". Left: People concentration. Right: Shop Places.

When applying the “shops” hypothesis to the A area we find only some coincidence between the starting points and the shops. This is however explained by the fact that there is no bus and subway stations near some shops, as can be seen comparing the maps in figure 9 left, showing the stops, and figure 13 left showing the shops. In this way we can estimate a lack of proper public transportation stops in an area which according to the hypothesis should have a great demand of them.

Fig. 13 Comparison at "A". Left: People concentration. Right: Shop Places.

5 Conclusions

In this work we presented a method to use existing crowdsourcing data to support a transportation network decision making process. The method uses the Dempster-Shafer Theory to provide a framework to model transportation demand based on the OpenStreetMap Data. The method also provide a simple way of use the Waze application data to provide a congestion probability value to each segment of a route.

In this work we propose that the use of croudsourcing data to build the transportation demand metric and the congestion probability it is possible to support a transportation network decision making process.

References

1. Heilig, G.K., World urbanization prospects the 2011 revision. United Nations, Department of Economic and Social Affairs (DESA), Population Division, Population Estimates and Projections Section, New York, 2012.

2. Harrison, C., et al., Foundations for smarter cities. IBM Journal of Research and Development, 2010. 54(4): p. 1-16.

3. Chourabi, H., et al. Understanding smart cities: An integrative framework. in System Science (HICSS), 2012 45th Hawaii International Conference on. 2012. IEEE.

4. Shafer, G. (1976). A mathematical theory of evidence (Vol. 1). Princeton: Princeton university press.

5. Piro, G., et al., Information centric services in Smart Cities. Journal of Systems and Software, 2014. 88: p. 169-188.

6. Santos, L., J. Coutinho-Rodrigues, and C.H. Antunes, A web spatial decision support system for vehicle routing using Google Maps. Decision Support Systems, 2011. 51(1): p. 1-9.

7. Haklay, M. and P. Weber, Openstreetmap: User-generated street maps. Pervasive Computing, IEEE, 2008. 7(4): p. 12-18.

8. Neis, P., D. Zielstra, and A. Zipf, The street network evolution of crowdsourced maps: OpenStreetMap in Germany 2007–2011. Future Internet, 2011. 4(1): p. 1-21.

9. Ciepłuch, B., et al. Comparison of the accuracy of OpenStreetMap for Ireland with Google Maps and Bing Maps. in Proceedings of the Ninth International Symposium on Spatial Accuracy Assessment in Natural Resuorces and Enviromental Sciences 20-23rd July 2010. 2010. University of Leicester.

10. Zilske, M., A. Neumann, and K. Nagel. OpenStreetMap for traffic simulation. in Proceedings of the 1st European State of the Map–OpenStreetMap conference. 2011.

11. Klug, M. CS Transport-Optimisation–A Solution for Web-based Trip Optimization Basing on OpenStreetMap. in 19th ITS World Congress. 2012.

12. Joubert, J.W. and Q. Van Heerden, Large-scale multimodal transport modelling. Part 1: Demand generation. 2013.

13. Boye, J., et al., Walk this way: Spatial grounding for city exploration, in Natural Interaction with Robots, Knowbots and Smartphones. 2014, Springer. p. 59-67.

14. Jacob, R., et al., Campus guidance system for international conferences based on openstreetmap, in Web and Wireless Geographical Information Systems. 2009, Springer. p. 187-198.

15. Silva, T.H., et al., Traffic Condition Is More Than Colored Lines on a Map: Characterization of Waze Alerts, in Social Informatics. 2013, Springer. p. 309-318.

16. Fire, M., et al. Data mining opportunities in geosocial networks for improving road safety. in Electrical & Electronics Engineers in Israel (IEEEI), 2012 IEEE 27th Convention of. 2012. IEEE.

17. Frez, J., et al., Dealing with Incomplete and Uncertain Context Data in Geographic Information Systems, in Computer Supported Cooperative Work in Design (CSCWD), IEEE, Editor 2014, IEEE: Hsinchu, Taiwan. p. 129-134.

18. Yang, L. and B. Wan. A Multimodal Composite Transportation Network Model and Topological Relationship Building Algorithm. in Environmental Science and Information Application Technology, 2009. ESIAT 2009. International Conference on. 2009. IEEE.

19. Liu, S., et al. Modeling and simulation on multi-mode transportation network. in Computer Application and System Modeling (ICCASM), 2010 International Conference on. 2010. IEEE.

20. Xu, L. and Z. Gao. Bi-objective urban road transportation discrete network design problem under demand and supply uncertainty. in Automation and Logistics, 2008. ICAL 2008. IEEE International Conference on. 2008. IEEE.

21. Castillo, E., et al., The observability problem in traffic models: algebraic and topological methods. Intelligent Transportation Systems, IEEE Transactions on, 2008. 9(2): p. 275-287.