Impacts of Large Vehicles on Traffic Safety in Freeway Interchange Merging Areas and Improvement Measures

In order to work out the characteristics of traffic operation on freeway interchange merging areas under different ratio of large vehicles, aerial photography technique based on unmanned aircraft and other observation device were used to investigate the merge section traffic data firstly. Based on survey data and actual traffic characteristics, the paper then established simulation models with VISSIM with calibrated car-following and lane changing behaviour models. Next, the paper analysed the influence of traffic volume and acceleration lane length under various traffic compositions to evaluate safety status of interchange merge sections by traffic conflict technique and speed consistency index. Finally, two kinds of safety promotion strategies, speed limit and setting forbidden line, were evaluated and the results indicated these measures can raise safety level by about 10%~15% under certain conditions, which were determined as the best applicable conditions.


Introduction
Freeway interchange merging area is the bottleneck and accident-prone spot.In general, at merg ing section, due to aggressive lane changing behaviours, vehicles on mainline may slow down or pull out to overtake, which increases the risk of t raffic accident.Therefore, a wellfunctioning merging area has a significant effect on improving traffic safety [1].
Currently, large vehicles have increasingly become one of the research hotspots with construction of freeway and development of modern logistics industry.Owing to poor performance of large vehicles, there is serious interference in traffic flow, especially at interchange merging areas, wh ich leads to a decline in stability of traffic flow.Agent et al. evaluated geometric align ment design of expressway with Heavy goods vehicle (HGV) as the objective [2].Gazis and Herman first proposed the concept of "Moving Bottleneck" and attempted to exp lain it according to traffic flow wave theory when they found that a HGV may cause a slow motorcade in 1992 [3]. Later, nu merous researches on formation mechanism of moving bottleneck have been conducted, and some traffic phenomena and experimental data were analysed fro m the perspective of moving bottleneck theory [4][5][6][7][8].However, most of these findings focus on basic sections of highway, not on interchange areas.
Regarding merging area, many previous studies have focused on influence factors of traffic safety and capacity [9][10][11].S. Li et al. established crash risk predict ion model and defined HCRI index based on traffic conflict technique for freeway interchange merging area [12].L. Lei analysed the gear-alternating control regulat ion based on FI cellular automaton model and concluded that gearalternating control can imp rove traffic situation significantly under large traffic volu me [13].Sun et al. simu lated the operation characteristics of merging area under open boundary condition by calibrated carfollowing model and lane changing model [14].
Fro m the literature rev iew, there seems to be three main issues in Ch ina: 1) few studies have linked large vehicles to interchange merg ing areas; 2) the models used to analyse merg ing segment in Ch ina are based on research results fro m other countries, but the differences of traffic co mposition and other conditions between China and other countries, isn't taken into consideration; 3) the research on merg ing areas is mainly focused on the capacity rather than safety.
In this paper, with regard to a directional on-ramp, simu lations are conducted with the calibrated car following model and lane changing model in VISSIM.Then traffic conflict technique and velocity consistency index are selected to evaluate the impact of HGV on traffic safety at confluence area.Finally, some measures are proposed to improve safety status of merg ing section and their applicable conditions are given.

Data collection
The traffic survey was conducted at a semi-cloverleaf interchange with a typical merging area, which is composed of a three-lane mainline and a one-lane ramp if emergency lane is neglected.The width of every lane is ICTTE 2017 3.75m.The horizontal curve radius of main line is 1500m and ramp radius is 650m.Un manned aerial vehicle (UA V) photography was applied to shoot video in good conditions.Finally three hours high-precision video was obtained.Moreover, radar velocimeters were used to measure speed of gore nose, midpoint and end point of merge area, as shown in the Figure 1.

Safety evaluation
Co mmon traffic safety evaluation methods include three categories, namely accident-based method, conflict-based method and traffic flow characteristic based method.In this paper, the latter two methods are considered.The traffic conflict and traffic characteristic indicator are defined as follows.

Traffic conflict
Traffic conflict technique, as an approach that overcomes the lack of reliab le accident records, is widely used for safety evaluation in China [15][16][17].However, most previous studies on traffic conflict pay attention to the intersection rather than merg ing section.In fact, there exists great difference between them because of distinct traffic and road conditions.Therefore, the key of this method is how to identify the effective conflict at merging segments.
TTC (time to collision) is the critical parameter to determine a traffic conflict.In this paper, two typical types of conflicts are defined according to the angle between the vehicle travelling directions: 1) rear -end conflict, the angle is in [0°, 15°); 2) sideswipe conflict, the angle belongs to [15°, 85°].Relevant studies pointed that the TTC thresholds of these two conflicts are 4.7s and 4.2s, respectively [16].To make study convenient, this paper selects 4.2s as the TTC threshold for all conflicts.

Speed consistency
Based on traffic characteristics, s ome studies found that speed consistency can well reflect the traffic flow stability and safety situation of highway [18,19].Speed consistency indicators commonly used include the following two categories.
• SDI, speed difference index, can reflect directly safety of adjacent sections of highway, mainly including ΔV 85 and 85(ΔV).
• SSI, speed statistic index, reflect traffic safety by analysing distribution of velocity statistics such as σ and σ/μ, which is more applicable to a single segment.As to confluence area, which is different fro m the basic section in traffic characteristics, SDI index may be more suitable.The equation of ΔV 85 and 85(ΔV) are listed as follows: where V 1 and V 2 are respectively the speed of the same vehicle on d ifferent segments.In fact, ΔV 85 overestimates the safety while 85(ΔV) is more precise because the relationship and interaction of adjacent sections is taken into account [20].Meanwh ile , a novel indicator SRC based on 85(ΔV) is proposed to reflect the reduction coefficient of speed.
3 VISSIM-based merging area model

Model assumption
In this paper, the purpose is to investigate the safety situation under different ratio of HGV.Therefo re, a freeway merge section model was established.Specific to the characteristics of merge sections and the functions of the software VISSIM some assumptions are listed as follows.
• Assume that both the mainline and on-ramp are straight and the slope of merge section is zero.
• Assume that the angle of mainline and ramp is 30° and parallel type acceleration lane is adopted.
• Neglect the effect of weather conditions and lateral clear width on the traffic flow characteristics.
• Assume that the design speed of main line and ramp respectively are 100km/h and 60km/h.
• Assume that the traffic co mposition on the entrance ramp is the same as the main line.

Vehicle parameter calibration
We divide the vehicles into two types, cars and large vehicles (HGV), and their various parameters are shown in Table 1.

Car following model calibration
In this paper, Wiedemann99 car-fo llo wing model, which is widely used in highway, is adopted to simu late the car

ICTTE 2017
following behavior at on-ramp bottleneck.Table 2 gives the calibrated parameters of car following model.

Lane changing model calibration
In general, at a merging section, the lane changing behaviours have the most significant influence on the traffic flow.Due to aggressive lane changing behaviours, vehicles on the main line may slow down or enter into another lane.Lane changing model in VISSIM includes mandatory and discretionary lane changing behaviours and thus it is adopted in this study.The parameters of model is calibrated as Table 3.The merg ing segment model is built in VISSIM and the investigation and statistical results (traffic volu me, ratio of HGV, etc.) were input into simu lation model.For more stable output results and smaller random erro r, each simu lation lasts for 3600s and simulat ion in the same condition is carried out 10 t imes, thus the average value of the total 10 simulations is adopted as final result.

Model verification
To test the validity of the model, this paper compared the simu lation results with the observations by two indicators, average speed of section and number of t raffic conflicts.The verificat ion results are shown in Table 4.It can be seen that Error of section speed and traffic conflict respectively are less than 5% and 15%, which are within acceptable range in terms of engineering application.Therefore the traffic simulation model is valid.

Simulation and results analysis 4.1 Simulation scheme
This paper aims to explore the influence of large vehicle mixing rate on traffic safety in freeway interchange merging areas.Nevertheless, the security situation is not only affected by large vehicles, but also by the traffic conditions (traffic volu me, merg ing ratio , etc.), road conditions (acceleration lane length, curve radius, number of lanes, etc.) and control conditions (speed control, channelization, traffic markings, etc.).
This paper establishes kinds of sub-models of simu lation by inputting different traffic conditions and road conditions, and thus the elements of different simulation scenarios are shown in Table 5.

Relationship of ratio of HGV vs. mainline traffic volume
The relationship between the ratio of HGV and traffic volume at the merg ing area is studied in this section.The confluence area with the accelerat ion lane length of 250m and merg ing ratio of 0.25 was taken as an examp le.The simulation results are shown in Figure 2 and Figure 3.
Fro m Figure 2, it can be observed that with the increase of ma inline traffic volu me, the conflict quantity at merge segment also increases obviously when the traffic volu me is less than a critical value.When ratio of HGV is not greater than 30%, total amount of collisions increases roughly in proportion to traffic volu me.Once the ratio of HGV reaches or exceeds 40%, the collisions will encounter a decline after a rapid growth.The difference is that the peak value of curve occurs at 1500veh/h/ln when proportion of HGV is 40% while it occurs at 1250veh/h/ln under 50%~60% of HGV.The decline is in line with expectations because the interaction and interference will increase when volume is large enough, which may cause larger traffic density, slower velocity and smaller possibility o f collision.In addition, traffic conflicts also increases significantly with DOI: 10.1051/matecconf/201712404004 ICTTE 2017 the increase of the rate of large vehicles under the traffic volume of 1000veh/h/ln~1500veh/h/ln, which means the large vehicles, as the moving bottleneck of merging area, exactly have great influence on safe operation of traffic flow, particularly when ratio of HGV arrived at 40%.According to Figure 3, under the same t raffic volu me, the SRC increases along with the increment of the rat io of HGV as a whole, which implies the decline in security.Nevertheless, the curve showed different gro wth trend under different traffic volu me.When volume is less than 750veh/h/ln, the SRC gro ws slowly, which suggests that the moving bottleneck caused by HGV has not yet formed.The SRC increases approximately in a line with the raise of rat io of HGV when the volu me is between 1000~1250veh/h/ln.Specially, when volu me continues to mount up, the SRC will increase exponentially which indicates HGV will have a greater impact on t raffic stability under larger volume.Meanwhile, the curves under 1500veh/h/ln and 1750veh/h/ln have a similar shape and small spacing, indicating that the influence of traffic volume is weakened.

Relationship of ratio of HGV vs. acceleration lane length
To exp lore the relat ionship of HGV and accelerat ion lane length, the simu lations under the situation that the mainline volu me is 1250veh/h/ln, merging rat io is 0.25 and the acceleration lane length is between 200m and 450m were carried out with the results shown in the Figure 4 and Figure 5.
In Figure 4, it can be seen that the length of acceleration lane has significant influence over the traffic operation in the confluence area.In general, with the increase of the acceleration lane length, the number of traffic conflicts tends to decline when the length is less than a critical value and then it remains nearly stable.When the ratio of HGV is 20%, the conflicts are no longer reduced when accelerat ion lane length reaches 300m, while this crit ical value will climb to 350m when HGV account for 30%~40% and 400m when HGV exceed 50%.Fro m another perspective, the acceleration lane exists a border, not the longer the better.Fro m Figure 5, it is found that the speed stability is improved with acceleration lane length increasing.A lso, under the same length of accelerat ion lane and traffic volume, the more large vehicles, the lower the speed consistency and stability.Th is because the fact that large vehicles will cause more interference when changing lane at merging area.

The impact of speed limit measure
Under the same condition, a higher proportion of large vehicles may cause lo wer security on interchange onramp joints.By the analysis of section 4.3, setting reasonable length of acceleration lane can significantly improve traffic condit ions.However, there are great difficult ies in reconstructing an acceleration lane for an existing freeway interchange.
In addition to b ig size and poor performance of HGV, the main reason for the safety problems is that the vehicle enter into the main line too early, which cause large speed difference between ramp vehicles and main line vehicles due to inadequate acceleration.Thus, a possible measure is limiting the main line speed to reduce speed difference.DOI: 10.1051/matecconf/201712404004

ICTTE 2017
Because the design speed of mainline and ramp are respectively 100km/h and 60km/h, the speed of inner, middle and outer lane of main line are respectively limited as 90km/h, 80km/h and 70km/h in this paper.Then the simu lations under different conditions were conducted when merg ing rat io and acceleration lane length were fixed to 0.25 and 300m.The reduction rate of t raffic conflicts is selected as the indicator to evaluate effcts and the result is shown in Figure 6.
Fro m Figure 6, at the same traffic volu me, a better improvement effect of sub-lane speed limit measure is observed at the condition of enhanced HGV proportion.And this measure can be implemented for optimu m results when traffic volu me is 1000veh/h/ln~1500veh/h/ln and ratio of HGV isn't less than 40%.

The impact of forbidden line
Regarding the parallel type accelerat ion lane, too large merging angle also may increase driv ing risk.Therefore M UTCD points out that the solid white line, namely forb idden line, setting at the end of optional diagonal or chevron approach markings on gore nose (Figure 7), can avoid premature merging into main line [21].However, the length of forb idden line and its adaptability in China and imp lementation effect have not been proven and evaluated.Taking the merge section under the condition that mainline volu me arrives at 1250veh/h/ln, merging rat io is 0.25, accelerat ion lane is 300m long and HGVs account for 40% as an examp le, the simulat ion was carried out to study the effectiveness of forbidden line, with the results shown in Figure 8. Obviously, when forbidden line is 15m long, the best imp rovement has been achieved.And the effectiveness will decline when continuing to increase the length because this actually reduces acceleration lane length and the choice of merging space.
The merge section with forb idden line length of 15m, acceleration lane length of 300m, merg ing rate of 0.25, traffic volu me o f 500veh/h/ln~1750veh/h and HGV ratio of 20%~60% was taken as an example to co mpute the reduction rate of traffic conflicts.And the K-means cluster analysis method was applied for the effectiveness evaluation of the data obtained.The final clustering result obtained from ten iterations is shown in Table 6.According to the cluster results, the effectiveness of forbidden line is d ivided into four levels, excellent, good, med iu m and poor.The effectiveness under various conditions is shown in Table 7.

ICTTE 2017
Fro m Table 7, when the large vehicle rat io is 30%~50% and mainline t raffic volu me is 1000veh/h/ln~ 1500veh/h/ln, setting forbidden line has excellent or good effect.When the traffic volu me is relatively s mall or large, setting forbidden line has poorer effect along with the increment of the large vehicle rat io.The reasons are following: when the traffic volu me is relatively small, the traffic flow is under free state and large vehicles slightly influence small vehicles.Ho wever, when the traffic volume is relatively large, the tra ffic flo w is under saturated state, and forbidden line cannot effectively reduce traffic conflicts.

Conclusion
On the freeway interchange merg ing area, the mov ing bottleneck caused by large cars is not independent, which may affect traffic safety when lateral lane changing behavior and longitudinal car-following behavior occur.The impact of large vehicles and the improvement measures were systematically researched in this paper with the goal of traffic safety.
The influence of the moving bottleneck caused by large vehicles on the traffic flow is significant.W ith the increment of large vehicle rat io, traffic safety decreases for the traffic conflicts and SRC increases significantly.When traffic volu me is less than 750veh/h/ln or more than 1500veh/h/ln, the impact of HGV is relatively small.This indicates that moving bottleneck has not yet formed at merging area at that time.
Setting appropriate length of acceleration lane can enhance traffic safety level effectively.Under the condition of traffic vo lu me of 1250veh/h/ln, the length of acceleration lane should be set to 300m, 350m and 400m when large vehicle ratio are 20%, 30%~40% and 50%~ 60%, respectively.
Besides, speed limit is an effective method to improve traffic operetion at merge sections, especially when traffic volu me of main line is 1000veh/h/ln~1500veh/h/ln and ratio of HGV is less than 40%.Forbidden line proposed by MUTCD is also considered in this paper and we conclude that 15m long forbidden line may have better improvement effect under mediu m traffic volu me condition.Meanwhile, by K-means cluster analysis, when traffic volu me is 1000veh/h/ln~1500veh/h/ln and the ratio of HGV is 30%~50%, setting forb idden line on freeway interchange merg ing area can effectively pro mote the traffic safety; especially, when the large vehicle ratio is about 40%, the positive effect is the most obvious.
However, this paper was limited to analyzing data fro m the perspective of part of road and traffic conditions.In reality, merging ratio, nu mber of lanes, ramp radius, type of acceleration lane and other traffic control or environment conditions should be taken into account.Future studies may look into these influencing factors as variables for modeling.

Figure 1 .
Figure 1.M erging segment and speed measuring locations.

Figure 2 .
Figure 2. Relationship of traffic conflict vs. traffic volume.

Figure 3 .
Figure 3. Relationship of SRC vs. traffic volume.

Figure 4 .
Figure 4. Relationship of conflict vs. acceleration lane length.

Figure 5 .
Figure 5. Relationship of SRC vs. acceleration lane length.

Figure 6 .
Figure 6.The effect of speed limit at merging area.

Figure 8 .
Figure 8.The effect of forbidden line at merging area.

Table 1 .
Parameters of car and HGV in model.

Table 2 .
Parameters of car following model.

Table 3 .
Parameters of car following model.

Table 4 .
Result of speed and conflict test.

Table 5 .
Conditions of traffic simulation.

Table 6 .
Final cluster centres in simulation.

Table 7 .
Effect analysis for the setting of the forbidden line.