Acceleration and Deceleration Rates in Interrupted Flow Based on Empirical Digital Tachograph Data

Abstract

The acceleration and deceleration rates are crucial road design reference values in terms of traffic safety. The purpose of acceleration and deceleration lanes is to reduce the speed difference between the mainline and ramp to minimize rear-end collision. Thus, in traffic safety and traffic flow operational aspects, the acceleration and deceleration lane lengths need to be long. Moreover, excessively long minimum acceleration or deceleration lane length regulations may cause hesitancy to connect new facilities with the road. Despite their relevance, acceleration and deceleration rates have not been updated for many decades. In this study, we analyze the digital tachograph data of vehicles stopped at signalized intersections at red lights and empirically deduce the acceleration and deceleration rates that reflect recent vehicle performance and driver behavior. Finally, we suggest a new corner clearance distance to safely connect the new road near a signalized intersection in urban and rural areas derived from our empirical acceleration and deceleration rates.

1. Introduction

The acceleration and deceleration rates of a vehicle are key parameters in determining the acceleration or deceleration lane length and corner clearance distance of an intersection or interchange. The purpose of acceleration and deceleration lanes is to reduce the speed difference between the mainline and ramp to minimize rear-end collision. AASHTO [1] recommended the merging speed to be within 10 km/h slower than the mainline operating speed. According to Lee [2], when the merging speed is 20 km/h or more slower than the mainline operating speed, the total delay increases more than five times. Furthermore, the regulations about the corner clearance distance from the intersection or interchanges required a sufficient offset to reduce the conflict between acceleration or deceleration vehicles from the intersection and access vehicles. Thus, in traffic safety and traffic flow operational aspects, the acceleration and deceleration lane lengths need to be long. Moreover, excessively long minimum acceleration or deceleration lane length regulations may cause hesitancy to connect new facilities with the road. When there is strict limited spare space in a highly urbanized area, it causes a dramatic increase in road construction costs. Despite their relevance, acceleration and deceleration rates have not been updated for many decades. It is argued that acceleration and deceleration rates should be changed by reflecting recently improved vehicle performance. Another opinion is that the reference values of acceleration and deceleration should be a comfortable rate rather than vehicles’ engine performance.

In this study, we analyze the digital tachograph (DTG) data and empirically deduce the acceleration and deceleration rates that reflect recent vehicle performance and driver behavior. Unlike uninterrupted flow, vehicles suffer from stop-and-go states in signalized intersections due to the various acceleration and deceleration performances between vehicle types. The vehicles approach the signalized intersection while decelerating to stop at red lights and accelerating after the lights change to green [3,4]. These acceleration or deceleration behaviors differ depending on traffic composition and driver characteristics [5,6]. Thus, this study analyzes the various type of vehicles’ DTG data whose vehicles stopped at signalized intersections at red lights to show a driver’s preference behavior when accelerating and decelerating. Finally, we suggest the corner clearance distance of a signalized intersection reflecting an empirically observed acceleration and deceleration rate.

2. Literature Review

2.1. Current Regulations about Auxiliary Lane

Regulations in South Korea on road structure and facility standards provide guidelines on the mainline and ramp designs, such as acceleration or deceleration lanes. According to MOLIT [7], most vehicles approaching a deceleration lane drive slower than the free-flow speed of the mainline. The Korean road standards consider 1.96 m/s², which is 20% of gravity acceleration; drivers usually slow down with a comfortable deceleration rate [7]. The minimum acceleration lane length is determined by the required acceleration of vehicles. In Korean regulation, vehicle performance is based on a 13-ps/ton truck, reflecting a high vehicle composition of trucks. On level terrain, the acceleration rate of a vehicle is calculated using Equation (1), and the average acceleration to the cruising speed is shown in

Table 1. Driving speed and average acceleration rate in Korean road design regulation [7].

Driving Speed (km/h)	70	63	60	55	51	50	45	42	40	35	30	28	20
Average Acceleration Rate (m/s2)	0.28	0.34	0.36	0.41	0.46	0.47	0.54	0.59	0.63	0.74	0.88	0.95	1.38

.

$$\mathrm{a}=\frac{dv}{dt}=\frac{g}{1+ϵ}[\frac{75\times 3.6\xi \left(BHP\right)}{W\times V}-\mu -\frac{R\times A}{{3.6}^2\times W}{V}^2=\frac{29.484}{V}-0.0933-\frac{0.134}{14000}{V}^2$$

(1)

where g: gravitational acceleration (9.8 m/s²), μ: rolling friction coefficient (0.01), ε: acceleration resistance ratio (0.05), R: air resistance coefficient (0.03 kg·sec²/m⁴), ξ: mechanical efficiency (0.9), A: projected area (6.2 m²), W: vehicle weight (14,000 kg), BHP: engine poser (PS), BHP/W = 0.013 (PS/kg), V: free-flow speed (km/h).

Based on the Korean regulations, the minimum acceleration and deceleration lane length on level terrain is calculated using Equation (2), and the calculated values are shown in

Table 2. Korean minimum deceleration and acceleration lane length [7].

Mainline Design Speed (km/h)			120	110	100	90	80	70	60
Ramp Design Speed (km/h)	80	Deceleration Lane Length (m)	120	105	85	60	-	-	-
			245	120	55
	70		140	120	100	75	55	-	-
			335	210	145	50	-	-	-
	60		155	140	120	100	80	55	-
			400	285	220	130	55	-	-
	50	Acceleration Lane Length (m)	170	150	135	110	90	70	50
			445	330	265	175	100	50	-
	40		175	160	145	120	100	85	65
			470	360	300	210	135	85	-
	30		185	170	155	135	115	95	80
			500	390	330	240	165	110	70

.

$$L=\frac{v_2^2-v_1^2}{2\times {3.6}^2\times d}=\frac{V_2^2-V_1^2}{2\times {3.6}^2\times a}$$

(2)

where L: acceleration or deceleration lane length (m), v₁: decelerated speed (km/h), V₁: initial speed (km/h), v₂: initial speed (km/h), V₂: accelerated speed (km/h), d: deceleration rate (1.96 m/s²), a: acceleration rate (m/s², Table 1).

The American standards are similar to the Korean standards. According to ASSHTO [1], approximately two-thirds of drivers apply deceleration rates greater than 2.0 m/s² to come to a stop line. The deceleration lane length in ASSHTO [1] is estimated based on such deceleration rates. For the acceleration lane, they consider the merging speed that is within 10 km/h of the mainline operating speed. ASSHTO’s [1] guideline could be seen that acceleration and deceleration lanes mainly depend on the design speed. The minimum acceleration and deceleration lane lengths on level terrain are presented in

Table 3. Minimum deceleration and acceleration lane lengths [1].

Mainline Design Speed			Design Speed on Ramp
			Stop Condition	20 km/h	30 km/h	40 km/h	50 km/h	60 km/h	70 km/h	80 km/h
			Initial Vehicle Speed on Ramp
	Design Speed	Merge/Diverge Speed	0 km/h	20 km/h	28 km/h	35 km/h	42 km/h	51 km/h	63 km/h	70 km/h
Deceleration Lane Length (m)	50 km/h	47 km/h	75 m	70 m	60 m	45 m	-	-	-	-
		37 km/h	60 m	50 m	30 m	-	-	-	-	-
	60 km/h	55 km/h	95 m	90 m	80 m	65 m	55 m	-	-	-
		45 km/h	95 m	80 m	65 m	45 m	-	-	-	-
	70 km/h	63 km/h	110 m	105 m	95 m	85 m	70 m	55 m	-	-
		53 km/h	150 m	130 m	110 m	90 m	65 m	-	-	-
	80 km/h	70 km/h	130 m	125 m	115 m	100 m	90 m	80 m	55 m	-
		60 km/h	200 m	180 m	165 m	145 m	115 m	65 m	-	-
Acceleration Lane Length (m)	90 km/h	77 km/h	145 m	140 m	135 m	120 m	110 m	100 m	75 m	60 m
		67 km/h	260 m	245 m	225 m	205 m	175 m	125 m	35 m	-
	100 km/h	85 km/h	170 m	165 m	155 m	145 m	135 m	120 m	100 m	85 m
		74 km/h	345 m	325 m	305 m	285 m	255 m	205 m	110 m	40 m
	110 km/h	91 km/h	180 m	180 m	170 m	160 m	150 m	140 m	120 m	105 m
		81 km/h	430 m	410 m	390 m	370 m	340 m	290 m	200 m	125 m
	120 km/h	98 km/h	200 m	195 m	185 m	175 m	170 m	155 m	140 m	120 m
		88 km/h	545 m	530 m	515 m	490 m	460 m	410 m	325 m	245 m
	130 km/h	103 km/h	215 m	210 m	205 m	195 m	185 m	170 m	155 m	135 m
		92 km/h	610 m	580 m	550 m	530 m	520 m	500 m	375 m	300 m

. The method of calculation is basically the same as that of Korea.

2.2. Current Regulations about Corner Clearance Distance from Intersection

The South Korean Road Access Management Rules restrict road access from an at-grade intersection within the corner clearance distance [8]. The access-restricted area and corner clearance distance are shown in Figure 1 and

Table 4. Road access restricted distance in South Korean act.

Design Speed (km/h)	Road Access Restricted Distance (m)
	Urban	Rural
50	25	40
60	40	60
70	60	85
80	70	100

. The corner clearance distance in the approaching area is settled from the stop line, and the departure area is settled from the outbound of the geometrical area of the intersection. The corner clearance distance along road design speed differs depending on urban or rural areas (Table 4).

In the US, the road access restricted distance is called corner clearance. The concept of corner clearance is the spacing gap between an intersection and the nearest access connection, as in Figure 2 [9,10]. The clearance type A and C are deceleration areas, and type B and D are acceleration areas. The upstream clearance distance, such as type A, should be longer than the sum of reaction time distance, deceleration distance and queue distance. The downstream clearance distance, such as type B, includes the stopping sight distance [10].

2.3. Acceleration and Deceleration Behavior of Vehicles

There are various studies related to vehicle acceleration and deceleration behavior [5,11,12,13,14,15,16,17,18,19,20,21,22,23,24]. A recent study [5] found that the acceleration rate gradually increases from lowest to maximum depending on the initial speed of the vehicle. This phenomenon is observed in all vehicle types except trucks. The maximum acceleration rate of each vehicle types were 1.0 m/s² for truck, 0.64 m/s² for motorized three-wheeler, 1.96 m/s² for motorized two-wheeler, 2.23 m/s² for diesel car, and 2.87 m/s² for gasoline car, respectively. Wang et al. [23] show a similar acceleration rate to Bokare and Maurya [5], and the other studies [12,15] report lower values. Bokare and Maurya [5] propose a negative exponential acceleration model for trucks, diesel cars and gasoline cars. For the deceleration maneuver process, maximum deceleration rates differ depending on vehicle types such as trucks and passenger cars. Bokare and Maurya [5] proposed the polynomial deceleration model. However, the constant acceleration or deceleration model is the simplest to adapt to the road design [22].

The acceleration and deceleration behavior at signalized intersections depends on various factors such as driver tendency, vehicle type, geometric design, etc. [18]. There are limited studies related to signalized intersections due to the complex experimental variables that influence the acceleration or deceleration [19,25,26]. Mondal and Gupta [19] found that a small motorized vehicle achieves maximum acceleration rate at low speed and highly decelerates during a stopping process at the intersection. It is observed that the average acceleration and maximum deceleration of vehicles vary from 0.9 to 1.40 m/s² and 2.21 to 3.01 m/s².

3. Data Collection and Screening

3.1. DTG

Korean commercial vehicles, such as trucks, taxis and buses, should install DTGs according to the Traffic Safety Act. A DTG records data such as vehicle type, speed, longitude, latitude, heading, acceleration, engine revolutions and break signals. The DTG data columns are shown in

Table 5. The DTG data column.

No.	Description	Type	Unit
1	Vehicle ID	String	-
2	Vehicle Type Code	String	Bus/Truck/Taxi/Etc.
3	Speed	Integer	km/h
4	Engine Revolution	Integer	RPM
5	Break Signal	Binary	On (1)/Off (0)
6	GPS Longitude	Integer	WGS84 (00.000000)
7	GPS Latitude	Integer	WGS84 (000.000000)
8	Heading	Integer	Degree
9	Longitude Acceleration	Integer	m/s2 (1 decimal place)
10	Latitude Acceleration	Integer	m/s2 (1 decimal place)
11	Date	Integer	YYMMDD hhmmssss (0.01 s)

. The driver’s personal information, such as gender, is protected by the Personal Information Protection Act. Thus, researchers can limitedly access the anonymous vehicle ID. Furthermore, since the DTG is installed only in commercial vehicles, it is impossible to observe the non-commercial driver’s behavior. Despite these limitations of DTG raw data, it is valuable raw data for empirically deducing the driving patterns according to vehicle type. This study collected 24-h DTG data for 1 month in April 2018.

3.2. Study Sites

To observe drivers’ acceleration and deceleration behavior, 77 study sites were selected inside the managing area of Korea’s Regional Construction and Management Administration (Figure 3). Those study sites are 1–3 flat grade lanes in one direction, and the speed limits are 50–80 km/h. The traffic conditions vary for a 24-h 1 month. Thus, it is not described in this section, but it is generalized by a large population of data size. The 77 study sites are signalized intersections that have 27 approach areas (in deceleration) and 50 departure areas (in acceleration) (Figure 4). The 27 approach areas comprise 7 urban and 20 rural areas, whereas the 50 departure areas comprise 18 urban and 32 rural areas. The latitude and longitude of those study sites are presented in

Table 6. Latitude and longitude of study sites.

Approaching						Departure
City			Rural			City			Rural			Rural
#	Latitude	Longitude	#	Latitude	Longitude	#	Latitude	Longitude	#	Latitude	Longitude	#	Latitude	Longitude
1	37.72394	127.0976	1	37.18925	127.5607	1	37.88737	127.1944	1	37.14497	127.5859	21	37.29773	127.2334
2	37.71236	127.1063	2	37.14929	127.5792	2	37.74994	127.0846	2	38.09039	127.08	22	37.11436	127.428
3	37.64546	127.1287	3	36.96333	127.8685	3	37.64594	127.1285	3	38.07761	127.2707	23	38.03541	127.2626
4	37.75593	127.0812	4	36.91914	127.9448	4	37.73332	126.7306	4	37.98989	127.2425	24	37.98027	127.2444
5	37.63339	126.8251	5	38.08305	127.2653	5	37.73571	126.7379	5	37.90179	127.2093	25	37.12298	126.9041
6	37.44244	126.7788	6	38.07767	127.2711	6	37.73288	126.7306	6	37.84019	127.1561	26	37.03031	127.9245
7	37.03973	127.0685	7	38.07764	127.2707	7	37.72938	126.7237	7	37.83776	127.2877	27	36.93195	128.2054
			8	37.97938	127.2444	8	37.72857	126.7291	8	38.01607	127.1509	28	36.85829	127.9231
			9	37.94273	127.2329	9	36.9791	126.931	9	38.02147	127.1833	29	36.83129	127.77
			10	37.90797	127.2127	10	37.35839	126.7312	10	38.04843	127.3675	30	36.9852	126.9345
			11	37.17683	126.9398	11	37.35166	126.7283	11	38.02462	127.3628	31	37.77306	127.4562
			12	37.83117	127.2911	12	37.3145	126.8521	12	37.97296	127.3216	32	37.04245	127.3603
			13	38.01835	127.1597	13	37.271	127.0628	13	37.71588	127.4075	33	37.55244	126.675
			14	37.90043	127.2083	14	37.32441	126.8045	14	37.01141	126.9937	34	37.55195	126.6798
			15	37.76978	127.4483	15	37.40899	126.8238	15	37.0232	127.3295	35	37.48959	126.6473
			16	37.04685	127.3636	16	37.01965	127.0716	16	37.34237	127.4901	36	37.49089	126.648
			17	37.44091	127.2634	17	37.03354	127.0701	17	36.959	127.0651	37	37.17142	127.3566
			18	37.11088	127.4305	18	37.06642	127.0641	18	36.96051	127.0661	38	36.3062	127.5749
			19	37.22813	127.3023				19	36.96691	127.0712	39	36.2925	127.5665
			20	37.08091	127.4366				20	36.96839	127.0719	40	36.16964	127.7732
												41	37.0067	127.0782
												42	37.03691	127.0716

.

3.3. Data Screening Algorithm

In the signalized intersection, vehicles approach with deceleration and stop during the red signal. After changing to the green light, they accelerate to the desired speed. Figure 5 shows vehicle movements time–distance and speed–time diagrams at a signalized intersection [5]. To settle a target by analyzing the DTG data, study sites were modeled with closed polygon areas. DTG data, which had GPS coordinates inside the closed polygon area, were extracted. The closed polygons were drawn widely along some core waypoints to collect DTG data without omission of the acceleration or deceleration process. The types of core waypoints include stop lines, taper start/end points, centerline and a maximum 450 m length. The maximum length of the closed polygon follows the influence area of the merging or diverging area in TRB [27]. The waypoints of closed polygons are set as shown in Figure 6a, and data extracted from there are shown in Figure 6b.

After extracting DTG data from the study sites, the analysis data should be screened (Figure 7). The data screening threshold values are set large enough in order to mitigate the analyst’s subjectivity and just filter the outliers. First, data from the approach areas were grouped by identical vehicle IDs and timeslots. Second, only data that showed perfect stops at the intersections were selected. Third, data with a deceleration of higher than 20 m/s² (approximately 2 g, many vehicle manufacturers report warning the drivers when they accelerate or decelerate over 10 m/s²; thus, this threshold is set to roughly double the vehicle manufacturers reporting threshold) were excluded as outliers. Fourth, data with speed values above 200 km/h were excluded. Finally, the decelerations of vehicles at red lights were analyzed, and the 90th percentile value was set as the representative value. Data from the departure areas were processed similarly. First, they were grouped by vehicle IDs and timeslots, and data with an initial vehicle speed of higher than 35 km/h, which are through traffic, were excluded. Then, data with acceleration greater than 20 m/s² (approximately 2 g, many vehicle manufacturers report warning to the drivers when they accelerate or decelerate over 10 m/s²; thus, this threshold is set roughly double than vehicle manufacturers reporting threshold) were excluded as outliers. Finally, the accelerations of vehicles at red lights were analyzed, and the 90th percentile value [28] was set as the representative value.

4. Analysis Result

4.1. Collected Data

A total of 18,542 and 27,019 vehicles from the approach and departure areas, respectively, were identified through the data screening algorithm. Among the vehicles identified in the approach areas, 6968 (37.6%), 4384 (23.6%) and 7190 (38.8%) were buses, taxis and trucks, respectively. Meanwhile, the departure area vehicles comprised 12,594 (46.6%) buses, 5753 (21.3%) taxis and 8672 (32.1%) trucks. The analysis results were categorized into 12 classes (

Table 7. Data analysis cases.

Area	Vehicle Type
	Bus + Taxi + Truck	Bus	Taxi	Truck
Urban + Rural	Case 1	Case 2	Case 3	Case 4
Urban	Case 5	Case 6	Case 7	Case 8
Rural	Case 9	Case 10	Case 11	Case 12

). Because the raw DTG data are collected in 1-km/h units, the acceleration/deceleration error is approximately 0.28 m/s². Non-commercial vehicles, such as passenger cars, are not equipped with DTG. Thus, passenger car behaviors are not represented in these results, but they might be similar to taxis, which have similar mechanical performance. These raw DTG data show deceleration and acceleration processes at signalized intersections. The DTG data were collected in various traffic conditions for 24 h for one month. Thus, 90th percentile values [28] were set as the representative value to filter the variation of observed acceleration/deceleration values. Furthermore, observed values significantly differ from the data in emergency stop-and-go situations.

4.2. Approach Area

As shown in

Table 8. Analysis results of deceleration in approach areas.

Case (Vehicle, Area)	Deceleration Rate (m/s2)								Total	Avg.	Max.	90th mPCTL
	0.28	0.56	0.83	1.11	1.39	1.67	1.94	2.22
Case 1 (All, All)	1820	4392	5580	4191	1919	565	74	1	18,542	0.74	1.97	1.20
Case 2 (Bus, All)	995	2172	2039	1322	375	61	4	0	6968	0.63	1.76	1.04
Case 3 (Taxi, All)	77	372	989	1421	1056	411	57	1	4384	0.99	1.97	1.41
Case 4 (Truck, All)	748	1848	2552	1448	488	93	13	0	7190	0.68	1.88	1.09
Case 5 (All, Urban)	490	1506	2260	1859	1039	352	50	1	7557	0.80	1.97	1.28
Case 6 (Bus, Urban)	321	862	905	447	89	13	0	0	2637	0.62	1.57	0.98
Case 7 (Taxi, Urban)	57	316	837	1192	882	333	48	1	3666	0.98	1.97	1.41
Case 8 (Truck, Urban)	112	328	518	220	68	6	2	0	1254	0.67	1.74	1.03
Case 9 (All, Rural)	1330	2886	3320	2332	880	213	24	0	10,985	0.69	1.89	1.13
Case 10 (Bus, Rural)	674	1310	1134	875	286	48	4	0	4331	0.64	1.76	1.07
Case 11 (Taxi, Rural)	20	56	152	229	174	78	9	0	718	1.00	1.89	1.44
Case 12 (Truck, Rural)	636	1520	2034	1228	420	87	11	0	5936	0.69	1.88	1.09

and Figure 8, the deceleration rate of taxis is higher than those of buses and trucks; they significantly differ under 0.05 significance. The 90th percentile of deceleration rates is 1.04, 1.41 and 1.09 m/s² for buses, taxis and trucks, respectively. According to our T-test results (

Table 9. Deceleration rate T-test results.

	Bus	Taxi	Truck	Urban	Rural
Variance	1.21	1.39	1.21	1.56	1.40
N	6968	4384	7190	7557	10,985
Degree of Freedom	14,156	8757		15,644
F	1.00	1.15		1.12
P (F ≤ f) One-tail	0.48	8.4 × 10−8		6.53 × 10−8
F Critical One-tail	1.04	1.05		1.04
T Statistic	−10.64	49.26		22.71
P (T ≤ t) One-tail	1.23 × 10−26	0		1.3 × 10−112
T Critical One-tail	1.64	1.65		1.64
P (T ≤ t) Two-tail	2.46 × 10−26	0		2.5 × 10−112
T Critical Two-tail	1.96	1.96		1.96

), the deceleration rates of buses and trucks are homoscedastic, whereas that of taxis is heteroscedastic.

All types of vehicle deceleration rates in urban and rural areas are heterogeneous and have different average values under 0.05 significance.

4.3. Departure Area

As shown in

Table 10. Analysis results of acceleration in departure areas.

Case (Vehicle, Area)	Acceleration Rate (m/s2)								Total	Avg.	Max.	90th PCTL
	0.28	0.56	0.83	1.11	1.39	1.67	1.94	2.22
Case 1 (All, All)	431	10,175	11,810	3766	765	67	5	0	27,019	0.64	1.90	0.94
Case 2 (Bus, All)	176	5021	6253	1111	33	0	0	0	12,594	0.61	1.30	0.82
Case 3 (Taxi, All)	30	544	2154	2235	718	67	5	0	5753	0.86	1.90	1.16
Case 4 (Truck, All)	225	4610	3403	420	14	0	0	0	8672	0.55	1.28	0.77
Case 5 (All, Urban)	200	5534	5703	2114	494	41	2	0	14,088	0.65	1.86	0.96
Case 6 (Bus, Urban)	107	3580	3318	405	4	0	0	0	7414	0.58	1.28	0.78
Case 7 (Taxi, Urban)	21	417	1561	1635	490	41	2	0	4167	0.85	1.86	1.15
Case 8 (Truck, Urban)	72	1537	824	74	0	0	0	0	2507	0.52	1.11	0.71
Case 9 (All, Rural)	231	4649	6098	1651	273	26	3	0	12,931	0.64	1.90	0.91
Case 10 (Bus, Rural)	69	1449	2924	709	29	0	0	0	5180	0.65	1.30	0.88
Case 11 (Taxi, Rural)	9	127	595	596	230	26	3	0	1586	0.88	1.90	1.18
Case 12 (Truck, Rural)	153	3073	2579	346	14	0	0	0	6165	0.56	1.28	0.78

and Figure 9, the acceleration rate of taxis is higher than those of buses and trucks; they significantly differ under 0.05 significance. The 90th percentile of acceleration rates is 0.82, 1.16 and 0.77 m/s² for buses, taxis and trucks, respectively. According to our T-test results (

Table 11. Acceleration rate T-test results.

	Bus	Taxi	Truck	Urban	Rural
Variance	0.35	0.71	0.34	0.64	0.54
N	12,594	5753	8672	14,088	12,931
Degree of Freedom		8428	18,864	27,017
F		0.49	1.04	1.20
P (F ≤ f) One-tail		0	0.03	1.82 × 10−25
F Critical One-tail		0.96	1.03	1.029
T Statistic		−73.90	24.28	3.98
P (T ≤ t) One-tail		0	1.6 × 10−128	3.52 × 10−5
T Critical One-tail		1.65	1.64	1.64
P (T ≤ t) Two-tail		0	3.3 × 10−128	7.03 × 10−5
T Critical Two-tail		1.96	1.96	1.96

), the acceleration rates of buses, taxis and trucks are heteroscedastic.

All types of vehicle acceleration rates in urban and rural areas are heterogeneous and have different average values under 0.05 significance.

5. Discussion

The 90th percentiles of acceleration and deceleration values are selected as representative values [28]. The empirically observed acceleration and deceleration values are shown in Table 8 and Table 10. In all cases, the acceleration and deceleration rates are heterogeneous between the urban and rural areas. Reflecting the heavy vehicle turbulence, the lowest acceleration and deceleration rates between heterogeneous vehicle types are selected as design values. The design values of acceleration for trucks are 0.71 and 0.78 m/s² in urban and rural areas, respectively. The design values of deceleration for buses are 0.98 and 1.07 m/s² in urban and rural areas, respectively. The corner clearance distances of a signalized intersection reflecting an empirically observed value are calculated from Equation (2). The suggested corner clearance distances of a signalized intersection, derived from the empirical analysis results of drivers’ behaviors, are shown in

Table 12. Suggestion of corner clearance distance of signalized intersection reflecting DTG analysis results.

Type			Design Speed (km/h)	80	70	60	50
Approaching Area (m)	Suggestion	Urban	a = 0.98 m/s2	193	156	119	87
		Rural	a = 1.07 m/s2	177	143	109	80
	MOLIT [7]			96	78	60	43
	AASHTO [1]			130	110	95	75
Departure Area (m)	Suggestion	Urban	a = 0.71 m/s2	196	153	110	74
		Rural	a = 0.78 m/s2	178	139	100	68
	MOLIT [7]			None
	AASHTO [1]			200	150	95	60

This comparison with MOLIT [7] and AASHTO [1] is based on stop conditions. MOLIT [7] does not provide the minimum length under stop conditions.

.

6. Conclusions

Road management authorities, such as AASHTO and MOLIT, have their own reference acceleration and deceleration rate values. However, vehicles’ driving performance has significantly improved along with technology development since these reference values were set. The concept of comfortable driving behavior has also changed over time. In this study, we analyzed the acceleration and deceleration behaviors with empirical DTG data from commercial vehicles and adapted them to the corner clearance distance from/to the signalized intersection. The major limitation of this study is that the non-commercial passenger cars’ behavior cannot be observed from DTG data. Furthermore, despite the importance of a driver’s personal information factors, such as driving experience, gender, etc., it was excluded from the raw data column for the respect for anonymity of the Personal Information Protection Act. However, these empirical study results are still valuable because this study deduced the representative acceleration and deceleration values by monitoring multiple vehicle maneuvers from 24-h DTG data for 1 month in multiple study sites.

To improve traffic safety, access management objectives should be updated following the suggested corner clearance distance. The latest acceleration and deceleration behaviors introduced in this study will also contribute to road design guidelines’ revisions, such as auxiliary lanes. When updating the current road design guidelines, additional data collection and analysis would be required to comprehensively reflect gender, age and non-commercial drivers’ behavior to overcome the limitation of this study. The already existing road infrastructures are difficult to immediately reconstruct following revised regulations. However, it is necessary to gradually upgrade those geometric architectures step by step. For example, during a regular maintenance work period, road management authorities should change the road design by reflecting recent regulations for road users’ safety and traffic operational efficiency.

In the close future, driving behaviors are expected to change significantly with the introduction of autonomous vehicles and electric vehicles. For the road infrastructure guidance services to advise the autonomous vehicle’s safe and harmonized driving under mixed traffic conditions with conventional vehicles, it is crucial to predict the conventional vehicles driving path via V2V or V2I cooperation. In this aspect, empirical analysis of acceleration and deceleration rate of conventional buses, trucks and taxis in this study can significantly improve the accuracy of driving path prediction for moderate driving guidance such as the right of way recommendation. This study has some limitations that cannot show the non-commercial drivers’ behavior. Furthermore, this study only focuses on the vehicles that stopped at signalized intersections at red lights and analyze the 90th percentile acceleration and deceleration rate. Thus, various intersection designs and road speed limit variations are not considered in detail. For the more precise prediction of conventional vehicle movements and to accomplish the advanced road design and accompany the driving service, further empirical research using roadside unit detection data needs to be followed in the near future.