Travel time prediction for buses in Rio de Janeiro : incident recognition and travel time change caused by incidents

Haarsman, Giel (2009) Travel time prediction for buses in Rio de Janeiro : incident recognition and travel time change caused by incidents.

Abstract:In Rio de Janeiro more people are choosing for individual transport and this results in a decrease in bus passengers and other negative consequences. To let the passengers choose for the bus, IFluxo is developing a service which gives the potential bus passenger information via text message about the departure time of a bus of interest. Three important aspect in this project are: 1) Stakeholder demands and wishes 2) Software for the text message service 3) Estimating the time of arrival of the buses In this thesis the first and third aspect are viewed. Demands for the public transport operator and the potential passenger have to be matched and the input of one influence the output of the other en the other way around. The estimation of the arrival time of buses in Rio de Janeiro consists of two parts, the continuous prediction and the incidental prediction. This thesis will start with the analysis of the incidental prediction. The objective of this research is to develop a table for all possible incidents during bus rides, whereby the table is quantified in order to determine the nature of the incident and as a result of that estimate a reliable time of arrival. In this thesis the events on a bus ride are viewed, as well visual as with a GPS Travel Recorder. Incidents and irregularities are tried to be recognized and attempted to be distinguished in the data and the influence on the travel time is examined. Result of the literature research where as follows. One can say that there is done a lot of research on the subject of predicting arrival times, but every research uses other types of data collecting and most researches have their focus on the continuous part of travel time prediction and not on the incidental part. In this thesis, because of the lack of a big GPS database, the focus will be on the change in the arrival time caused by incidents, but as well caused by irregularities. Herby, incidents are non-recurring and irregularities are recurring traffic events. To recognize the irregularities and incidents in the GPS data and to give the delay caused by them first, all travel times of the different bus lines (2 lines, 2 directions) had to be matched to the same Points of Interest (PoI) which are in this case bus stops. This is called map-matching and this is done in combination with the nearest neighbourhood method. This method chooses the nearest point to the PoI and allocates this point to the PoI. Finally, this will result in average sections times between the points of interest. Results of the analysis of recognizing the irregularities and incidents in the GPS data, separated from the visual data are negative. It is impossible to recognize the irregularities or incidents in the GPS data, because first, one incident / irregularity has the same characteristics as another incident / irregularity. Second, the incident / irregularity has within the incident / irregularity different characteristics. Because incidents and irregularities can not be distinguished in the GPS data, the delays are predicted using both, GPS data and visual data. Only (1) ‘a lot of passengers getting in or out the bus’, (2) ‘road maintenance’ and (3) ‘congestion’ occurred. Incident (1) always occurs in the same section, incident time (time difference with the average section time) is small and fluctuations are minimal. This incident can be included in the continuous prediction. Incident (2) sometimes occurs in the same section, incident time is short and fluctuations are minimal. This can be included in the continuous prediction. However, there are times, appearing in the data, that road maintenance not occur in the same sections and incident times are big, varying from one to two minutes. Incident (3) always occurs in the same section, but incident time is big and is fluctuating from one minute to 7,5 minutes. Because only three kinds of incident occur in the data an interview is held. This interview gives more insights in the procedures, number of occurrence and incident times of the other incidents. All irregularities can be taken within the continuous prediction. When more data is available the incidents can maybe distinguished and the delay can be predicted, but in this stage it is not possible.
Item Type:Essay (Bachelor)
IFluxo, Rio de Janeiro, Brazil
Faculty:ET: Engineering Technology
Subject:56 civil engineering
Programme:Civil Engineering BSc (56952)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page