16
Smart Arrival Notification System for ADA Passenger Paratransit Service Using 1 Consumer Mobile Device 2 Slobodan Gutesa 1 3 Ph.D Candidate 4 Phone: (973) 596-8497, Fax: (973) 596-6454 5 Email: [email protected] 6 7 Branislav Dimitrijevic 1 8 Senior Research Scientist 9 Phone: (973) 596-6463, Fax: (973) 596-6454 10 Email: [email protected] 11 12 Joyoung Lee 1 , Ph.D. (Corresponding Author) 13 Assistant Professor 14 Phone: (973) 596-2475, Fax: (973) 596-6454 15 Email: [email protected] 16 17 Yuchuan Zhang 2 18 Graduate Research Assistant 19 Phone: (973) 596-5211, Fax: (973) 596-6454 20 Email: [email protected] 21 22 Cecilia Feeley 3 , Ph.D 23 Transportation Autism Project Manager 24 Phone: 848-445-2975, Fax: 732-445-3325 25 Email: [email protected] 26 27 Lazar Spasovic 1 , Ph.D 28 Professor 29 Phone: (973) 642-7214, Fax: (973) 596-6454 30 Email: [email protected] 31 32 33 Paper submitted for consideration for publication in the Journal of Transportation Research 34 Board and presentation for the TRB Annual Meeting in January 2017 35 36 7300 words = 4800 + 2500 (=250 X 6 Figures + 250 X 4 Tables) 37 *Paper revised from original submittal 38 1 Department of Civil and Environmental Engineering, New Jersey Institute of Technology, University Height, Newark, NJ 07102 2 Department of Computer Science, New Jersey Institute of Technology, University Height, Newark, NJ 07102 3 Center for Advanced Infrastructure and Transportation, Rutgers, The State University of New Jersey, 100 Brett Road Piscataway, NJ 08854-8058

Smart Arrival Notification System for ADA Passenger ...docs.trb.org/prp/17-04593.pdf1 Smart Arrival Notification System for ADA Passenger Paratransit Service Using 2 Consumer Mobile

  • Upload
    lamnhu

  • View
    223

  • Download
    0

Embed Size (px)

Citation preview

Smart Arrival Notification System for ADA Passenger Paratransit Service Using 1 Consumer Mobile Device 2

Slobodan Gutesa1 3 Ph.D Candidate 4

Phone: (973) 596-8497, Fax: (973) 596-6454 5 Email: [email protected] 6

7 Branislav Dimitrijevic1 8

Senior Research Scientist 9 Phone: (973) 596-6463, Fax: (973) 596-6454 10

Email: [email protected] 11 12

Joyoung Lee1, Ph.D. (Corresponding Author) 13 Assistant Professor 14

Phone: (973) 596-2475, Fax: (973) 596-6454 15 Email: [email protected] 16

17 Yuchuan Zhang2 18

Graduate Research Assistant 19 Phone: (973) 596-5211, Fax: (973) 596-6454 20

Email: [email protected] 21 22

Cecilia Feeley3, Ph.D 23 Transportation Autism Project Manager 24

Phone: 848-445-2975, Fax: 732-445-3325 25 Email: [email protected] 26

27 Lazar Spasovic1, Ph.D 28

Professor 29 Phone: (973) 642-7214, Fax: (973) 596-6454 30

Email: [email protected] 31 32

33 Paper submitted for consideration for publication in the Journal of Transportation Research 34 Board and presentation for the TRB Annual Meeting in January 2017 35 36 7300 words = 4800 + 2500 (=250 X 6 Figures + 250 X 4 Tables) 37 *Paper revised from original submittal 38

1 Department of Civil and Environmental Engineering, New Jersey Institute of Technology, University Height, Newark, NJ 07102 2 Department of Computer Science, New Jersey Institute of Technology, University Height, Newark, NJ 07102 3 Center for Advanced Infrastructure and Transportation, Rutgers, The State University of New Jersey, 100 Brett Road Piscataway, NJ 08854-8058

1

ABSTRACT 1

This research presents an arrival notification system for paratransit passengers with disabilities. 2 Almost all existing curb-to-curb paratransit services have significantly large pick-up time 3 window ranging from 20 to 40 minutes from the scheduled time producing substantial passenger 4 waiting times. The arrival notification system presented in this study delivers an automated voice 5 call to a registered user once the paratransit vehicle is in the near proximity to the pick-up 6 location. The system utilizes Google Traffic API for the vehicle arrival estimation. Unlike other 7 vehicle arrival notification systems in the state-of-the-practice, the proposed system is compact 8 and does not require additional equipment such as radio transmitting and positioning devices. 9 Using consumer mobile devices with Android or iOS platform, the proposed system is designed 10 to exploit commercial cellular network service (i.e., 3G and 4G-LTE). In addition to the 11 passenger notification, the proposed system provides paratransit drivers with real-time route 12 guidance information developed through Google Maps API. Field evaluation conducted in Essex 13 County, New Jersey, reveals significant reduction in passenger waiting time. The passenger 14 waiting time was reduced by 15 to 20 minutes. In addition, accuracy of the notification system 15 was tested: during the test, in almost all cases, vehicle arrived 1 minute earlier from the proposed 16 arrival time. 17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

2

INTRODUCTION 1

Background 2

Unlike fixed-route transit, paratransit is a public transportation service that offers flexible routes 3 to accommodate specific categories of users such as elderly or disabled population. The main 4 characteristic of paratransit is that it usually provides door-to-door service for users who reserve 5 a ride. Basic paratransit offers service between user-specified origin and destination that is within 6 the paratransit operating area (1). Although there are some paratransit services open to the 7 general public, most of the service have eligibility requirements, and can be restricted to 8 individuals with disabilities or senior citizens. ADA paratransit service is accessible in the same 9 service area and during the same hours as regular, fixed-route service (1). The Americans with 10 Disabilities Act (ADA) of 1990 (2) that protects the rights of people with disabilities requires 11 public transportation system to offer ADA paratransit service to individuals who are unable to 12 use local bus service as a result of their disability. According to ADA, not everyone with 13 disability is eligible to use ADA paratransit services. Individuals with disabilities who can access 14 regular fixed-route public services do not qualify for this paratransit category. Specifically, a 15 person qualifies for paratransit if it falls into one of three categories: 1) persons who cannot get 16 on or off the regular public transit vehicle due to their disability (e.g., person with cognitive 17 disabilities or a person with visual impairment; 2) persons who are capable of using regular 18 public transit but accessible vehicles are not available or are down to maintenance; and 3) 19 persons with disability that prevents them from approaching public transit stops (e.g., 20 impairment conditions that prevent them from overcoming barriers such as curbs or steps). In 21 most cases paratransit providers use small, lift-equipped vans or minibuses with flexible route 22 deviations that are suitable for curb-to-curb service in residential areas. Although most 23 paratransit agencies have the wide variety of vehicles ranging from small sedans or vans to 24 minibuses, proper organization and fleet utilization represents a complex organizational 25 challenge (3). Despite the drastic advancement of demand forecasting techniques for the past two 26 decades, precisely predicting the demand of disabled people for paratransit service represents a 27 complex problem, especially for moderate to large sized American cities where demand variation 28 can be large.(4). Such organizational practices, as well as adverse traffic conditions can result in 29 unpredictable waiting times on the customer side. Aforementioned facts are forcing paratransit 30 agencies to develop policies that often require customers to accept long waiting times when 31 scheduling a ride. 32

Existing Paratransit Practice 33

Paratransit demand increased as a result of ADA-supported expansion of paratransit services. 34 Various “dial a ride” agencies are operating around the country with similar or even identical 35 scheduling and dispatching practices. Access-a-Ride Service operated by Metropolitan 36 Transportation Agency (MTA) in New York City is a curb-to-curb service that requires 37 customers to schedule their ride in advance (5). According to the policies developed by MTA a 38

3

vehicle can arrive at the pick-up location any time within the 30 minutes pick-up window (5). 1 Similarly, San Francisco Paratransit rules and policies defined a 20 minutes “on-time” window 2 where vehicle is allowed to arrive 5 minutes prior or 15 minutes after the scheduled time (6). 3 Chittenden County Transportation Authority (CCTA) paratransit for individuals with disabilities 4 in Vermont and Specialized Community Area Transportation (SCAT) of North Carolina have a 5 30 minutes pick-up window that requires customers to be available 10 minutes before or 20 6 minutes after the scheduled time (7-8). Chicago Paratransit service known as Pace, defined a 20 7 minutes pick-up window (9). NJ Transit’s Access Link has a 40 minutes time window policy, 8 where vehicle can arrive 20 minutes before or 20 minutes after the scheduled time (10). Different 9 agencies have different penalties for customers who do not show up at the pick-up locations and 10 they vary from written warnings to suspension of services (5). Almost all the agencies commit to 11 wait for maximum 5 minutes after vehicle arrival. Those policies are the main reason for 12 customer complaints or even service cancelation. In situations where a customer is required to 13 come to the curb to meet the vehicle within five minutes, customers tend to wait for the vehicle 14 outdoors close to the pick-up location. The long waiting time is often difficult during inclement 15 weather, especially for persons with different cognitive or psychological limitations or persons 16 with significant health impairments. 17

The main purpose of this study is to develop a smart arrival notification system for ADA 18 passenger. The proposed system will send a text message or initiate an automated telephone call 19 when paratransit vehicle is close to the pick-up location. Thus the proposed notification system 20 will enable ADA passenger to spend less time outdoors and leave their house or waiting area 21 when a transit vehicle is in the near proximity. In addition the proposed system also allows ADA 22 passengers to customize an advance notification setup that will accommodate their personal 23 needs, which significantly reduce the passenger waiting time. 24

The main structure of this paper is the following. In the next section, both state-of-the-art 25 and practice efforts for transit arrival notification technologies will be reviewed. The overall 26 system architecture of proposed smart arrival notification system developed by the research team 27 will be presented along with details for core components in the next section. The section of field 28 evaluations will discuss experimental design and data collection approach to evaluate the 29 performance of the proposed system, followed by the evaluation results. Findings and future 30 research will be addressed in the section of concluding remarks. 31 32

LITERATURE REVIEW 33

Beginning with state-of-the-art efforts for bus arrival time estimation, this section presents 34 literature review on relevant technologies for the proposed arrival notification system. 35

Until now, several authors developed mathematical algorithms to predict vehicle arrival 36 time. Lin et al (11) developed a mathematical method to gain real-time vehicle arrival 37 information by utilizing regression models. In this study, a bus location, scheduled arrival time, 38 difference between predicted and actual arrival time, and waiting time at time-check stops are 39

4

used to predict the arrival time. Although providing acceptable accuracy, the model can only be 1 applied in the rural areas with low traffic intensity. Dailey et al (12) used time series model to 2 predict vehicle arrival time up to 1-hour in advance. The model uses data from an automated 3 vehicle location (AVL) system comprising vehicle location and time. Shalaby et al (13) used 4 Kalman Filter (14) with AVL data for the prediction of vehicle arrival time. The model is used 5 for user-interactive system in Toronto, Canada which provides continuous information on the 6 expected arrival and departure times of buses at downstream stops. Similarly, using GPS-based 7 AVL system, Schimer and Freda (15) developed a real-time notification system for passengers 8 waiting at bus stops along the route. To estimate the arrival time of each bus, the algorithm 9 employed historical inter-stop travel time data obtained from the AVL system. The solution was 10 later introduced to public through Next Bus web service. Chien and Kuchipudi (16) used Kalman 11 Filter with historic path-based data to predict vehicle travel time. Prediction models were also 12 studied by Yu et al (17) who compared several models for the arrival time prediction using the 13 data collected from multiple bus routes. Among Support Vector Machine (SVM) (18), artificial 14 neural network (ANN)(19), k nearest neighbors (k-NN) (20) and linear regression, the SVM 15 model provided the most accurate estimates to predict vehicle arrival time. Jeong and Rilett (21) 16 provided another performance comparison for several methods. The historical data based model, 17 the regression models, and the ANN model were compared. The results showed that the ANN 18 model outperformed the historical data based model and the regression model in terms of 19 accuracy. In the state-of-the practice, several solutions for arrival notification that utilize Radio 20 Frequency (RF) transmitting equipment have been proposed. Initially, Bishop et al (22) 21 developed a school bus approach notification system using radio transmitted signal which 22 activates a visual or audio alarm system instrumented inside of the passenger’s place of 23 residence. This solution requires the installation of a radio transmitter on each school bus 24 vehicle. Winkler et al. (23) presented a system for notifying passengers consisted of GPS module 25 and RF transmitter with antenna, mounted on a vehicle. The system notifies passengers at a bus 26 stop over displays and speakers. Jones (24) developed an advance notification system that 27 notifies users of the impending arrival of a vehicle. The patent is a GPS-based solution designed 28 to be implemented to a delivery truck system that communicates with the base station using RF 29 modem. This notification system generally comprises a vehicle control unit, a base station 30 control unit, and a user computer where the arrival information is displayed. Recently, Dow et al 31 (25) proposed a location based paratransit notification system where visual and audio notification 32 services can be provided to a passenger based on the vehicle location. The proposed 33 methodology requires dedicated short-range communications (DSRC) device and internet access. 34 The arrival notification is delivered to users though mobile application. 35

In summary, the research concepts and market-ready solutions reviewed in this paper rely 36 heavily on historical travel time data to predict the arrival time. This often becomes inappropriate 37 as non-recurrent roadway congestions can produce unpredicted vehicle delay and inaccurate 38 vehicle arrival time estimations. Furthermore, the existing solutions require bulky on-board 39 components to be mounted in a vehicle, such as GPS unit, radio antenna, transmitting and 40

5

receiving modules, and power supply equipment. It is worth noting that such components are 1 readily available in the all-in-one board of a consumer mobile device (e.g., smartphone, Tablet 2 PC). 3

SMART ARRIVAL NOTIFICATION SYSTEM 4

NJIT research team developed a personalized notification system for ADA paratransit passenger 5 that provides highly accurate arrival time estimation. The bus arrival time is calculated using real 6 time traffic information through Google API (26). One of the main advantages of this solution is 7 that it provides a personalized notification through a telephone call or SMS to each ADA 8 passenger individually. Each passenger has option to setup the advance notification time to meet 9 their own needs. The predicted arrival time is calculated using current vehicle location and the 10 pick-up location of an ADA passenger along with an estimated time of arrival (ETA) obtained 11 from the Google Traffic and Google Distance Matrix API (27). Google Traffic obtains prevailing 12 traffic speed and travel time data through their crowd-sourcing environment, where data is 13 obtained by analyzing the GPS-determined locations transmitted to Google by a large number of 14 mobile phone users. Google processes the incoming raw data about mobile phone device 15 locations, and then excludes anomalies such as traffic signals, postal vehicles or regular bus 16 service which make frequent stops. This makes Google traffic model one of the most accurate 17 models currently in the market (28). 18

System Architecture 19 20 Figure 1 depicts high-level overall system architecture for the arrival notification system that the 21 NJIT research team developed. Integrating with commercial tools provided by Google and 22 Twilio (29), the primary components of the system are consisted of 1) Mobile Application and 2) 23 Service Manager. Twilio is a private company which provides automated programmable phone 24 call and text message services. The Twilio’s programmable voice and message services have 25 been adopted by numerous private companies providing similar services, such as Uber, Home 26 Depot, Box, etc. 27

Running on a mobile device (e.g., smartphone and tablet PC), the mobile application 28 handles vehicle positioning, route guidance, and arrival time estimation. It also triggers 29 automatic phone call and text message services to ADA passengers based on the estimated 30 arrival time and passenger pick-up schedules obtained from the service manager. The service 31 manager deals with ADA customer information, such as address, pick-up time, phone numbers, 32 and passenger to pick-up location assignment. Details for both components are discussed in the 33 next sections. 34

6

1 Figure 1. Passenger advanced notification- system architecture 2

3

Mobile Application 4

The NJIT research team developed the mobile application on both Android and iOS platforms. 5 Figure 2 presents the actual deployment of the mobile application running on an Android Tablet 6 for a case study that will be discussed in the next section. The application obtains instantaneous 7 vehicle position information through a GPS receiver embedded in the mobile device to update 8 the origin of the trip to the next pick-up location. The customer pick-up location is obtained from 9 the service manager. Once the next destination is identified, the estimated arrival time is 10 calculated through Google Traffic and Google Distance Matrix API. Combining the customer 11 data from the service manager and real-time information from the mobile application (e.g., 12 instantaneous position, estimated travel time, and route to the destination), the mobile application 13 triggers a phone call request message to the service manager once the vehicle arrives to the 14 desired proximity to the passenger. 15

Reflecting prevailing traffic congestion condition, the mobile application also provides 16 the driver of paratransit vehicle with real-time route guidance information as shown on Figure 3. 17 Customized for our notification system, the route guidance is handled by standard Google Maps 18 navigation panel (30) by displaying a recommended route on the display of the mobile device. 19 The customer pick-up location and the current vehicle position are also visible on the mobile 20 application display for driver’s convenience as shown on Figure 3. Once a notification phone-21

7

call to the next ADA passenger has been initiated, the driver is notified over the display with a 1 simple message, “Calling”. 2 3

4

Figure 2 Deployment of the Mobile Application 5

6

Figure 3 Navigation Tool for the Mobile Application 7

8

8

Service Manager 1 The service manager is computer system embedded with a SQL database and software developed 2 by the research team to conduct automatic service management. The service manager builds a 3 roster for each transit driver on a daily-basis to provide passenger information such as pick-up 4 location, pick-up time, and contact number for text message and/or phone call. Table 1 show the 5 hypothetical ADA customer information, which is typically collected from the user while 6 scheduling a ride. The data was manually fed into the service manager database to be retrieved 7 by the mobile application during the test. Once the mobile application obtains vehicle position 8 from the embedded GPS receiver, it sends a query to the service manager to obtain the next 9 passenger information such as pick-up location (e.g., either address or geocode), pick-up time, 10 and notification preference. It is noted that the next pick-up locations are further used for the 11 mobile application to calculate the estimated arrival time to the next passenger. Once the 12 estimated arrival time is equal to the desired advance notification time as shown in the last field 13 of Table 1, the mobile application sends a triggering message to the service manager. With the 14 triggering message, the service manager places a request for Twilio to make a phone call to the 15 ADA customer by using a programmable voice message or SMS text: e.g., Dear Mr. “Name” 16 your scheduled bus service will arrive in “ETA” minutes. Employing Twilio Cloud 17 Communication API (29), an automated voice call is initiated by sending a secured Hypertext 18 Transfer Protocol (HTTP) request which is placed by the service manager based upon the 19 triggering message from the mobile application. 20 21

Table 1 ADA Customer Information for the service manager 22

ID Name Geocode Address Phone # Pick-up Time Date Notification

Preference

1 John Smith

45.7859, -74.33456

123 Davidson Rd

896-989-XXXX 13:10:00 5/10/2016 10 min

2 Rodger Mccline

45.7859, -74.33456 1467 Grove St 332-669-

XXXX 13:35:00 5/10/2016 5 min

3 Chris Timms

45.7859, -74.33456

1445 Bloomfield Av

369-998- XXXX 14:05:00 5/10/2016 15 min

4 Orlando Rodriquez

45.7859, -74.33456

1238 Central Av

336-987- XXXX 14:20:00 5/10/2016 20 min

5 Monica Isaacs

45.7859, -74.33456

168 Harrison Rd

698-963- XXXX 14:45:00 5/10/2016 15 min

23 Figure 4 conceptually illustrates the interaction between the mobile application and the 24

service manager of the smart arrival notification system that the NJIT research team developed. 25

9

1 Figure 4. Interaction between the mobile application and service manager of the notification system 2

3

FIELD EVALUATIONS 4

Evaluation Design 5 The performance of the proposed arrival notification system was examined through the field test 6 by taking into consideration different prevailing traffic conditions. To measure the performance, 7 passenger waiting time at each pick-up location was collected. The test site, designated in Essex 8 County, New Jersey, covered several locations corresponding to typical ADA passenger 9 distribution for a paratransit agency. The test site comprises a highly urbanized area in the 10 downtown of Newark and multiple suburbanized areas with slightly lower traffic congestion 11 conditions as shown in Figure 5. The test started at University Hospital in the Newark downtown 12 area and went through several townships using urban highway (i.e. Garden State Parkway), 13 signalized arterials (i.e., US-22), and local streets in the county. 14 15

10

1 Figure 5 Notification system test site 2

3 A total of four hypothetical ADA passengers were assumed to be coming to the 4

designated pick-up address as summarized in Table 2. For the base case where no notification is 5 used, the scheduled pick-up time for each ADA passenger was determined based on field traffic 6 congestion conditions experienced by the research team. For the case of the notification system, 7 the preferred advance notification time was set to 10 minutes for every ADA passenger. That is, 8 a notification call is placed when the transit vehicle is within the 10-minute proximity of the 9 pick-up location. In addition, it was assumed that the ADA passengers arrive 5 minutes after the 10 notification call was received. On the other hand, the ADA passengers without the notification 11 system were assumed to arrive at the pick-up location 20 minutes prior to the scheduled pick-up 12 time as it is required by most paratransit agencies following the ADA protocol. 13

14 Table 2 Scheduled passenger pick-up times and locations 15

Passenger

Scheduled Passenger Pick-up Time for the

Base Case

Preferred Advance Notification Time

(Min)

Pick-up Address

1 11:49 AM

10

120-125 Bergen Street, Newark 2 12:05 PM 443-450 South Orange Avenue, South Orange 3 12:35 PM 422-430 Prospect Street, Westfield 4 12:55 PM 1744-1747 Whittier Street, Rahway, NJ

16 17

11

Results 1 Tables 3 and 4 summarize the passengers’ waiting times measured from the cases 1) without and 2 2) with the notification system, respectively. For the base case without the notification system, it 3 was observed that the passenger waiting times are ranging from 16 to 23 minutes as shown in 4 Table 3 and Figure 6. Furthermore, the differences between the scheduled and actual pick-up 5 times appeared up to 3 minutes for the passengers 1 and 2. That is, although the passengers reach 6 the pick-up locations at the scheduled pick-up time, they would need to wait additional 3 minutes 7 to be picked up due to the uncertainty of prevailing traffic congestion condition. 8 9

Table 3 Field test results for the base case (without notification system) 10 Passenger

Scheduled Pick-up

Time

Actual Vehicle Arrival Time

Time Difference

(Scheduled-Actual)

Passenger Arrival Time at the Pick-up Location

Passenger Waiting

Time (Min)

1 11:50 AM 11:53 AM +3 min 11:30 AM 23 2 12:05 PM 12:08 AM + 3min 11:45 AM 23 3 12:35 PM 12:34 PM -1 min 12:15 PM 19 4 12:55 PM 12:51 PM -4 min 12:35 PM 16

Mean 20.3 11

On the other hand, the results shown in Table 4 and Figure 6 clearly demonstrate that the 12 notification system achieves substantial reduction in passenger waiting time. During the test, it 13 was observed that the test vehicle arrived 8 to 11 minutes after the notification call was placed. 14 Knowing that the preferred notification time for the call is given by 10 minutes, these arrivals 15 produced 1 to 2 minutes difference, which resulted in dramatic waiting time reductions ranging 16 from 3 to 6 minutes. Compared to the base case yielding average 20.3 minutes of passenger 17 waiting time, the notification system produced promising benefits reducing the average waiting 18 time by 16 minutes. In addition, from the maximum 1-minute of time difference between 19 estimated and actual pick-up time, the notification system helped improve the punctuality of 20 ADA paratransit service. By utilizing smart arrival notification all passengers can benefit 21 throughout waiting time reduction due to shared-ride nature of paratransit. Beside passenger 22 benefits, transportation provider benefits are also expected through increased attractiveness of 23 the service which can further produce increase in ridership. In addition, smart notification system 24 can produce significant savings for the paratransit fleet due to reduction in dwelling time of 25 vehicles. 26

Table 4 Advanced passenger notification field test results 27

Passenger

Call Placed Time

Estimated Pick-up

Time

Actual Pick-up

Time

Time Difference

(Estimated-Actual)

Passenger Arrival Time at the Pick-up Location

Passenger Waiting

Time (Min)

1 11:44 AM 11:55 PM 11:53 AM -2 min 11:49 AM 4 2 12:00 PM 12:10 PM 12:08 AM -2 min 12:05 PM 3 3 12:23 PM 12:33 PM 12:34 PM +1 min 12:28 PM 6 4 12:42 PM 12:52 PM 12:51 PM -1 min 12:47 PM 4

Mean 4.3

12

1

2 Figure 6 Passenger waiting time w/o and w/ notification system 3

CONCLUDING REMARKS 4

Conclusions 5

Existing paratransit services in American metropolitan areas are experiencing difficulties in 6 maintaining on-schedule arrival time determined by dispatchers and fleet managers. Adverse 7 traffic conditions along the route are often unpredictable, thereby resulting in high variation of 8 the vehicle arrival time. Many paratransit agencies are forced to develop strict policies that 9 require customers to be prepared for boarding within 5 minutes of the vehicle arrival. On the 10 other hand, those agencies cannot guarantee exact time of the vehicle arrival, resulting in wide 11 pick-up time window ranging from 20 to 40 minutes. In order to board vehicle on time, ADA 12 passengers usually arrive at the pick-up location much earlier. For passengers with different 13 health difficulties, such long waiting time might become unacceptable. 14

The main idea behind the development of the proposed smart arrival notification system 15 was to decrease unnecessary waiting time for ADA passengers. Discovered by the field 16 evaluations for the proposed notification system, the waiting time can be significantly reduced if 17 the passengers have accurate information for the vehicle arrival time. During the test, for 4 18 locations in Essex County, New Jersey, waiting time was reduced by 15 to 20 minutes depending 19 on location. Such reduction in waiting time can certainly improve the quality of paratransit 20 services for ADA passengers and decrease the potential dangerous conditions. The field test 21 revealed that the accuracy of an arrival time prediction is within 1 minute, which significantly 22 reduces passenger waiting time and increases passengers’ confidence that the paratransit vehicle 23 will arrive as notified. By using accurate notification system, ADA passengers can spend less 24 time outdoors and leave the place where they reside in a more convenient fashion. 25

26

13

Future Research 1

By taking into consideration the prevailing traffic congestion condition, this study primarily 2 focused on developing automatic notification system to reduce ADA passengers waiting time 3 based on presumed pick-up schedules. Thus, minimal effort to deal with optimal pick-up 4 schedule was conducted for the field evaluation. However, it must be noted that the proposed 5 notification system could be enhanced by employing advanced scheduling techniques in the 6 future. In addition, the notification system requires additional field tests in a variety of 7 metropolitan areas and with a diverse group of paratransit consumers. 8

REFERENCES 9

[1] Lave, Roy, and Rosemary Mathias "State of the Art of Paratransit. “Transportation in the 10 New Millennium 478 (2000). 11

[2] Simon, Rosalyn M. "Integrating Americans with disabilities act paratransit services and 12 health and human services transportation." TCRP Research Results Digest 10 (1997). 13

[3] Fu, Liping, and Gary Ishkhanov. "Fleet size and mix optimization for paratransit services.” 14 Transportation Research Record: Journal of the Transportation Research Board 1884 (2004): 39-15 46. 16

[4] Bearse, Peter, Shiferaw Gurmu, Carol Rapaport, and Steven Stern. "Paratransit demand of 17 disabled people." Transportation Research Part B: Methodological 38, no. 9 (2004): 809-831. 18

[5] MTA New York (2016, January 01). Access-a-Ride No-Show/Late Cancellation Policy. 19 Retrieved June 7, 2016, from 20 http://web.mta.info/nyct/paratran/policies/NoShowLateCanPolicy.htm 21

[6] San Francisco Paratransit. (2015, August 1). Paratransit Rules and Polices At A Glance. 22 Retrieved June 22, 2016, from http://www.sfparatransit.com/rules-and-policies.htm 23

[7] CCTA Transit Authority (2016, January 01). ADA Paratransit Guide. Retrieved June 22, 24 2016, from http://cctaride.org/wordpress/wp-content/uploads/adaparatransitguide1.pdf 25

[8] SCAT Paratransit Service North Carolina. (n.d.). SCAT PARATRANSIT SERVICES. 26 Retrieved June 23, 2016, from http://www.greensboro-nc.gov/index.aspx?page=2188 27

[9] "Pace ADA Paratransit Service." Pace ADA Paratransit Service. July 15, 2015. Accessed 28 June 26, 2016. http://pacebus.com/pdf/paratransit/City_ADA_Service_Guide_July2016.pdf. 29

[10] New Jersey Transit. (n.d.). Access Link ADA Paratransit Service. Retrieved June 22, 2016, 30 from http://www.njtransit.com/tm/tm_servlet.srv?hdnPageAction=AccessLinkTo 31

14

[11] Lin, Wei-Hua, and Jian Zeng. "Experimental study of real-time bus arrival time prediction 1 with GPS data." Transportation Research Record: Journal of the Transportation Research Board 2 1666 (1999): 101-109. 3

[12] Dailey, D., S. Maclean, F. Cathey, and Z. Wall. "Transit vehicle arrival prediction: 4 Algorithm and large-scale implementation." Transportation Research Record: Journal of the 5 Transportation Research Board 1771 (2001): 46-51. 6

[13] Shalaby, Amer, and Ali Farhan. "Prediction model of bus arrival and departure times using 7 AVL and APC data." Journal of Public Transportation 7, no. 1 (2004): 3. 8

[14] Grewal, Mohinder S. Kalman filtering. Springer Berlin Heidelberg, 2011. 9

[15]. Schmier, Kenneth J., and Paul Freda. "Public transit vehicle arrival information system." 10 U.S. Patent 6,006,159, issued December 21, 1999. 11

[16] Chien, Steven I-Jy, and Chandra Mouly Kuchipudi. "Dynamic travel time prediction with 12 real-time and historic data." Journal of transportation engineering 129, no. 6 (2003): 608-616. 13

[17] Yu, Bin, William HK Lam, and Mei Lam Tam. "Bus arrival time prediction at bus stop with 14 multiple routes." Transportation Research Part C: Emerging Technologies 19, no. 6 (2011): 15 1157-1170. 16

[18] Vanajakshi, Lelitha, and Laurence R. Rilett. "Support vector machine technique for the 17 short term prediction of travel time." In 2007 IEEE Intelligent Vehicles Symposium, pp. 600-18 605. IEEE, 2007. 19

[19] Smith, Brian L., and Michael J. Demetsky. "Short-term traffic flow prediction: neural 20 network approach." Transportation Research Record 1453 (1994). 21

[20] Oswald, R. Keith, William T. Scherer, and Brian L. Smith. "Traffic flow forecasting using 22 approximate nearest neighbor nonparametric regression." Final project of ITS Center project: 23 Traffic forecasting: non-parametric regressions (2000). 24

[21] Jeong, Ranhee, and R. Rilett. "Bus arrival time prediction using artificial neural network 25 model." In Intelligent Transportation Systems, 2004. Proceedings. The 7th International IEEE 26 Conference on, pp. 988-993. IEEE, 2004. 27

[22] Bishop, Michael K. "School bus approach notification method and apparatus." U.S. Patent 28 4,325,057, issued April 13, 1982. 29

[23]. Winkler, Josef K., Vernon L. Stant, and Stephen G. Wilson. "Bus arrival notification 30 system and methods related thereto." U.S. Patent 6,700,506, issued March 2, 2004. 31

15

[24] Jones, M. Kelly. "Advance notification systems and methods utilizing a computer network." 1 U.S. Patent No. 6,748,320. 8 Jun. 2004. 2

[25] Dow, Chyi-Ren, Po-Yu Lai, and Jhen-He Ye. "A moving context-aware and location-based 3 paratransit system." International Journal of Internet Protocol Technology 9.1 (2015): 34-43. 4

[26] Svennerberg, Gabriel. Beginning Google Maps API 3. Apress, 2010. 5

[27] Wang, Fahui, and Yanqing Xu. "Estimating O–D travel time matrix by Google Maps API: 6 implementation, advantages, and implications." Annals of GIS 17, no. 4 (2011): 199-209. 7

[28]. Subramanian, Karthik; Srikanth, R. (January 21, 2014). "Now, Apps for Live Traffic Feed". 8 The Hindu. 9

[29] "Twilio." - APIs for Text Messaging, VoIP & Voice in the Cloud. Accessed Jun 26, 2016. 10 https://www.twilio.com/ 11

[30] "Google Maps JavaScript API V3 Reference." Google Developers. Accessed June 12, 2016. 12 https://developers.google.com/maps/documentation/javascript/reference. 13

14