You are on page 1of 37

CAPACITY ESTIMATIONS FOR TYPE B WEAVING AREAS BASED ON GAP ACCEPTANCE

by Ponlathep Lertworawanich1 The Pennsylvania Transportation Institute The Pennsylvania State University 201 Transportation Research Building University Park, PA 16802 U.S.A Phone: (814) 863-1043 Email: ponlathep@psu.edu and

Lily Elefteriadou, Ph.D. The Pennsylvania Transportation Institute The Pennsylvania State University 201 Transportation Research Building University Park, PA 16802 U.S.A Phone:(814)-863-7923 Fax: (814) 863-3039 Email: axe11@psu.edu

Total number of Words is 7496. November 11, 2000

Paper submitted for presentation at the 80th Annual Meeting of the Transportation Research Board and publication in the Transportation Research Record
1

Corresponding Author

1 Lertworawanich, P. and Elefteriadou, L.

ABSTRACT Though weaving areas are one of the major highway facilities that have long been investigated by many researchers, the estimation of capacity along weaving areas has not been well researched and validated. Most of the previous literature concentrates on speed estimation methods for weaving and non-weaving vehicles, and level of service estimation. The Highway Capacity Manual (HCM) 2000 weaving methodology (due to be published toward the end of 2000) includes capacity estimates for weaving segments, which are based on the assumption that the density at capacity is the boundary of Level of Service or LOS E/F, 27 pc/km/ln .

The objective of this study is to develop a method for estimating capacity of Type B weaving areas based on gap acceptance and linear optimization. In addition, traffic data were obtained from a site located on the QE Expressway in Toronto, Canada, and analyzed to identify its capacity. Field estimates of capacity were compared to those resulting from the new methodology, and from the HCM 2000 methodology. It was concluded that the proposed methodology provided better capacity estimates of the study site than those of the HCM 2000 methodology, when compared to field observations. Additional data collection is required to validate the proposed model for a wide variety of Type B weaving segments, and for various traffic and highway design conditions.

2 Lertworawanich, P. and Elefteriadou, L.

1. BACKGROUND Weaving is defined as the crossing of two or more traffic streams traveling in the same general direction along a significant length of highway without the aid of traffic control devices (Highway Capacity Manual HCM, 1997). Weaving areas are formed when a merge area is closely followed by a diverge area, or when an on-ramp is closely followed by an off-ramp and the two are joined by an auxiliary lane (HCM, 1997). Classification of weaving areas is based on the number of lane changes required by each weaving movement. Figure 1 illustrates the method for determining the configuration type of a weaving area (HCM, 1997).

Though weaving areas are one of the major highway facilities that have long been investigated by many researchers, the estimation of capacity along weaving areas has not been well researched and validated. Most of the previous literature concentrates on speed estimation methods for weaving and non-weaving vehicles, and level of service estimation. The Highway Capacity Manual (HCM) 2000 weaving methodology (due to be published toward the end of 2000) includes capacity estimates for weaving segments, which are based on the assumption that the density at capacity is the boundary of Level of Service or LOS E/F, 27 pc/km/ln (Roess and Ulerio, 1999). Thus, the capacity is defined as any combination of flow that causes the density to attain 27 pc/km/ln. Lertworawanich (2000) used gap acceptance theory and linear optimization to address lane-changing processes on ramp weaves (Type A weaving segments) and developed a model to estimate the capacity of ramp weaves. The high level of lane-changing activity around weaving areas is one of the major contributors to the complication of weaving area analysis. Gap acceptance theory can be used to consider potential lane changing activity and thus estimate the capacity of the weaving segment. The objectives of this paper are:

3 Lertworawanich, P. and Elefteriadou, L.

To develop a capacity model for type B weaving areas based on gap acceptance modeling and linear optimization.

To compare the results of the proposed methodology to observed values at a type B weaving segment in Toronto, Canada. Speed and flow data were obtained at this site over several days, which provided a range of possible capacity values.

To compare the estimates of the proposed method and field observations to capacity estimates from the HCM 2000 weaving methodology.

The next section in this paper presents the methodology and the development of the capacity model for type B weaving areas. The third section presents a discussion on observing capacity, and summarizes the capacity values obtained from the weaving section in Toronto, Canada. Next, capacity estimates using the proposed methodology and the HCM 2000 methodology are presented, followed by conclusions and recommendations.

2. METHODOLOGY AND MODEL DEVELOPMENT Figure 2 presents a Type B weaving segment. In general, traffic demand from ramp and freeway lanes of a type B weaving section can be decomposed into several components as shown in Figure 2.

4 Lertworawanich, P. and Elefteriadou, L.

In the HCM, capacity is defined as the maximum flow of vehicles that can reasonably be expected to traverse a point or uniform section of a lane or roadway during a given time period under prevailing roadway, traffic, and control conditions (HCM 1997). In this study, the capacity of type B weaving areas comprises of two parts: a) the maximum flow rate of vehicles for the three weaving lanes (ramp lane, lane 1, and lane 2), and b) the maximum flow rate of vehicles for the other lanes in the weaving area. The first part of the weaving segment capacity is estimated by the procedure described in this paper, while the second part is assumed to be the equivalent basic freeway segment capacity. Thus, it is assumed that very few vehicles weave from the inside freeway lanes, and that the capacity of the weaving segment is not affected by them. When only the three lanes adjacent to the shoulder of the weaving sections are considered, the maximum traffic flows are decomposed into seven components as illustrated on Figure 3. Note that these are the resultant flow rates and not the demands shown in Figure 2. These would be equal to or lower than the respective demands depending on whether capacity is reached or not.

Based on the configuration shown in Figure 3, the capacity of the three lanes adjacent to the shoulder can be mathematically expressed as:

CAPACITY = MAX (VFF1 + VRF + VRR +VFR11 + VFR10 +VFF2 + VFR2 )

(1)

To compute the capacity based on Equation (1), an optimization technique can be deployed along with some constraints. In this study the linear optimization technique is used to accomplish this task. From Figure 3, the constraints that have to be imposed are: 1) The capacity of the weaving segment cannot exceed the capacity of the equivalent basic freeway segment and can be expressed as:

5 Lertworawanich, P. and Elefteriadou, L.

VRF + VRR BFC (Basic Freeway Segment Capacity) VFF 1 + VFR11 + VFR 10 BFC VRF + VFF 1 + VFR 11 + VFR 2 BFC VFF 2 + V FR 2 BFC VFR10 + V RR BFC In this paper, the ideal capacity of the basic freeway segment is 2185 pc/h/ln, which is obtained from the data. 2) Capacity is highly dependent on the proportion of weaving traffic, thus the volume ratios (defined as the ratio of weaving vehicles to total vehicles), of the traffic flows must remain the same as the respective ratio of the demands. These constraints are expressed as: DRF VRF = W0 = D VRF + VRR R DFR 11 + DFR 10 + DFR 2 VFR 11 + VFR 10 + VFR 2 = W1 = VFR 11 + VFR 10 + VFR 2 + VFF 2 + VFF 1 DF 1 + DF 2

3) The bounds of each component specified in Equation (1) should not exceed their maximum values. For example, the resultant weaving flow cannot exceed the maximum weaving flow for the respective demand volume ratios. These constraints are mathematically expressed as: VFF 1 BFC (Basic Freeway Segment Capacity) VFF 2 BFC VFR11 BFC VFR10 MAX (VFR10 ) VRR BFC VRF MAX (VRF ) VFR 2 MAX (VFR 2 )

6 Lertworawanich, P. and Elefteriadou, L.

Based on the capacity equation and constraints, the final linear programming optimization equations are expressed as follows:
CAPACITY = MAX (V RR + V RF + V FF 1 + V FR 11 + V FR 10 + V FR 2 + V FF 2 ) (Objective Function) subject to capacity constraint 1 : capacity constraint 2 : capacity constraint 3 : capacity constraint 4 : capacity constraint 5 : V RF + V RR BFC (Basic Freeway Segment Capacity) V FF 1 + V FR11 + V FR10 BFC V RF + V FF 1 + V FR11 + V FR 2 BFC V FF 2 + V FR 2 BFC V FR 10 + V RR BFC V FR11 V FR11 + V FR 10 + V FR 2 = W1 + V FR 10 + V FR 2 + V FF 2 + V FF 1

lanes 1 and 2 volume ratio constraint : ramp volume ratio constraint :

V RF = W0 V RF + V RR

bound of V RR , V FF 1 , V FF 2 , V FR 11 : BFC bound of V FR10 bound of V RF bound of V FR 2 :V FR 10 MAX(V FR10 ) : V RF MAX (V RF ) : V FR 2 MAX (V FR ) (2)

To obtain the constraint values for the last three constraints, we need to estimate MAX(VRF), MAX(VFR2 ), and MAX(VFR10). These can be expressed as the potential capacity for each of the traffic streams, and they are related to the gap availability, gap size and gap acceptance characteristics. Thus, gap acceptance theory will be used to estimate these three terms. To estimate these, the time headway distributions of traffic flows from both the freeway and the ramp, as well as the critical gap must be known in advance. Thus, the proposed methodology is based on gap acceptance parameter estimation to calculate the maximum lane changes for each traffic stream, and finally compute capacity by using the linear programming optimization equations above. The chart shown in Figure 4 provides a step-by step outline of the methodology. As shown in Figure 4, the first step is to obtain traffic-related parameters, traffic demands (including origin destination demands), and speeds of traffic on each lane. Note that in the process of critical lag

7 Lertworawanich, P. and Elefteriadou, L.

estimation, which will be described later in this section, speeds of traffic on each lane are crucial parameters. The second step is to compute the critical lag for lane-changing processes by using the equations developed by Drew et.,al (1967) and the findings of Raff and Hart (1950). These will also be used in the fourth step, in estimating the maximum lane changes. Thus, all the parameters required for the optimization Equation (2) can be obtained. The last step is to calculate the capacity of the weaving areas using the linear programming technique. The remainder of this section provides detailed information on each of these steps.

2.1 Critical Lag Estimations As mentioned above, the critical lag is estimated here using the critical gap estimation method presented by Drew et al (1967) and the findings of Raff and Hart, (1950) that the critical lag is approximately 0.2 seconds less than the average critical gap. This is the only methodology found in the literature for estimating the critical gap. From the definition of the merging process (Drew et al, 1967), the most favorable types of gap for a driver to merge would be both the optional merge and the ideal merge because they would not cause turbulence in the freeway traffic. It was assumed in that paper that the typical vehicles average acceleration, when approaching at a speed lower than the speed of the main traffic or the weaving vehicle has to accelerate, is a linear function of the speed of the merging vehicle. Using equations of motion, the minimum ideal gap for merging was developed based on the time required for safe stopping and the time loss due to acceleration, which can be expressed as:

8 Lertworawanich, P. and Elefteriadou, L.

1) time required for safe stopping Time Required for Safe Stop = L f + Lr + 2 RT v where L f and Lr = the length of freeway and ramp vehicles v = speed of merged stream RT = driver reaction time 2) time loss due to acceleration a b v ( a / b) v ln( ) u+ v a b u Time Loss due to Acceleration = + b v b v where v = speed of merged traffic stream u = speed of merging traffic stream a and b = constants

The equation for estimating the ideal gap for merging was expressed as (Drew et al, 1967): L f + Lr u+v + 2 RT + + with ( u < v) v bv b v where L f and Lr = the length of the freeway and ramp vehicles respective ly v = speed of merged traffic stream u = speed of merging traffic stream RT = driver reaction time a and b = constants ( a / b) v ln( a b v ) a bu

TT =

9 Lertworawanich, P. and Elefteriadou, L.

Subsequently, the equation for the ideal lag for merging, considering the findings of Raff and Hart (1950), can be expressed as: ( a / b) v ln( a bv ) a b u - 0.2

TT =

u+v + with v bv b v where L f and L r = the length of the freeway and ramp vehicles respective ly + 2 RT + v = speed of merged traffic stream u = speed of merging traffic stream RT = driver reaction time a and b = constants

L f + Lr

( u < v)

(3)

To determine the ideal lag with Equation (3), the parameters a and b must be estimated. From experiments in Australia performed by Knox (1964), the suggested values of a and a/b were 4.8 mph/sec and 80 mph respectively and a reaction time of 1 second was recommended. This recommendation might not be a good representation for current traffic conditions; however, there are no recent studies conducted to re-evaluate these findings. In order to address the case that the merging vehicle is approaching at a speed higher than the main traffic stream or when the merging vehicle has to decelerate during the lane-changing process, some modification of equation (3) is required. In this study, the relationship between speed and deceleration is assumed and expressed as: dv = f g dt where f = coefficien t of friction between vehicles' tires and pavement g = gravitatio nal accelerati on (32.2 ft/s 2 ) Using the same concept, the ideal gap for merging process is composed of two elements; 1) time required for safe stopping, as defined above, and 2) time due to deceleration , which can be expressed as:

10 Lertworawanich, P. and Elefteriadou, L.

Time Loss = Normal Travel Time Required Travel Time for the same dist ance during Decelerati on 1 u T + ( f g) T 2 1 f g 2 2 Normal Travel Time Required = =T + T u 2 u vu Travel Time during D eceleratio n (T) = f g 1 (v u )2 Time Loss = 2 u f g where v = speed of merged traffic stream u = speed of merging traffic stream f = coefficien t of friction between vehicles' tires and pavement (-0.2) g = gravitatio nal accelerati on (32.2 ft/s 2 ) Thus, the critical lag estimation equation for the case when the speed of the merging vehicle is higher than the speed of the main traffic stream can be expressed as: TT = 1 (v u )2 0.2 * * * * * * * *(u > v) v 2 u f g where L f and L r = the length of freeway and ramp vehicles + 2 RT L f + Lr (4)

RT = driver reaction time u = speed of merging traffic stream v = speed of merged traffic stream f = coefficien t of friction between of vehicles' tires and pavement g = gravitatio nal accelerati on

2.2 Development of Rear Lag Distribution In this section, the procedure to estimate the distribution of rear lags is presented. The rear lag distributions are subsequently used to compute the maximum lane-changes based on gap acceptance theory. To estimate the maximum lane changes from ramp to freeway, the distribution of the rear lag RL01, as shown in Figure 5, must be computed. To compute the maximum lane changes from freeway lane 2 to freeway lane 1, the distribution of the rear lag RL21, as shown in Figure 6, must be computed. To compute the maximum lane changes from

11 Lertworawanich, P. and Elefteriadou, L.

freeway lane 1 to ramp 0, the distribution of the rear lag RL10, as shown in Figure 7, must be computed.

Time headway distributions for the oncoming traffic stream on Lanes 0 and 1 are symbolized as FHL0 (t) and FHL1(t) respectively. In this study, the time headways are represented by Cowan M3 distributions since the Cowan M3 distribution provides a good fit for time headways greater than the critical lag, which are crucial in this study (14). Also, it is tractable in terms of integration. Time headway distributions are expressed as: FHL 0 (t ) = 1 e ( t t m ) ; t > t m =0 ; otherwise where FHL0 (t) = cdf of HL0 time headway = q ( 1-t m q)

q = flow rate of lane 0 in veh/sec ond t m = platoon ve hicles' time headway in second 1- = percentage of platoon ve hicles FHL1 ( t ) = 1 e (t t m ) ; t > tm =0 ; otherwise where FHL1 (t) = cdf of HL1 time headway q = ( 1-t m q) q = flow rate of lane 1 in veh/sec ond t m = platoon ve hicles' time headway in second 1- = percentage of platoon ve hicles

2.2.1 Determination of Rear Lag RL01 for Lane 1 From Figure 5, assuming that the arrival process of vehicles on lane 1 is independent of the arrival process of vehicles on lane 0 of the freeway, the rear lag distribution RL01 can be calculated based on the random arrival process and it is expressed as:

12 Lertworawanich, P. and Elefteriadou, L.

f RL01 (t ) = =

1 FHL1 (t ) E (T )

e ( tt m ) ; t tm (5) tm + 1 = ; t < tm tm + where f RL01 (t) is the rear lag RL01 distributi on FHL1 (t) is the cdf of time headways for lane 1 E(T) is the expected value of the time headway for lane 1 2.2.2 Determination of Rear Lag RL21 for Lane 1 From Figure 6, assuming that the arrival process of vehicles on lane 1 is independent of the arrival process of vehicles on lane 2 of the freeway, the rear lag distribution RL21 can be calculated based on the random arrival process and it is expressed as: 1 FHL1 (t ) E (T )

f RL21 ( t ) = =

e ( tt m ) ; t tm (6) tm + 1 = ; t < tm tm + where f RL21 (t) is the rear lag RL21 distributi on FHL1 (t) is the cdf of time headways for lane 1 E(T) is the expected value of the time headway for lane 1

13 Lertworawanich, P. and Elefteriadou, L.

2.2.3 Determination of Rear Lag RL10 for Lane 0 From Figure 7, assuming that the arrival process of vehicles on lane 1 is independent of the arrival process of vehicles on lane 0 of the ramp, the rear lag distribution RL10 can be calculated based on the random arrival process and it is expressed as: f RL10 (t ) = 1 FHL 0 (t ) E(T )

e ( tt m ) = ; t tm (7) tm + 1 = ; t < tm tm + where f RL10 (t) is the rear lag RL10 distributi on FHL0 (t) is the cdf of time headways for lane 0 E(T) is the expected value of the time headway for lane 0

2.3. Maximum Lane Change Estimations As mentioned at the beginning of this section, the maximum lane changes can be estimated by gap acceptance theory. In this study, the general gap acceptance equation provided in the revised Traffic Flow Theory Monograph (1992) is used to estimate the maximum lane changes and can be expressed as: Qm = Q p f ( t ) g ( t )dt
0

(8)

where Qm = maximum traffic volume merging into main stream (veh/sec). Q p = main stream volume (veh/sec). f (t ) = probabilit y density function of the gaps in the major stream. g ( t ) = the number of merging vehicles which can enter into a gap of duration t.

14 Lertworawanich, P. and Elefteriadou, L.

2.3.1 Estimation of Maximum lane changes From Lane 0 to Freeway Lane 1 The maximum lane changes from Lane 0 to freeway Lane 1 can be estimated by plugging the distribution of the rear lag RL01 computed earlier into Equation (8):

MAX (VRF ) = = Q p f ( t ) g ( t ) dt
0

= q1 f RL01 ( y ) (
t0

y t0 )dy t1 (9)

q1 e ( t0 t m ) 2 ( + tm ) t1

2.3.2 Estimation of Maximum lane changes From Freeway Lane 2 to Freeway Lane 1 The maximum lane changes from freeway Lane 2 to freeway Lane 1 can be estimated by plugging the distribution of the rear lag RL21 computed earlier into Equation (8):

MAX (VFR 2 ) = = Q p f ( t ) g ( t )dt


0

= q1 f RL21 ( y ) (
t0

y t0 )dy t1 (10)

q1 e ( t0 t m ) 2 ( + tm ) t1

15 Lertworawanich, P. and Elefteriadou, L.

2.3.3 Estimation of Maximum lane changes From Freeway Lane 1 to Ramp Lane 0 The maximum lane changes from freeway Lane 1 to ramp lane 0 can be estimated by plugging the distribution of the rear lag RL10 computed earlier into Equation (8): MAX (VFR10 ) = = Q p f ( t ) g (t )dt
0

= q 0 f RL10 ( y ) (
t0

y t0 ) dy t1 (11)

q0 e ( t0 t m ) 2 ( + t m ) t 1

After the maximum lane changes are available, a linear programming technique is used to calculate the capacity of three weaving lanes of weaving section. The solution of the optimization Equation (2) is the required capacity of the weaving section (for three lanes: lane 0, lane 1, and lane 2). The total capacity of the whole section of the weaving area is the sum of the capacity of the three weaving lanes and the equivalent basic freeway segment capacity of all the lanes except these three lanes.

3. FIELD DATA In this study, a site along the Highway 401 freeway system in Toronto, Canada was selected for data collection to compare capacities observed, to those estimated by the proposed methodology, and those estimated by the HCM 2000 methodology. Detector stations are located as shown on Figure 8.

Each station can continuously provide information on speeds, flows, and densities of the oncoming traffic from both the freeway and the ramp for every 20-second interval. Complete data sets were available for ten days (October 22-31, 1997). The posted speed limit on the 401

16 Lertworawanich, P. and Elefteriadou, L.

freeway is 100 km/h. Free-flow speeds during off-peak periods were found to vary from 100 km/h to 120 km/h. The percentage of heavy vehicles in the traffic stream in the area of the study site is roughly 10 percent according to the Ministry of Transportation of Ontario. From the traffic data, observed on October 22-31, 1997, the speed and vehicle count data were summarized in 20second intervals for each of the lanes. The 20-second interval data were aggregated into 5-minute intervals. The study site regularly operates under congested conditions and appears to experience breakdown conditions due to the weaving operations, as evident from the data. Variations of the traffic demands and volume ratios throughout the study period resulted in different capacities for the weaving segment for every day of the study. These were extracted for each day using time-series plots of speed and flow rates. In addition, in this study, three types of flow rates, corresponding to three proposed capacity definitions, were extracted from the data each day of the data collection: 1) The maximum sustained 15-Minute rate of flow that can be accommodated by a freeway segment under prevailing traffic and roadway conditions (the HCM 1997). This is the HCMdefined capacity, which corresponds to the maximum sustained pre-breakdown flow rate, and not necessarily the flow immediately preceding breakdown. 2) The maximum queue-discharge flow rate of the weaving section. This is the flow rate under congested conditions, which were defined here as corresponding to speeds below 80 km/h for at least a 15-minute period. 3) The breakdown flow rate, defined here as the 15-minute flow rate immediately prior to the breakdown. Breakdown is defined as the traffic condition when the average speed of all lanes on the section dropped below 80 km/h for at least a 15-minute period. Based on these three definitions, three different flow rates were extracted from the data and are shown in Table 1.

17 Lertworawanich, P. and Elefteriadou, L.

According to the HCM definition, (maximum sustained 15-minute flow rate), the capacity of the weaving section ranges from 5970 to 6820 veh/h. The maximum queue-discharge flow rates of the weaving section range from 5540 to 6620 veh/h, which were invariably lower than the maximum sustained 15-minute flow rates. This confirms the existence of the two-capacity phenomenon for this weaving segment. Table 1 also shows that there are variations of the observed flow rates from day to day, especially on October 25-26, 1997, Saturday and Sunday. On October 25, 1997, the observed maximum queue-discharge flow rate was much lower than the maximum sustained 15-minute flow rate. On October 26, 1997, the observed maximum sustained 15-minute flow rate was lower than those of the other days. The breakdown flow rates on these two days were also lower than those on the weekdays. This is consistent with observations regarding impact on capacity due to non-commuter driver populations. In general, the maximum observed flow rates for each of the days varied within 9 percent of the average.

4. CAPACITY ESTIMATION USING THE NEW METHOD AND THE HCM 2000 METHOD In this section, the capacity of the weaving area is estimated by the method developed here and the HCM 2000 method using the available traffic data as inputs. The last portion of this section presents the comparison between these two and the field data.

4.1 Capacity Estimates from the Proposed Methodology To implement the proposed methodology, the demand data, speeds of traffic, and volume ratios on each of the weaving lanes, must be extracted from the data. One of the major difficulties in obtaining data for weaving area traffic analysis, is that speed-flow databases do not typically

18 Lertworawanich, P. and Elefteriadou, L.

include volume ratios (i.e., origins and destinations for the approaching traffic streams), which are a critical input in the capacity estimation. Thus, to be able to estimate the capacity of the section in this study, a possible range of capacity values was estimated, using all possible values of volume ratios for each lane, ranging from 0 to 1. Figure 9 illustrates one sample result from the proposed methodology using the observed data from October 24, 1997 as inputs.

As shown in Figure 9, the estimated capacity on October 24, 1997 ranges from 4948 to 8320 veh/h for the 4-lane weaving area shown in Figure 8, depending on the volume ratios. When the volume ratio of lane 0 (W0) is fixed, the estimated capacity increases with the increase of volume ratio of lanes 1 and 2 (W1). This occurs because as weaving vehicles move from freeway lane 1 to lane 0 they create more available space for weaving vehicles from lane 0. But this is valid until the volume ratio of lanes 1 and 2 (W1) reaches 0.1. The estimated capacity remains constant when the value of W1 is between 0.1 and 0.5. This reflects the fact that weaving vehicles from the freeway are not required to change lanes in order to complete their weaving maneuvers, which is a major characteristic of type B weaving areas; hence, the capacity of the section does not change with changes of W1 in this range. However, if the number of weaving vehicles from the freeway plus the number of weaving vehicles from lane 0, desiring to occupy lane 1 of the freeway is greater than the capacity of lane 1, the capacity of the weaving section is starting to drop. The same trend is not observed when the volume ratio of lanes 1 and 2 (W1) is fixed, because weaving vehicles from lane 0 must change lanes to complete their maneuvers along the weaving section. Thus, the capacity model developed here incorporates the fact that weaving traffic streams from the freeway and ramp have different impacts on the capacity of the weaving area due to different lane-changing characteristics. The capacity estimates for the other days of data collection did not differ more than 200 veh/h from those of

19 Lertworawanich, P. and Elefteriadou, L.

October 24, 1997, for the same volume ratios. In general the capacity of the weaving section depends mostly on the volume ratios and speeds of traffic streams.

4.2 Capacity Estimates from the HCM 2000 Weaving Methodology According to the HCM 2000, the capacity of weaving areas can be estimated using free-flow speed, length, number of lanes, and volume ratios of the weaving sections as input parameters. Capacity estimates (based on the HCM 2000) of the weaving area shown in Figure 8, are summarized in Table 2. Note that these estimates correspond to a free-flow speed of 100 km/h, and a 4-lane weaving segment with length 550 m. The fourth column of the table presents the final capacity values (interpolated from the HCM 2000 tables).

Table 2 shows that capacity of the weaving area is decreasing with the increase of the volume ratio of the section. When the volume ratio is less than 0.1, the capacity of the weaving section is constrained by the basic freeway segment capacity.

4.3 Comparison of Capacity Estimates against Observed Capacities Even though the available traffic data did not directly provide the information on volume ratios which are an important input for the proposed and the HCM 2000 methodologies, the values of these parameters could be confined to some reasonable ranges by using traffic data from the upstream and downstream ramps and the freeway. From Figure 8, the maximum number of vehicles traveling from the on-ramp to the off-ramp was restricted by traffic counts from upstream and downstream ramp detector stations. In other words, the maximum number of rampto-ramp vehicles could not exceed the minimum number of vehicle counts from detector stations on upstream and downstream ramps. With this information, ranges of volume ratios of traffic

20 Lertworawanich, P. and Elefteriadou, L.

could be extracted from the data for any given day. Flow rates, speeds, and ranges of volume ratios on lanes 0, 1, and 2 were input into the proposed methodology to compute the corresponding capacity for each day. With the application of the HCM 2000, ranges of volume ratios for the whole weaving section were input to estimate capacity of the section for each of the days. In this study, only the maximum sustained 15-minute flow rates were considered so that the comparison of the results from both the HCM and the proposed methodology could readily be examined. Table 3 shows the results of capacity estimates from both methodologies together with the observed capacities (the maximum sustained 15-minute flow rates.)

The results indicated that the proposed methodology provided narrower ranges of the capacity estimates, which were relatively closer to the observed capacities than those provided by the HCM 2000 methodology. From Table 3, the HCM 2000 capacity estimate ranges did not cover the capacities observed in the field on any given day. The proposed methodology provided better capacity estimate ranges, which included the capacities observed in the field six out of ten days of the observations. In addition, the HCM methodology frequently overestimated the capacities and gave broader ranges of capacity estimates relative to the observed capacities.

5. CONCLUSIONS AND FUTURE RESEARCH The major findings of this study are as follows: 1. A capacity estimation model was developed for Type B weaving segments, as a function of volume ratios, speeds, and traffic flow rates on each lane of the freeway and the ramp. 2. The capacity model shows that weaving traffic streams from the freeway and the ramp have different impacts on capacity of weaving areas; therefore, capacity

21 Lertworawanich, P. and Elefteriadou, L.

estimates from the model require both volume ratios as input parameters. In the HCM 2000, capacity estimates are a function of the demand volume ratio for the entire weaving segment. 3. The model shows that the volume ratio of the traffic on lane 0 has more influence on capacity estimates of the weaving section than that of the traffic on lanes 1 and 2 up to a certain value. This reflects the main characteristics of the type B weaving areas in that weaving vehicles from freeway do not require lane changes along weaving areas when weaving demand is not high. The HCM 2000 does not consider the impacts of varying demands on capacity estimates for Type B weaving areas. 4. Based on traffic data collected at the study site over a period of days, the proposed capacity model provided relatively better results of the capacity estimates than the HCM 2000 methodology did. 5. The methodology developed and outlined here can be expanded to address other weaving configurations that require one lane-change for weaving vehicles. Based on the theoretical concept of the proposed methodology, the following research would enhance the methodology to provide better results: 1. Collect data to evaluate the critical gap estimation models and subsequent capacity estimations. The gap acceptance model used here was developed over 30 years ago, and may not be valid under current traffic conditions. 2. Consider other types of time headway distributions. 3. Expand the methodology to address other configurations of weaving areas.

22 Lertworawanich, P. and Elefteriadou, L.

ACKNOWLEDGEMENTS The authors wish to sincerely thank David Tsui, Gilbert May, and Phil Masters of the Ministry of Transportation of Ontario, for providing the traffic data used in this study.

REFERENCES 1. Cassidy, M.J. and May, A.D. Proposed Analytical Technique for Estimating Capacity and Level of Service of Major Freeway Weaving Sections. Transportation Research Record 1320, TRB, National Research Council, Washington, D.C. 1991. 2. Blumenfeld, D.E., and Weiss, G.H. The Effects of Gap Acceptance Criteria on Merging Delay and Capacity at an Uncontrolled Junction. Traffic Engineering & Control, January 1975, pp. 16-20. 3. Drew, D.R., Buhr, J.H., and Whitson, R.H. Determination of Merging Capacity and Its Applications to Freeway Design and Control. Transportation Research Record 244, HRR, National Research Council, Washington D.C., 1968 4. Drew, D. R., Lamotte, L. R., Buhr, J. H. and Wattleworth, J. A. Gap Acceptance in the Freeway Merging Process. Transportation Research Record 208, HRR, National Research Council, Washington D.C., 1967, pp. 1-36. 5. Fazio, J. and Rouphail, N. Freeway Weaving Sections: Comparison and Refinement of Design and Operations Analysis Procedures. Transportation Research Record 1091, TRB, National Research Council, Washington, D.C., 1986 6. Special Report 209: Highway Capacity Manuals 1985 Edition, TRB, National Research Council, Washington D.C., 1985. 7. Special Report 209: Highway Capacity Manuals 1997 Edition, TRB, National Research Council, Washington D.C., 1997. 8. Iizuka, T., and Makigami, Y. Evaluation of Weaving Traffic Stream Using Merging Probability. Transportation and Traffic Theory 1993, pp. 53-69 10. Makigami, Y., and Matsuo, T. A Merging Probability Calculation Method Considering Multiple Merging Phenomena. Transportation and Traffic Theory 1990, pp. 21- 38. 11. Mimura, T., and Hisashi, M. Highway Merging Problem with Acceleration Area. Transportation Science, Vol. 4, No.3, 1970, pp. 205-213.

23 Lertworawanich, P. and Elefteriadou, L.

12. Pignataro, L.J., McShane, W.R., and Roess, R.P., Weaving Areas Design and Analysis, NCHRP Report 159 (1975). 13. Roess, R.P. and Ulerio, J. Weaving Area Analysis in the Year 2000 Highway Capacity Analysis. Preprints of the 79th Annual Meeting of Transportation Research Board, TRB, National Research Council, Wasington D.C., 2000. 14. Special Report 165: Traffic Flow Theory Monograph. Transportation Research Board, TRB, National Research Council, Wasington D.C., 1992, http://wwwcta.ornl.gov/cta/research/trb/tft.html. 15. Wang, M.H., Cassidy, M.J., Chan, P., and May, A.D. Evaluation of the Capacity of Freeway Weaving Sections. Journal of Transportation Engineering 1993. 16. Lertworawanich, P. Capacity Estimations for Ramp Weaves Based on Gap Acceptance. M.S. Thesis, Pennsylvania State University, State College PA, 2000. 17. HCM 2000 - Weaving Segments, Chapter 24, Highway Capacity and Quality of Service Committee, 2000

24 Lertworawanich, P. and Elefteriadou, L.

TABLE 1 Observed Flow-Rates of the Weaving Section

Max 15 Min. Date Flow Rate a 22-Oct-97 23-Oct-97 24-Oct-97 25-Oct-97 26-Oct-97 27-Oct-97 28-Oct-97 29-Oct-97 30-Oct-97 31-Oct-97 6820 6390 6450 6870 5970 6540 6680 6800 6620 6460 Avg. Speedb 81 63 87 104 105 68 97 91 78 88

Max. Queue-Discharge Flow Rate 6620 6390 6200 5540 5850 6540 6580 6480 6280 6160 Avg. Speed 72 63 49 45 51 68 55 50 58 47

Breakdown Flow Rate Flow Rate 6480 5990 6450 5650 5930 5910 6680 6510 6510 6230 Avg. Speed 95 97 87 96 97 94 97 103 87 93

Flow Rate in veh/h a Speed in km/h b

25 Lertworawanich, P. and Elefteriadou, L.

TABLE 2 Capacity Estimates of the Weaving Section Based on the HCM 2000 (From the HCM 2000 Manuscript)

100 km/h Free-Flow Speed Length of Weaving Area (m) Volume Ratio 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
a b

450a 9200 9010 8280 7650 7120 6670 5760 5000

600a 9200 9200 8530 7900 7370 6670 5760 5000

550b 9200c 9137 8447 7817 7287 6670 5760 5000

From the HCM 2000 manuscript (17) From Interpolation c Capacity expressed in pcph

26 Lertworawanich, P. and Elefteriadou, L.

TABLE 3 Comparison of Capacity Estimates against Observed Capacities

Observed Capacity Date (Max. sustained 15-min. ) 22-Oct-97 23-Oct-97 24-Oct-97 25-Oct-97 26-Oct-97 27-Oct-97 28-Oct-97 29-Oct-97 30-Oct-97 31-Oct-97 6820 6390 6450 6870 5970 6540 6680 6800 6620 6460 6850-7280 6760-7050 6340-7180 6660-6970 6550-8170 6740-7060 6530-7000 6640-6960 6620-6960 6330-6770 6940-8740 7250-8740 6540-8760 6940-8740 5995-8760 7100-8730 6870-7960 6950-8710 7010-8710 6870-7310 Proposed Capacity Estimates HCM Capacity Estimates

All capacity estimates expressed in veh/h (based on 10% heavy vehicles)

27 Lertworawanich, P. and Elefteriadou, L.

FIGURE 1 Determination of Configuration Type for Weaving Areas .

Vo1 or V02 Vw1 or Vw2

Vw1 or Vw2 Vo1 or Vo2

Number of Lane-changes Required by Movement Vw1 0 1 2

Number of Lane-Changes Required by Movement Vw2 0 Type B Type B Type C 1 Type B Type A N/A 2 Type C N/A N/A

28 Lertworawanich, P. and Elefteriadou, L.

FIGURE 2 Type B weaving area with traffic demand.

DF2 DF1

DFF2 DFR2 DRF DRR DFF1 DFR11 DFR10

Lane 2 Lane 1

DR DF1 represents the oncoming traffic demand from lane 1 of the freeway DF2 represents the oncoming traffic demand from lane 2 of the freeway DR represents the oncoming traffic demand from the ramp DFF1 represents the through traffic demand (from freeway to freeway lane 1) DFR11 represents the weaving traffic demand (from freeway lane 1 to off ramp without lane-change) DFR10 represent the weaving traffic demand (from freeway lane 1 to off ramp with lane-change) DFF2 represents the through traffic demand (from freeway to freeway lane 2) DFR2 represents the weaving traffic demand (from freeway lane 2 to freeway lane 1) DRF represents the weaving traffic demand from the ramp DRR represents the ramp to ramp through traffic demand W 1 represents the volume ratio of traffic demand of freeway lanes 1 and 2 (DFR11+DFR2 +DFR10)/(DF1 +DF2 ) W 0 represents the volume ratio of traffic demand in ramp lane (DRF/DR)

29 Lertworawanich, P. and Elefteriadou, L.

FIGURE 3 Type B weaving area with traffic flow rates at capacity.

VFF2 VFR2 VRF VRR VFF1 VFR11 VFR10

Lane 2 Lane 1

VFF1 represents the through traffic flow rate (from freeway to freeway of lane 1), at capacity VFR11 represents the weaving traffic flow rate (from freeway lane 1 to off ramp without lane-changes), at capacity VFR10 represents the weaving traffic flow rate (from freeway lane 1 to off ramp with lane-changes), at capacity VFF2 represents the through traffic flow rate (from freeway to freeway of lane 2), at capacity VFR2 represents the weaving traffic flow rate (from freeway lane 2 to freeway lane 1), at capacity VRF represents the weaving traffic flow rate from the ramp, at capacity VRR represents the ramp to ramp through traffic flow rate, at capacity

30 Lertworawanich, P. and Elefteriadou, L.

FIGURE 4 Methodology flow chart.

Traffic Parameter Inputs Traffic Flow Rates of Each Lane Volume Ratios Geometric Information of Weaving Areas Speeds of Traffic on Each Lane

Critical Lags for Merging Processes

Rear Lag Distribution Estimation

Maximum Lane Changes Using Gap Acceptance Th eory

Capacity Estimations

Definition of Capacity

Objective Function

Constraints

Basic Freeway Segment Capacity

Linear Programming Optimization

Capacity of Weaving Areas

31 Lertworawanich, P. and Elefteriadou, L.

FIGURE 5 Rear Lag RL01 of Freeway Lane 1.

HL1

Lane 2 A Lane 1

RL01

FL01

Lane 0

Represents an oncoming main stream vehicle. Represents a merging vehicle. Represents a main stream vehicle.

32 Lertworawanich, P. and Elefteriadou, L.

FIGURE 6 Rear Lag RL21 of Freeway Lane 1.

HL1 FL21

RL21

Lane 2 A Lane 1 B

Lane 0
Represents a vehicle on Lane1 of the freeway. Represents a lane-changing vehicle on Lane2 of the freeway . Represents a vehicle on Lane1 of the freeway .

33 Lertworawanich, P. and Elefteriadou, L.

FIGURE 7 Rear Lag RL10 of Ramp Lane 0

HL0

Lane 2 A Lane 1

RL10

FL10

Lane 0

Represents a merging vehicle. Represents an oncoming vehicle from ramp. Represents a vehicle from ramp.

34 Lertworawanich, P. and Elefteriadou, L.

FIGURE 8 Study Site.

Detector Station
Lane 3 Highway 401 "through" Lane 2 Lane 1 Lane 0

Detector Station

Highway 400 ramp lanes

120 meters

Detector Station

550 meters

Detector Station

35 Lertworawanich, P. and Elefteriadou, L.

FIGURE 9 Capacity Estimates from the New Method.

Capacity Estimates Based on the Proposed Methodology

9000 8000 7000 6000 w0=0.0 w0=0.2 w0=0.4 Volume Ratio of Lane 0 w0=0.6 w1=0.0 w0=0.8 w1=1.0 w1=0.9 w0=1.0 w1=0.8 w1=0.7 w1=0.6 w1=0.5 w1=0.4 w1=0.3 w1=0.2 5000 Capacity (veh/h) 4000 3000 2000 1000 w1=0.1 0

Volume Ratio of Lanes 1 and 2

capacity w0=0.0 w0=0.1 w0=0.2 w0=0.3 w0=0.4 w0=0.5 w0=0.6 w0=0.7 w0=0.8 w0=0.9 w0=1.0

w1=0.0 8320 8112 7902 7210 6864 6656 6517 6418 6344 6287 6240

w1=0.1 8320 8320 8318 7625 7279 7071 6933 6834 6760 6702 6656

w1=0.2 8320 8320 8318 7625 7279 7071 6933 6834 6760 6702 6656

w1=0.3 8320 8320 8318 7625 7279 7071 6933 6834 6760 6702 6656

w1=0.4 8320 8320 8318 7625 7279 7071 6933 6834 6760 6702 6656

w1=0.5 8320 8320 8318 7625 7279 7071 6933 6834 6760 6702 6656

w1=0.6 7875 7875 7875 7553 7207 6999 6860 6761 6687 6630 6583

w1=0.7 7291 7291 7291 6969 6623 6415 6276 6178 6103 6046 5999

w1=0.8 6854 6854 6854 6531 6185 5977 5838 5740 5665 5608 5561

w1=0.9 6513 6513 6513 6190 5844 5636 5498 5399 5325 5267 5221

w1=1.0 6240 6240 6240 5918 5572 5364 5225 5126 5052 4994 4948

36 Lertworawanich, P. and Elefteriadou, L.

Lists of Tables Table 1 Table 2 Table 3 Observed Flow-Rates of the Weaving Section Capacity Estimates of the Weaving Section Based on the HCM 2000 Comparison of Capacity Estimates against Observed Capacities

Lists of Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Determination of Configuration Type for Weaving Areas Type B Weaving Area with Traffic Demand Type B Weaving Area with Traffic Flow Rates at Capacity Methodology Flow Chart Rear Lag RL01 of Freeway Lane 1 Rear Lag RL21 of Freeway Lane 1 Rear Lag RL10 of Ramp Lane 0 Study Site Capacity Estimates from the New Method

You might also like