Travel Time Prediction Models and Reliability Indices for Palestine Urban Road in Baghdad City

Travel Time estimation and reliability measurement is an important issues for improving operation efficiency and safety of traffic roads networks. The aim of this research is the estimation of total travel time and distribution analysis for three selected links in Palestine Arterial Street in Baghdad city. Buffer time index results in worse reliability conditions. Link (2) from Bab Al Mutham intersection to Al-Sakara intersection produced a buffer index of about 36% and 26 % for Link (1) Al-Mawall intersection to Bab AlMutham intersection and finally for link (3) which presented a 24% buffer index. These illustrated that the reliability get worst for link (2), (1) and (3) respectively during the peak period. Extra delay is observed on link(1), (2) and (3) in terms of 95% percentile travel time of about (301.9, 219.4, and 193.8)sec. for Link (1, 2 and 3) respectively. Higher value for 95% travel time is obtained for link (1). Travel time index (TTI) of 4.2 %, 4.9% and 4% is obtained for Link (1, 2 and 3) respectively. Maximum value for delay per km that obtained for link (1) which is about 266 sec/km and 268 sec./km for link (3) and 244 sec/km for link(2). Different predicted model for the three studied links of Palestine street have been developed based on the obtained field data. A best fit is presented as compared the predicted models with the observed field travel time data for all the models of studied links which illustrated that the predicted model can present the actual field data.


Research Objective
The aim of this research is the estimation of total travel time and distribution analysis for three selected links in Palestine arterial street in Baghdad city which is considered as one of the most important residential and commercial area in Baghdad city due to dramatic change that produce potential pressure on daily trip generation and attraction.A statistical methods is needed to model travel time distribution on which reliability indices measurements including buffer index, buffer time and 95% percentile travel time were developed based on.

Introduction
The travel time of urban arterial in urban city played an important role in measuring the performance of traffic transportation system.Different studies are employed to model the distribution for travel time, [1], [2], [3] and [4] concluded for a lognormal distribution.Polus, (1979) concluded for a Gamma distribution; Al-Deek and Eman (2006) proposed a Weibull one [5], [6].In Taylor and Susilawati (2012) and Susilawati et.al. (2012) the Burr distribution is adopted [7], [8].Federal Highway Administration (FHWA) has defined travel time reliability is as consistency or dependability in travel time as measure from day to day and across different times of the day [9].The 95th percentile travel time and Buffer Index (BI) and Planning Time Index (PTI) considered as performance indicators for travel time reliability.Travel time distribution and empirical based approach is the base for development of these indices.Study of travel time reliability can help in understanding the variation in travel time and aid in transportation system management [10].

Data Collection
The field data are collected for the selected sections of Palestine Street for Link (1), ( 2) and (3) respectively.Congestion of traffic conditions is taking the major part during peak hours of the day from (12:00 to 4:00 p.m.) on Monday 18 May and Tuesday 19 May 2016 which is selected to study the variations of total travel time and travel delay time for each link in the selected site of Palestine arterial street.GPS essentials measurement equipped with cell phone is applied to compute travel time with a 30 set of data point is recorded at peak period from (12:00 to 4:00 p.m.) on each selected link corridor.Figure (2) shows the control points for mapping the distance for acceleration and deceleration and starting of each link.

Results and Discussions 4.1. Analysis of Travel Time
In this research a consideration to congestion of traffic conditions is taking the major part during peak hours of the day from (12:00 to 4:00 p.m.) is preferred to study the variations of total travel time and travel delay time for each link in the selected site of Palestine arterial street.A number of sample run of 30 is provided for each link in the south direction as described from the link name in the previous section of site selection.Travel time analysis is required for measuring the performance and to have indication about the operation efficiency of Palestine arterial corridor.Figures (3), ( 4) and (5) clarified the variations of total travel time for each link including stop time and ( acceleration and deceleration time).Figure ( 6) and ( 7) illustrated the travel time delay for each link; Travel time delay estimated from total travel time minus ideal travel time for each link depend on posted speed limit and distance for each one.Based on the obtained results shown in figures, link (1) from Al-Mawall to Bab-Al-Mutham intersection produced the highest values for travel and delay time than other two links due to several reasons which can be referred to the surrounded of commercial, residential and educational mix land use that produced and attracted a large number of daily trips also its attributed to the traffic condition of the link itself which controlled by a check point at the start of the route for the link near Al-Mwaal intersection and cause excess delay during the peak hours reach about 5 minutes stopping causing the state of stop and slow moving conditions for vehicles.Also its appeared that travel time varied during the time period and three maximum peak point at (12:00-1:00p.m),(1:30-2:30p.m)for all links and additional third peak point from (3:00-3:30) for link (1) were observed.Maximum travel time of 518.5 sec.and delay time of 457sec.which is about 88% of travel time is lost due to condition of traffic congestion on link (1).Also 238sec.total travel time for link (2) with delay time of 206sec.which is about 86% loss of travel time.And for link (3) a maximum travel time of 210 sec.and delay time of 172 sec.which is about 81.9% is lost due to delay congestion.GPS essentials equipped with cell phone is applied to compute the delay component (acceleration, deceleration and stopped delay) of signalized intersections at the selected site by determining the critical points for acceleration , deceleration and stopping then measure the required time, see Figure (2) of control points for each link.Since HCM used equation for estimate the control delay which is applied for fixed signalized signal during the day and this not achieved in our locally signalized intersection which almost managed by police man that organized the traffic movement at signalized intersections otherwise its produced error in the obtained results and not matching the reality conditions.Figures (8),( 9) and (10) and tables (1), ( 2) and (3) presented the components of total delay for intersections in the selected site for ٣٠ sample run.It's clear that stopped delay compromise the major part of intersection delay for all links studied.Also the acceleration time delay is higher than deceleration time as shown in the obtained results due to the conflict vehicles in the intersection which make the driver more alert during passing the intersection and this take more time to accelerate.

Estimation of Travel Time Distribution
A 30 time data set of 15min.period is collected from field data using GPS essentials equipped with cell phone as explained earlier to study the behavior of travel time variation on the three links (1), ( 2) and ( 3) that provide significant variation and multimodal shape due to the delay at signalized intersection and impedance due to traffic jam at the selected links at peak period.A normal, lognormal distribution are applied and fitted for the field data using (SPSS ver.21 statistical software).The graphical representation of histogram with normal curve for travel time distributions are shown in Figures (10) to (12) for Link (1), ( 2) and ( 3) in this research.The travel time distribution is fitted to normal for Link (1) and , log-normal for Link(2) and (3).Based on normality test for Shapiro-Wilk column in Tables (4) to (9) for Link (1) ,( 2) and ( 3) respectively which illustrated the significant level of p -value greater than 0.05.See Table (10).

Travel Time Model
Different predicted model for the three studied links of Palestine street have been made based on the obtained field data as shown below: Travel Time for Link (1): Where: TT: Total Travel Time (sec.) DT: Delay Time at signalized Intersection (sec.)The summary of stepwise regression linear models are displayed in Tables (7) to (9) respectively for Link (1), link (2) and link (3).The validation of the three travel time model have been illustrated in Figures (13) to (15) between the estimated travel time and observed travel time from field data.An additional data have been measured and not included in the model building to complete the process of validations.A best fit is presented as compared the predicted model with the observed field travel time data for all the models of studied links which illustrated that the predicted model can present the actual field data.The goodness of fit for predicted model and field observed data have been checked using chi-square test as shown in Tables (10) to (12) for the study links.For < ; there are no significant difference between predicted model and observed field data.

Reliability Measurement
The variability of travel time effect on its reliability measurement and then produced extra arrival time which had a real cost.Furthermore reliability of travel time is an important topics for increasing safety, quality life for road users, to produce less delay for their trips.Also it's a good indicators for improving the overall system operations and management.Buffer time reliability measure is explained as [9]:  2) and (3).Increasing the buffer time index results in worse reliability conditions.Link (2) produced a buffer index of about 36% and 26 % for Link (1) and finally for link (3) which present a 24% buffer index.These illustrated that the reliability get worst for link (2) (1) and (3) respectively.Also buffer time of (62 , 59, and 38 ) sec. for Link(1, 2and 3) respectively is obtained based on average travel time for each link, that mean additional 62, 59 and 38 sec.buffer time is provided from the average value to ensure 95% arrive on time at the destination of arterial corridor for link (1),( 2) and (3).Also Figure (16) show the 95% percentile travel time for observed links which presents the extra delay that perceived on each link (301.9, 219.4,and 193.8)sec.for Link (1, 2 and 3) respectively.Higher value for 95% travel time is obtained for link (1)   Greater than 1.0 meaning taking a longer travel time by about 420, 490 and 400 percent of free travel time to travel the three segment length respectively with the higher travel time index for Link(2) and link(1) and finally with link(3).This demonstrated the heavily congested conditions.Also Figure (18) prove the above statement interms of estimating the delay per travelling kilometer for each segment length (1.03, 0.520, and 0.620) Km for link (1, 2 and 3) respectively studied in this research.

Conclusions
It can be drawn the following points: 1. Maximum travel time of 518.5 sec.and delay time of 457sec.which is about 88% of travel time for link (1); also 238sec.total travel time for link (2) with delay time of 206sec.which is about 86% loss of travel time.For link (3) a maximum travel time of 210 sec.and delay time of 172 sec.which is about 81.9% is lost due to delay congestion.2. Stopped delay compromise the major part of intersection delay for all links studied.Also the acceleration time delay is higher than deceleration time due to the conflict vehicles in the intersection.3. Buffer time index results in worse reliability conditions.Link (2) produced a buffer index of about 36% and 26 % for Link (1) and finally for link (3) which presented a 24% buffer index.4. The 95% percentile travel time of about (301.9, 219.4,and 193.8) sec.for Link (1, 2 and 3) respectively.Higher value for 95% travel time is obtained for link (1). 5. Travel time index TTI of 4.2 %, 4.9% and 4% is obtained for Link (1, 2 and 3) respectively.Higher travel time index for Link (2) and link (1) and finally with link (3) respectively.6. Maximum value for delay per km that obtained for link(1) which is about 266 sec/km and 268 sec./km for (3) and 244 sec/km for link (2).

Data Collected 3 . 1 .
Study AreaPalestine street is one of the most important major arterial streets in Baghdad city due to the majority of surrounded area of different mix land uses; commercial, educational and residential that provide potential pressure of generation and attraction additional daily trips.Palestine street is located in the East of Baghdad and it runs parallel to the west of Army Canal between Al-Mustansiriyah Sequare through Beirut square to the end of it at Maysalone sequar; 6 divided lane carriageway 3-lane in each direction.Figure (1) presents the three links that have been considered in this research namely as Link(1); From Al-Mawall Intersection to Bab Al-Muatham Intersection of 1.03 Km length, Link (2); From Bab Al-Muatham Intersection to Al-Sachara Intersection of 520m length, and Link (3); From Al-Sachara Intersection to Beirut Intersection of 620 m length respectively.The selected corridor for Palestine street links passed through three signalized intersection (Bab Al-Mutham Intersection, Al-Sachara Intersection and Bairuit Intersection) which also take into consideration their impedance and delay effect on travel time variability, reliability and distribution estimation.

Fig. 1 .
Fig. 1.Study Area Urban Arterial Palestine Street with Selected Three Links.

Fig. 2 .
Fig. 2. Control Points for Deceleration, Acceleration and Stopping Time Measurement for Studied Links.

Fig. 3 .
Fig. 3.Total Travel and Delay Time Variations During Peak period Time for Link (1).

Fig. 4 .
Fig. 4. Total Travel and Delay Time Variations During Peak Period Time for Link (2).

Fig. 5 .
Fig. 5.Total Travel and Delay Time Variations During Peak period Time for Link (3).
Fig. 11.Normal and Log-Normal Travel Time Distribution for Link (2).
Figure (17) depicted the maximum value for delay per km that obtained for link(1) which is about 266 sec/km and 268 sec./km for (3) and 244 sec/km for link (2).