This document discusses a method for dynamically forecasting bus arrival times to stations using historical Automatic Vehicle Location (AVL) data. Key factors like time of day, day of week, weather, holidays and school days that influence bus speeds are classified. Bus movement along each route is represented as positions in a "cylinder" with segment and time. Forecasts of arrival times for each bus to future segments are stored in a dynamic table lookup. Initial results show a 10% error rate within a 30 minute forecast window and 95% confidence level. Testing is still needed during the rainy season.
Online Bus Arrival Time Prediction Using Hybrid Neural Network and Kalman fil...IJMER
International Journal of Modern Engineering Research (IJMER) is Peer reviewed, online Journal. It serves as an international archival forum of scholarly research related to engineering and science education.
International Journal of Modern Engineering Research (IJMER) covers all the fields of engineering and science: Electrical Engineering, Mechanical Engineering, Civil Engineering, Chemical Engineering, Computer Engineering, Agricultural Engineering, Aerospace Engineering, Thermodynamics, Structural Engineering, Control Engineering, Robotics, Mechatronics, Fluid Mechanics, Nanotechnology, Simulators, Web-based Learning, Remote Laboratories, Engineering Design Methods, Education Research, Students' Satisfaction and Motivation, Global Projects, and Assessment…. And many more.
Predicting Operating Train Delays into New York City using Random Forest Regr...AI Publications
The Long Island Railroad operates one of the largest commuter rail networks in the U.S.[1]. This study uses data which includes the location and arrival time of trains based on onboard GPS position and other internal sources. This paper analyzes the GPS position of the train to gain insight into potential gaps in on time performance and train operations. This was done by developing a Random Forest Re-gression model [2] and an XGBoost regression model [3[. Both models prove to be useful to make such predictions and should be used to help railroads to prepare and adjust their operations.
passSepnegceirasl. pIut rcpoonsesi sotns lionfe tbwuos tarancdkroinidg saypsptelimca tisio dnesv. eFloirpset da ptop lpicroavtiiodne tehseta rbelaisl hteims ec oinmfomrumnaitciaotnio onf bbeutswese eton
central server and bus system which is capable of providing real-time data (GPS coordinates) regarding the
current location of buses. Second application is student side application which displays the current location of the
bus on Google map along with the estimated time of arrival to the specified bus stop hence it saves the time of
students. Both the side applications provide dynamically updated timetable. Another important feature of both the
applications is emergency service through which call or alert messages can be simultaneously sent to college,
police and ambulance in case of accidents. Thus this is a cost effective, efficient system and helps students to
catch their buses at the proper time.
Online Bus Arrival Time Prediction Using Hybrid Neural Network and Kalman fil...IJMER
International Journal of Modern Engineering Research (IJMER) is Peer reviewed, online Journal. It serves as an international archival forum of scholarly research related to engineering and science education.
International Journal of Modern Engineering Research (IJMER) covers all the fields of engineering and science: Electrical Engineering, Mechanical Engineering, Civil Engineering, Chemical Engineering, Computer Engineering, Agricultural Engineering, Aerospace Engineering, Thermodynamics, Structural Engineering, Control Engineering, Robotics, Mechatronics, Fluid Mechanics, Nanotechnology, Simulators, Web-based Learning, Remote Laboratories, Engineering Design Methods, Education Research, Students' Satisfaction and Motivation, Global Projects, and Assessment…. And many more.
Predicting Operating Train Delays into New York City using Random Forest Regr...AI Publications
The Long Island Railroad operates one of the largest commuter rail networks in the U.S.[1]. This study uses data which includes the location and arrival time of trains based on onboard GPS position and other internal sources. This paper analyzes the GPS position of the train to gain insight into potential gaps in on time performance and train operations. This was done by developing a Random Forest Re-gression model [2] and an XGBoost regression model [3[. Both models prove to be useful to make such predictions and should be used to help railroads to prepare and adjust their operations.
passSepnegceirasl. pIut rcpoonsesi sotns lionfe tbwuos tarancdkroinidg saypsptelimca tisio dnesv. eFloirpset da ptop lpicroavtiiodne tehseta rbelaisl hteims ec oinmfomrumnaitciaotnio onf bbeutswese eton
central server and bus system which is capable of providing real-time data (GPS coordinates) regarding the
current location of buses. Second application is student side application which displays the current location of the
bus on Google map along with the estimated time of arrival to the specified bus stop hence it saves the time of
students. Both the side applications provide dynamically updated timetable. Another important feature of both the
applications is emergency service through which call or alert messages can be simultaneously sent to college,
police and ambulance in case of accidents. Thus this is a cost effective, efficient system and helps students to
catch their buses at the proper time.
This paper proposes development of an android app to improve the transportation services for bus rental companies that lift Taibah University students. It intends to reduce the waiting time for bus students, thereby to stimulate sharing of updated information between the bus drivers and students. The application can run only on android devices. It would inform the students about the exact time of arrival and departure of buses on route. This proposed app would specifically be used by students and drivers of Taibah University. Any change in the scheduled movement of the buses would be updated in the software. Regular alerts would be sent in case of delays or cancelation of buses. Bus locations and routes are shown on dynamic maps using Google maps. The application is designed and tested where the users assured that the application gives the real time service and it is very helpful for them.
UNIVERSITY BUSES ROUTING AND TRACKING SYSTEMijcseit
This paper proposes development of an android app to improve the transportation services for bus rental
companies that lift Taibah University students. It intends to reduce the waiting time for bus students,
thereby to stimulate sharing of updated information between the bus drivers and students. The application
can run only on android devices. It would inform the students about the exact time of arrival and departure
of buses on route. This proposed app would specifically be used by students and drivers of Taibah
University. Any change in the scheduled movement of the buses would be updated in the software. Regular
alerts would be sent in case of delays or cancelation of buses. Bus locations and routes are shown on
dynamic maps using Google maps. The application is designed and tested where the users assured that the
application gives the real time service and it is very helpful for them.
This paper proposes development of an android app to improve the transportation services for bus rental
companies that lift Taibah University students. It intends to reduce the waiting time for bus students,
thereby to stimulate sharing of updated information between the bus drivers and students. The application
can run only on android devices. It would inform the students about the exact time of arrival and departure
of buses on route. This proposed app would specifically be used by students and drivers of Taibah
University. Any change in the scheduled movement of the buses would be updated in the software. Regular
alerts would be sent in case of delays or cancelation of buses. Bus locations and routes are shown on
dynamic maps using Google maps. The application is designed and tested where the users assured that the
application gives the real time service and it is very helpful for them.
Vehicle Headway Distribution Models on Two-Lane Two-Way Undivided RoadsAM Publications
The time headway between vehicles is an important flow characteristic that affects the safety, level of service, driver behavior, and capacity of a transportation system. The present study attempted to identify suitable probability distribution models for vehicle headways on 2-lane 2-way undivided (2/2 UD) road sections. Data was collected from three locations in the city of Semarang: Abdulrahman Saleh St. (Loc. 1), Taman Siswa St. (Loc. 2) and Lampersari St. (Loc.3). The vehicle headways were grouped into one-second interval. Three mathematical distributions were proposed: random (negative-exponential), normal, and composite, with vehicle headway as variable. The Kolmogorov-Smirnov test was used for testing the goodness of fit. Traffic flows at the selected locations were considered low, with traffic volume ranged between 400 to 670 vehicles per hour per lane. The traffic volume on Loc.1 was 484 vehicles per hour, that on Loc. 2 was 405 vehicles per hour, and that on Loc. 3 was 666 vehicles per hour. Random distribution showed good fit at all locations under study with 95% confidence level. Normal distribution showed good fit at Loc. 1 and Loc. 2, whereas composite distribution fit only at Loc. 1. It was suggested that random distribution is to be used as an input in generating traffic in traffic analysis at highway sections where traffic volume are under 500 vehicles per hour.
FORESEEING BUS ARRIVAL TIME IN VIEW OF TRAFFIC MODELING AND REAL-TIME DELAYijiert bestjournal
To know the entry time of bus to a bus stop is the fundamental data each travel clients need to know. Sitting tight for quite a while at a bus stop debil itates the individuals to depend upon open transpor ts. In this paper,we display a framework which can forese e the landing time of bus to a specific bus stop considering the ongoing parameters that influences the travel time of transport. With transport module,the ongoing parameters that influence travel time a re persistently gathered and used to foresee the bu s arrival time at different transport stops. A portab le application is created to help the questioning c lients in getting the arrival time of bus to a specific st op. As there will be postpone in travel time becaus e of the vehicles on street,street movement is displaye d utilizing M/G/1 lining hypothesis to ascertain th e delay created by other vehicles going in the same s treet. Server predicts the arrival time on the real time basis by bus module and the data shown in the datab ase of server. Server is coded such that it sends t he anticipated arrival time of transport to the questi oned users wireless at each bus stop it goes till i t achieves the stop questioned by client. Arrival tim e got by clients helps the travel clients to arrang e their schedule and reach the bus stop in time. Such an an ticipating framework propels the non clients of ope n transport frameworks to utilize them and avoid the utilization of private vehicles in their normal lif e.
JAVA 2013 IEEE DATAMINING PROJECT T drive enhancing driving directions with t...IEEEGLOBALSOFTTECHNOLOGIES
To Get any Project for CSE, IT ECE, EEE Contact Me @ 09849539085, 09966235788 or mail us - ieeefinalsemprojects@gmail.com-Visit Our Website: www.finalyearprojects.org
T drive enhancing driving directions with taxi drivers’ intelligenceIEEEFINALYEARPROJECTS
To Get any Project for CSE, IT ECE, EEE Contact Me @ 09849539085, 09966235788 or mail us - ieeefinalsemprojects@gmail.co¬m-Visit Our Website: www.finalyearprojects.org
Sensor based System for Computation of Speed, Momentum of Vehicle, Traffic De...iosrjce
This sensor based system can compute the speed of vehicle. It also computes the momentum of
vehicles and the traffic density on the approach road. This system will also detect any vehicular problems such
as brake failure, accidents on the roads and also detect the section which is affected or will be affected. The
system can be easily rectified in case there is a failure in the sensors. The system can also be used for estimating
any obstruction in between two ends of the road.
Congestion Effects of Parking and Pedestrians on Urban Road -A Case Study of ...ijsrd.com
Traffic congestion is a common phenomenon almost in all the cities of India. The situation has become to a very critical stage and has already assumed unbearable proportion in the cities of India. Cities of India cannot afford the economic and environmental loss resulted from this sever traffic Congestion. In this report, SBI to Bus-Station road in Modasa Town has been considered as study area. In the study area motorized and non-motorized vehicle creating congestion which has many negative effects. Traffic volume survey, Travel time delay study and informal occupancy survey has been conducted. A large number of vehicles occupy the road, reducing road capacity and creating congestion. In this study Average travel time delay measured due to Compared Speed-Density relationship with or Without Parking and Pedestrians and quantifies Congestion effect of Parking and Pedestrians.
Intelligent road surface quality evaluation using rough mereologyMohamed Mostafa
The road surface condition information is very useful for the safety of road users and to inform road administrators
for conducting appropriate maintenance. Roughness features of road surface; such as speed bumps and potholes, have bad effects on road users and their vehicles. Usually speed bumps are used to slow motor-vehicle traffic in specific areas in order to increase
safety conditions. On the other hand driving over speed bumps at high speeds could cause accidents or be the reason for spinal injury. Therefore informing road users of the position of speed bumps through their journey on the road especially at night or when lighting is poor would be a valuable feature. This paper exploits a mobile sensor computing framework to monitor and assess road surface conditions. The framework measures the changes in the gravity orientation through a gyroscope and the shifts in the accelerator's indications, both as an assessment
for the existence of speed bumps. The proposed classification approach used the theory of rough mereology to rank the modified data in order to make a useful recommendation to road users.
Cooperative Traffic Control based on the Artificial Bee Colony IJERA Editor
This paper studies the traffic control problem in an isolated intersection without traffic lights and phase, because the right-of-way is distributed to each vehicle individually based on connection of the Vehicle-to-Infrastructure (V2I), and the compatible streams are dynamically combined according to the arrival vehicles in each traffic flows. The control objective in the proposed algorithm is to minimize the time delay, which is defined as the difference between the travel time in real state and that in free flow state. In order to realize this target, a cooperative control structure with a two-way communications is proposed. First of all, once the vehicle enters the communication zone, it sends its information to the intersection. Then the passing sequence is optimized in the intersection with the heuristic algorithm of the Artificial Bee Colony, based on the arrival interval of the vehicles. At last, each vehicle plans its speed profile to meet the received passing sequence by V2I. The simulation results show that each vehicle can finish the entire travel trip with a near free flow speed in the proposed method.
Hybrid travel time estimation model for public transit buses using limited da...IAESIJAI
A reliable transit service can motivate commuters to switch their traveling mode from private to public. Providing necessary information to passengers will reduce the uncertainties encountered during their travel and improve service reliability. This article addresses the challenge of predicting dynamic travel times in urban areas where real-time traffic flow information is unavailable. In this perspective, a hybrid travel time estimation model (HTTEM) is proposed to predict the dynamic travel time using the predicted travel times of the machine learning model and the preceding trip details. The proposed model is validated using the location data of public transit buses of, Tumakuru, India. From the numerical results through error metrics, it is found that HTTEM improves the prediction accuracy, finally, it is concluded that the proposed model is suitable for estimating travel time in urban areas with heterogeneous traffic and limited traffic infrastructure.
This paper proposes development of an android app to improve the transportation services for bus rental companies that lift Taibah University students. It intends to reduce the waiting time for bus students, thereby to stimulate sharing of updated information between the bus drivers and students. The application can run only on android devices. It would inform the students about the exact time of arrival and departure of buses on route. This proposed app would specifically be used by students and drivers of Taibah University. Any change in the scheduled movement of the buses would be updated in the software. Regular alerts would be sent in case of delays or cancelation of buses. Bus locations and routes are shown on dynamic maps using Google maps. The application is designed and tested where the users assured that the application gives the real time service and it is very helpful for them.
UNIVERSITY BUSES ROUTING AND TRACKING SYSTEMijcseit
This paper proposes development of an android app to improve the transportation services for bus rental
companies that lift Taibah University students. It intends to reduce the waiting time for bus students,
thereby to stimulate sharing of updated information between the bus drivers and students. The application
can run only on android devices. It would inform the students about the exact time of arrival and departure
of buses on route. This proposed app would specifically be used by students and drivers of Taibah
University. Any change in the scheduled movement of the buses would be updated in the software. Regular
alerts would be sent in case of delays or cancelation of buses. Bus locations and routes are shown on
dynamic maps using Google maps. The application is designed and tested where the users assured that the
application gives the real time service and it is very helpful for them.
This paper proposes development of an android app to improve the transportation services for bus rental
companies that lift Taibah University students. It intends to reduce the waiting time for bus students,
thereby to stimulate sharing of updated information between the bus drivers and students. The application
can run only on android devices. It would inform the students about the exact time of arrival and departure
of buses on route. This proposed app would specifically be used by students and drivers of Taibah
University. Any change in the scheduled movement of the buses would be updated in the software. Regular
alerts would be sent in case of delays or cancelation of buses. Bus locations and routes are shown on
dynamic maps using Google maps. The application is designed and tested where the users assured that the
application gives the real time service and it is very helpful for them.
Vehicle Headway Distribution Models on Two-Lane Two-Way Undivided RoadsAM Publications
The time headway between vehicles is an important flow characteristic that affects the safety, level of service, driver behavior, and capacity of a transportation system. The present study attempted to identify suitable probability distribution models for vehicle headways on 2-lane 2-way undivided (2/2 UD) road sections. Data was collected from three locations in the city of Semarang: Abdulrahman Saleh St. (Loc. 1), Taman Siswa St. (Loc. 2) and Lampersari St. (Loc.3). The vehicle headways were grouped into one-second interval. Three mathematical distributions were proposed: random (negative-exponential), normal, and composite, with vehicle headway as variable. The Kolmogorov-Smirnov test was used for testing the goodness of fit. Traffic flows at the selected locations were considered low, with traffic volume ranged between 400 to 670 vehicles per hour per lane. The traffic volume on Loc.1 was 484 vehicles per hour, that on Loc. 2 was 405 vehicles per hour, and that on Loc. 3 was 666 vehicles per hour. Random distribution showed good fit at all locations under study with 95% confidence level. Normal distribution showed good fit at Loc. 1 and Loc. 2, whereas composite distribution fit only at Loc. 1. It was suggested that random distribution is to be used as an input in generating traffic in traffic analysis at highway sections where traffic volume are under 500 vehicles per hour.
FORESEEING BUS ARRIVAL TIME IN VIEW OF TRAFFIC MODELING AND REAL-TIME DELAYijiert bestjournal
To know the entry time of bus to a bus stop is the fundamental data each travel clients need to know. Sitting tight for quite a while at a bus stop debil itates the individuals to depend upon open transpor ts. In this paper,we display a framework which can forese e the landing time of bus to a specific bus stop considering the ongoing parameters that influences the travel time of transport. With transport module,the ongoing parameters that influence travel time a re persistently gathered and used to foresee the bu s arrival time at different transport stops. A portab le application is created to help the questioning c lients in getting the arrival time of bus to a specific st op. As there will be postpone in travel time becaus e of the vehicles on street,street movement is displaye d utilizing M/G/1 lining hypothesis to ascertain th e delay created by other vehicles going in the same s treet. Server predicts the arrival time on the real time basis by bus module and the data shown in the datab ase of server. Server is coded such that it sends t he anticipated arrival time of transport to the questi oned users wireless at each bus stop it goes till i t achieves the stop questioned by client. Arrival tim e got by clients helps the travel clients to arrang e their schedule and reach the bus stop in time. Such an an ticipating framework propels the non clients of ope n transport frameworks to utilize them and avoid the utilization of private vehicles in their normal lif e.
JAVA 2013 IEEE DATAMINING PROJECT T drive enhancing driving directions with t...IEEEGLOBALSOFTTECHNOLOGIES
To Get any Project for CSE, IT ECE, EEE Contact Me @ 09849539085, 09966235788 or mail us - ieeefinalsemprojects@gmail.com-Visit Our Website: www.finalyearprojects.org
T drive enhancing driving directions with taxi drivers’ intelligenceIEEEFINALYEARPROJECTS
To Get any Project for CSE, IT ECE, EEE Contact Me @ 09849539085, 09966235788 or mail us - ieeefinalsemprojects@gmail.co¬m-Visit Our Website: www.finalyearprojects.org
Sensor based System for Computation of Speed, Momentum of Vehicle, Traffic De...iosrjce
This sensor based system can compute the speed of vehicle. It also computes the momentum of
vehicles and the traffic density on the approach road. This system will also detect any vehicular problems such
as brake failure, accidents on the roads and also detect the section which is affected or will be affected. The
system can be easily rectified in case there is a failure in the sensors. The system can also be used for estimating
any obstruction in between two ends of the road.
Congestion Effects of Parking and Pedestrians on Urban Road -A Case Study of ...ijsrd.com
Traffic congestion is a common phenomenon almost in all the cities of India. The situation has become to a very critical stage and has already assumed unbearable proportion in the cities of India. Cities of India cannot afford the economic and environmental loss resulted from this sever traffic Congestion. In this report, SBI to Bus-Station road in Modasa Town has been considered as study area. In the study area motorized and non-motorized vehicle creating congestion which has many negative effects. Traffic volume survey, Travel time delay study and informal occupancy survey has been conducted. A large number of vehicles occupy the road, reducing road capacity and creating congestion. In this study Average travel time delay measured due to Compared Speed-Density relationship with or Without Parking and Pedestrians and quantifies Congestion effect of Parking and Pedestrians.
Intelligent road surface quality evaluation using rough mereologyMohamed Mostafa
The road surface condition information is very useful for the safety of road users and to inform road administrators
for conducting appropriate maintenance. Roughness features of road surface; such as speed bumps and potholes, have bad effects on road users and their vehicles. Usually speed bumps are used to slow motor-vehicle traffic in specific areas in order to increase
safety conditions. On the other hand driving over speed bumps at high speeds could cause accidents or be the reason for spinal injury. Therefore informing road users of the position of speed bumps through their journey on the road especially at night or when lighting is poor would be a valuable feature. This paper exploits a mobile sensor computing framework to monitor and assess road surface conditions. The framework measures the changes in the gravity orientation through a gyroscope and the shifts in the accelerator's indications, both as an assessment
for the existence of speed bumps. The proposed classification approach used the theory of rough mereology to rank the modified data in order to make a useful recommendation to road users.
Cooperative Traffic Control based on the Artificial Bee Colony IJERA Editor
This paper studies the traffic control problem in an isolated intersection without traffic lights and phase, because the right-of-way is distributed to each vehicle individually based on connection of the Vehicle-to-Infrastructure (V2I), and the compatible streams are dynamically combined according to the arrival vehicles in each traffic flows. The control objective in the proposed algorithm is to minimize the time delay, which is defined as the difference between the travel time in real state and that in free flow state. In order to realize this target, a cooperative control structure with a two-way communications is proposed. First of all, once the vehicle enters the communication zone, it sends its information to the intersection. Then the passing sequence is optimized in the intersection with the heuristic algorithm of the Artificial Bee Colony, based on the arrival interval of the vehicles. At last, each vehicle plans its speed profile to meet the received passing sequence by V2I. The simulation results show that each vehicle can finish the entire travel trip with a near free flow speed in the proposed method.
Hybrid travel time estimation model for public transit buses using limited da...IAESIJAI
A reliable transit service can motivate commuters to switch their traveling mode from private to public. Providing necessary information to passengers will reduce the uncertainties encountered during their travel and improve service reliability. This article addresses the challenge of predicting dynamic travel times in urban areas where real-time traffic flow information is unavailable. In this perspective, a hybrid travel time estimation model (HTTEM) is proposed to predict the dynamic travel time using the predicted travel times of the machine learning model and the preceding trip details. The proposed model is validated using the location data of public transit buses of, Tumakuru, India. From the numerical results through error metrics, it is found that HTTEM improves the prediction accuracy, finally, it is concluded that the proposed model is suitable for estimating travel time in urban areas with heterogeneous traffic and limited traffic infrastructure.
1. Dynamic forecasting of bus arrivals to stations
Hernández-Suárez, Carlos
Facultad de Ciencias, Universidad de Colima
Colima, Colima, 28140, MEXICO
E-mail: carlosmh@mac.com
and CIMAT- Unidad Monterrey
Centro de Investigación y Desarrollo en Ciencias de la Salud
Monterrey N.L. 66460, MEXICO
Introduction
Public transportation is strongly encouraged all over the world. In cities within developed
countries that have managed to have exclusive bus lanes, anticipating the arrival of a bus is
not a problem, since most of the time there is a schedule followed by the bus drivers, and
so, inter-arrival times are more or less uniform. This can be achieved for instance, when
buses belong to a public transportation system, that is, they belong to the governments and
the system is regulated and maintained by public authorities.
In some countries, buses belong to private owners, and the government acts only as a
regulator of fares and routes, but the administration of the system is left to the owners. This
autonomy results sometimes in a competition for passengers, resulting in irregular inter-arrival
times, due to variation in speed, or even to the fact that some owners decide to send their
buses to cover a different route at will. A more efficient service resulted from a self-
organization of bus owners, mainly due to the need of avoiding competition among them than
to please passengers. Still, the quality of service is limited because most buses have to
compete for space with regular traffic, and inter-arrival times are highly irregular.
Passengers would benefit from knowing how long they have wait for the next bus to
arrive for several reasons, including taking the decision of using a different transportation
alternative (i.e. taxis). This knowledge also helps to administer the waiting time. In any case,
it has been suggested that the knowledge of the arrival time tends to reduce stress.
The problem here is how to provide an accurate forecast even when in general, inter-
arrival times are highly variable, due to the aforementioned factors.
There is plenty of literature in the field of bus-arrival forecasts. (See 1-4 and references
therein) Instead of building models for bus displacement as a function of parameters as traffic,
route, speed, etc, we build our forecast system based on an analog technique.
The idea of seeking analogs is natural to many long-range weather forecast systems, but
transportation data is less limited than yearly weather data. The adapted key concept is this: a
bus will take the same time in traveling from point A to point B if all the actual conditions
repeat. Of course, since not all conditions can be repeated exactly and some conditions are
more important than other, we took as the first goal to collect all factors that may be
relevant to the forecast. Colima and Villa de Álvarez are two cities in west Mexico forming
a metropolitan area of pop. 255,000 where the system was tested. The city has 182 buses
covering 26 routes, but it has 110,500 vehicles, for one of the largest ratio of private
automobile per person, thus, bus circulation is continuously challenged.
2. Building efficient forecasts
The first task was to ask bus drivers what were the regular factors that they believe were
important in the variability of bus arrivals to bus stops. These were catalogued by route. The
most common factor detected from the interviews was time of the day followed by day of
the week. School day and Holidays were also mentioned. These factors influenced both the
amount of traffic and the amount of passengers, all of which affect the speed of a bus. The
second task was to classify the reported factors in categories, for instance: day of the week
was divided in 7 categories: (SU,MO,TU,WE,TH,FR,SA), time of the day was divided in 15
minutes categories starting at 00:00, weather was classified in no rain, mild rain and severe
rain (NR, MR, SR, School day (SY, SN) and Holiday (HY, HN) were also included.
There were historical records of AVL data, available at every two-minutes during the
last six months. In addition, the route was divided in 50-m segments and the coordinates
were translated to the actual segment the bus was in. The previous classification was added
to every record, a typical record line would be:
[40567 07/06/11 11:43:12 19.2877 -103.7380 47 280 21A TR 33 NR SY HN]
Which is read as:
[ID dd/mm/yy hh:mm/ss Latitude Longitude Quarter Segment Route Weekday Rain_status
School_day_status Holiday_status]
The position of a bus in a given day along a specific route is depicted as a position in a
cylinder where the basis of the cylinder is the position of the bus (segment) (Figure 1) and
its side is time (minutes). So, a bus “climbs up” the cylinder by turning around it giving a
“bus signature” (Figure 1)
Figure 1. The signature of eleven buses during a whole day. X-axis
is the segment the bus is in (1 = an arbitrary but fixed segment)
and the Y axis the time in minutes since the beginning of the day.
3. In advance forecasting
One of the most challenging problems in bus arrival forecasting, is dealing with the
massive amount of information required to make the forecast of every bus to every station, as
well as its nearly continuous update. Our approach was to anticipate the queries as follows: a
given route is divided in S segments and is transited by N buses. For every quarter of the
day and for every segment, we calculate using historical data, the average time it take for a
bus to arrive to every one of all the following segments for each of the possible
combinations of current Weekday, Rain status, School day status, and Holiday status. This is
computational time consuming but the table can be constructed days or weeks beforehand.
When a bus reports its position, the current condition of the bus is classified according to the
actual segment, the quarter, the weekday, the Rain status, School day status, and Holiday
status, and the forecast is taken from the table constructed beforehand, avoiding calculations.
With this we update a dynamic table, containing the forecast of each bus to each following
segment (see Table 1)
Table 1. The dynamic table
BUS NUMBER
33 35 77 18 221 324 42
1 38 22 61 23 5 40 17
2 39 27 63 30 9 46 24
SEGMENT NUMBER
3 42 36 7 34 17 49 33
4 48 38 8 44 25 3 43
5 58 43 17 45 27 10 49
6 68 53 27 8 32 17 51
7 9 61 34 10 37 19 53
8 19 71 42 12 44 21 56
9 21 2 50 13 52 26 8
10 31 12 54 14 60 36 11
The advantage of this strategy is that all calculations can be made beforehand: every
time a bus updates its position, this is classified according to the reported position and other
variables and the forecast is sought in a database with minor calculations. A person waiting
for the bus at segment number 7, for instance, will receive as a result to its query, the
numbers of the 7-th row, sorted. We can filter those arrivals exceeding some amount of time,
say 30 min. In practice, only those segments that contain a bus station are included in the
table.
Results
At the end, there were no difference in the forecasts between weekdays MO-FR, and
there were only three weekday categories: MO-FR, SA and SU. So far, we have an overall
4. 10% error in the forecast with a 95% confidence level, meaning that when the forecast is 30
minutes, the arrival will occur within (27 min -33 min) with a 95% probability, and for a
forecast of 10 minutes, within (9 min -11 min). Nevertheless, we lack of data for the rainy
season, where the system has not been tested yet.
REFERENCES
[1] Mei Chen, Xiaobu Liu, Jingxin Xia, Steven I.Chien, “A Dynamic Bus-Arrival Time Prediction
Model Based on APC Data”, Journal of Computer-Aided Civil and Infrastructure Engineering, United
Kingdom, Volume 19, 2004, pg 364-376
[2] Wei-Hua Lin, Jian Zeng, “An Experimental Study On Real Time Bus Arrival Time Prediction With
GPS Data”, Journal of Center for Transportation Research and Department of Civil and Environmental
Engineering, Virginia
[3] Jens Biesterfeld, Elyes Ennigrou, Klaus Jobmann, “Neural Networks for Location Prediction in
Mobile Networks”, Journal of Institute far Allgemeine Nachrichtentechnik, University Hannover,
Germany
[4] Lajos KISGYORGY, Laurence R.RILETT, “Travel Time Prediction By Advanced Neural Network”,
Journal of Budapest University of Technology and Economics, Texas, 10 October 2001