ADA Systems Application for Traffic Safety Improvement on Roadways
Hassan ZOGHI
Assistant Professor, Civil Engineering Department, Islamic Azad University, Karaj branch
Tehran, Iran
H_zoghi@kiau.ac.ir
Kianoush SIAMARDI
Civil Engineer and member of Young Rearchers Club Islamic Azad University, Karaj branch
Tehran, Iran
k.
Morteza TOLOUEI
Civil Engineer and member of Young Rearchers Club Islamic Azad University, Karaj branch
Tehran, Iran
Abstract— Com pared to the m otorway network, rural and urban roads are very unsafe. Advanced Driver Assistance System s can be ud to increa traffic safety, thereby im proving the conditions for effective deploym ent of the underlying network for integrated traffic management. In this study, we have identified ADA system s that are expected to have a high impact on traffic safety by coupling characteristics of accidents to the functionality of different ADA system s. Bad on the results of this study, Intelligent Speed Adaptation (ISA) and Interction Crash Avoidance system s appeared m ost likely to substantially im prove the safety on non-motorway roads. ISA was lected for further analysis due to the likelihood of its near future introduction. Microscopic sim ulation m odel was ud to quantify the im pacts on traffic safety for an urban highway with controlled interctions near the south of Tehran m etropolitan. Depending on the penetration level, the results of the sim ulation study showed that the decrea in the total number of accidents ranges from 4% to 19%.
Keywords- ADA System, ITS, road networks, motorways, accident.
I.I NTRODUCTION
Compared to the economical and societal cost of traffic congestion, the cost of traffic unsafety is very high. For example, in 2008 the cost for society of traffic congestion in the Iran was 280 milliard dollars; the costs due to accidents was much higher, namely 890 milliard dollars. Since most accidents do not occur on motorways but on the underlying traffic network, it ems logical to consider approaches to improve the traffic safety on urban and rural roads [1,2].
Although various alternatives can be chon to improve traffic safety, this study focus on the potential effects of Advanced Driver Assistance Systems (ADAS) on traffic safety. Many studies have focud on the impacts of ADAS on the motorway traffic network. Surprisingly, our analysis revealed that many of the accidents occur on urban and rural highways. Little is known about the ADAS impacts on the road types, justifying the rearch reported in this paper here. The approach taken in this paper is two-fold. At first, the safety potential of ADA is assd by a detailed analysis of accident reports, focusing both on the so-called conflict maneuver level [3] and the accident cau level [4].
Both maneuvers and caus are then linked to the functionalities of the identified ADA systems, and combined afterwards to provide an overall ADA safety potential figure.
A simulation study is subquently performed to provide a more detailed analysis of the ADA system that turns out to have a high safety potential, namely Intelligent Speed Adaptation (ISA). The main rearch objective of this paper is “to asss the impact of ADAS on the safety on urban and rural highways for different types of ADAS”.
First pha of this paper entails making an inventory of unsafe situations, maneuvers and conflicts (safety analysis), as well as an overview of the different ADA systems. In the cond pha, the results of the safety analysis are combined with the ADA system overview with the aim to provide a preliminary asssment of the potential ADA safety impacts. In this respect, safety potential is defined by the maximum attainable reduction in accident frequency under ideal circumstances, e.g. 100% ADAS penetration, faultless operation and u of ADAS, etc. The cond pha will provide one or more ADA systems with a high safety potential for the urban and rural networks. In the third pha, the considered system(s) will be analyzed in more detail by a microscopic simulation study, in terms of safety impacts as well as impacts of network efficiency. The last pha entails a synthesis of the previous steps.
II.A CCIDENT A NALYSIS
In general, it was obrved that more that half of all accidents occur on rural and urban highways. The absolute number of accidents in rural areas is much less than the number of accidents in urban areas. The accident risk, defined by the number of accidents per traveled km, is nevertheless higher on rural roads than on urban roads. The accident risks on both the rural and the urban highways are in turn much higher than the accident risk on motorways. It
was furthermore obrved that in rural areas, most accidents occur on roadway ctions, whereas in urbanized areas, most accidents occur on interctions. Of all 41270 (reported) non-lethal accidents, respectively 20930 and 20340 occurred on ctions and interctions, which is again in line with the national figures. Figure 1 shows in more detail the shares with respect to the total number of accidents per roadway type. It shows that most accidents occur on urban highways (33% of all lethal and 45% of all non-lethal accidents).
0%
10%20%30%40%50%Motorway
Rural highways Urban highways Suburban streets
Urban streets
Road network type
p e r c e n t a g e
Figure 1. Accident fraction for different roadway types
III. C LASIFICATION O F A DA S YSTEMS
It is expected that in the near future, systems will become available that will support different tasks of the driver. ADA (Advanced Driver Assistance) systems are a subt of AVG (Automated Vehicle Guidance) systems. The support that a driver will receive can have various forms. It can either be additional visual or auditory information collected via dedicated vehicle nsors, but also more active
support is possible. Examples of ADA systems are Intelligent Speed Adaptation (ISA), Adaptive Intelligent Crui Control (AICC), Lane Keeping Systems (LKS), Collision Avoidance Systems (CAS), etc. The system primarily affect the operational driving task (obrve, decide, control), and to a lesr extent the tactical and strategic levels (position on the road, navigation, etc.). It is expected that ADA will have a considerable effect of traffic safety. In illustration, in Japan a reduction in accidents of 80 percent is expected. With the introduction of AICC, the number of head-to-tail accidents can be reduced with 60%.
To ass the effects of ADA on traffic safety, the results of this analysis can be ud as a reference under the condition that the knowledge regarding the driver’s u of the considered ADA is sufficient. Also note that in the prent study, we address the maximal potential of ADA systems to increa safety. The categorization of ADA systems is chon to suit this goal, that is, to reveal which driver subtasks the ADA system effectively supports. The categorization propod in ADVISORS (2002) [5] is ud, since it suits this purpo adequately. Table 1 shows this overview, distinguishing between pre-crash, crash and post-crash support. Our rearch will focus primarily on pre-crash systems.
TABLE I.
C ATEGORIZATION OF ADA SYSTEMS , (PRE -CRASH , CRASH ,
AND POST -CRASH SYSTEMS )
PRE CRASH
7. Vehicle status monitoring • Tachograph (data recorder) • Vehicle diagnostics 1. Navigation bad function
• Enhanced navigation; navigation routing
• Real time traffic and traveler information;
distributed navigation 2. Longitudinal control • Speed control; Curve speed control • Advanced Crui Control • Stop-and-go 8. Perception
• Vision enhancement; night vision • Electronic mirror • Blind Spot detection
• Reversing Aid / Parking aid • State of the road surface ; Low friction warning
3. Lateral control • Road departure; Lane departure collision avoidance • Lane change and merge collision avoidance 9. Man-Machine Communication
• Driver convenience communication • On-board hands-free functions • Driver identification and automatic cockpit
configuration
• Automated transactions; Electronic toll collection
4. Collision avoidance • Rear-end collision avoidance; Pre-crash nsing
• Obstacle detection; Pedestrian detection
• Interction collision avoidance
• Rail-road crossing collision avoidance
CRASH • Smart Restraints 5. General control (longitudinal & lateral) • Platoon driving assistant • Overtake checker POST CRASH • Alerting systems • Automatic collision notification • Specific phone call number 6. Driver monitoring • Driver vigilance monitoring • Driver health monitoring
A. Accident cau analysis
Bad on occurred accidents that was reported to police, In total, 97 different caus have been identified, which are categorized into 6 different groups (A-F). In illustration, group A is subdivided into 11 subgroups. Subgroup 11 (keeping too little distance) reflects one of the most important accident caus. Note that the caus are indicated on the police reports ud for the prented rearch, and are thus bad on causality as subjectively interpreted by the police. Considering the caus for road ction accidents, we can conclude that the most common cau is keeping too little distance. While in rural areas, most accidents occur on roadway ctions, in urban areas, most accidents occur on interctions (22% of all lethal and 28% of all non-lethal accidents). Table 2 only shows the caus most frequently obrved.
TABLE II.
O VERVIEW OF MAIN ACCIDENT CAUSES FOR ROADWAY
SECTIONS AND INTERSECTIONS
Accident on roadway ctions
Accidents on interctions
A. Cau due to driver.(Distance keeping)
11 keeping too little distance Overtaking
18 overtake on the left
19 cutting of pasd vehicle Lateral location on the road 25 drive too much to the left 26 drive too much to the right Providing right-of-way
30 did not grant right-of-way 31 did not provide passageway Other caus
39 steering error in corner A. Cau due to driver(Traffic signs and signals)
01 neglect stopping sign / traffic light
Distance keeping 11 too little distance Providing right-of-way
30 did not grant right-of-way 31 did not provide passageway Speed-related caus 35 Speed too high
D. Road conditions 65 snow, black ice, etc.
66 road covered with oil, leaves, etc.
F. Other circumstances
77 loo control over steering wheel
F. Other circumstances
77 loo control over steering wheel
IV. S AFETY I MPACT A NALYSIS ON A C ONFLICT L EVEL Along the same line of thought, the safety potential of ADA can be assd by considering the maneuvers preceding the accidents. In total, 89 different maneuvers have been defined. In the police reports ud for this analysis, each maneuver is described by a number of keywords and a schematic overview of the situation at hand. According to guidelines, the maneuvers are categorized into conflict groups. In the rearch prented here, six conflict types have been identified. Table 3 gives an overview and a short description of the different conflict types that have been identified. For roadway ctions, the data shows that on rural highways, most accidents are a conquence of unidirectional conflicts (33%), followed by accidents caud by frontal conflicts and conflicts with obstacles. In the latter ca, it turns out that the maneuvers ‘ending up against a tree, fence, hou or streetlight’ are most important. On urban highways, we again e that unidirectional conflicts are most prominently causing accidents (31%). On the cond and third place we find respectively conflicts with obstacles (21%) and perpendicular conflicts (20%). In the latter category, maneuvers due to crossing pedestrians and crossing slow vehicles are most common (8% and 4%
respectively). For interctions on rural and urban highways,
most accidents can be contributed to perpendicular conflicts
(42% for rural highways and 53% for urban highways),
followed by accidents resulting from uni-directional conflicts.
TABLE III.
C ONFLICT TYPES AN
D ITS DESCRIPTION
V. ADA S AFETY P OTENTIAL A SSESMENT
This ction discuss the general results of ADA safety
asssment by linking the previous accident analysis results on the accident cau level and the conflict/maneuver level to the functionalities of the considered ADA systems. We furthermore address the impacts of ISA by an alternative approach, since it turned out that the formed analysis was not applicable to ISA. Finally, the results are synthesized,
yielding an overall estimate of the potential safety impacts of ADAS. To asss the potential safety i
mpacts, for all ADA systems the functionality of the system was considered cloly and coupled to the accident caus it was likely to prevent. Table 4 shows the Results of safety impact analysis by coupling ADA systems to accident caus for both rural and urban highways. The application results of this method must however be carefully considered. For one, the potential effect of navigation cannot be assd using this approach. Monitoring systems will have a maximum effect of 1 percent, while the potential effect of perception is between 3 and 7 percent. For longitudinal control, lateral control and general control systems, the expected effects are high (between 20% and 24% reduction in the number of accidents). Collision avoidance systems have an even higher safety potential. TABLE IV. R ESULTS SAFETY IMPACT ANALYSIS BY COUPLING ADA SYSTEMS TO ACCIDENT CAUSES FOR BOTH RURAL AND URBAN HIGHWAYS
ADA system x accident cau Main cau Potential safety impact Road ction Inter
ction
1. avigation
bad function 03 one-way road in wrong direction
04 driving opposite to traffic
0% 0% Driving monitoring
1% 0% Driver vigilance monitoring
80 sleep / illness 1% 0% Driver health monitoring 79 alcohol, medication, drug u,
80 sleep / illness
1% 0%
2. Vehicle status monitoring 40 incorrect lights, no lights, 70 mechanical problem
1% 0% 3. Longitudinal control
24% 13%
Speed control (ISA)
35 driving too fast 1% 1%
Curve speed control 35 driving too fast, 39 taking the turn incorrecty 6% 3% Advanced Crui Control 11 not maintaining sufficient distance
17% 9%
Stop and Go 11 not maintaining sufficient distance
12 unexpected need to
brake 18% 10%
4. Lateral control. 20% 5% Road departure / Lane departure CA 25 driving too much to the left 26 driving too muh to
the right
12% 3%
Lane change and merge collision avoidance 08 wrong merge (converge), 09 wrong merge (diverge),
17 overtake on the right,
18 overtake of the left, 19 cutting off vehicle
8% 2% 5. General control 24% 12%
Platooning driving assistant 11 not maintaining sufficient distance
12 unexpected / sudden need to brake
18% 10% Overtake checker 18 overtake of the left, 19 cutting off vehicle 6% 2% 6. Collision avoidance
6% 58% Rear-end CA / Pre-crash nsing
~ ~
Obstacle detection / Pedestrian detection 23 being on the road, 51 careless crossing,
52 playing on roadway, 53 careless walking on
roadway,
84 obstacle on
roadway
6% 2%
Interction collision avoidance 30 not providing right-of-way, 31 not providing passageway
~ 56% 8. Perception 4% 2%
Vision enhancement / night vision 22 parking without lights
81 too little street
lighting 0% 0%
Blind Spot detection
68 unclear corner 0% 0% Reversing Aid / Parking aid 21 incorrect parking maneuver
0% 0%
State of the road surface / Low friction warning 65 snow, black ice, frost on road,
66 oil, leaves, etc. on roadway. 3% 2%
(percentages shows fraction of total number of accidents on road ction or interction)
Similar to the analysis on the level of accident caus, we have analyzed the potential impact of ADA a similar analysis
has been applied to the accident analysis on the maneuver level, i.e. which maneuvers would have been prevented if a
certain ADA system would have been available. To clarify the results, let us briefly discuss the main results of the analysis: 1. Longitudinal control • Speed control (ISA) applies to each maneuver that
occurs when the speed is too high, but is does not necessarily
prevent a maneuver. In the remainder, an alternative approach is considered that quantifies the effec
ts of ISA. • Curve speed control can be coupled to the maneuver ‘vehicle off the road in/after curve’; the appear to be only a minor share of all accidents
• Advanced Crui Control (ACC) applies to longitudinal
conflicts and can be coupled to veral maneuvers within this
category. As such, ACC applies to 16% of all accidents on roadway ctions and 8% of all accidents on interctions. • Stop-and-Go cannot be coupled to maneuvers. This system has been developed for u in congestion and is mainly focud at reducing accidents at low speeds (only damage to the car). 2. Lateral control
• Road departure collision avoidance / Lane departure Warning Assistant (LDWA) works for a number of conflicts and can be coupled to veral maneuvers. In total, 3% of all road ction accidents can be prevented in theory.
• Lane change and merge collision avoidance apply to conflicts in which converging / diverging traffic streams are relevant, and applies to 7% of all road ction maneuvers and
1% of all interction maneuvers. 3. General control • Platooning driving assistant can be linked to
20% of all maneuvers preceding accidents on roadway ctions and to 10% of all maneuvers preceding accidents on interctions. • Overtaking checkers apply to the all kinds over overtaking maneuvers, which only reflect 1% of all maneuvers preceding accident on roadway ctions.
4. Collision avoidance • Rear-end collision avoidance applies to longitudinal conflicts and can be coupled to veral maneuvers (left- or right-turning vehicle is hit from behind by vehicle). This applies to 16% of all accidents on roadway ctions and 8% of all accidents on interctions. • Obstacle detection / pedestrian detection will ideally
detect all obstructions and can as such be coupled to veral
maneuvers from the group of longitudinal conflicts as well as lateral conflicts. It applies to 21% of all road ction accidents and 11% of all interction accidents. • Interction collision avoidance (I-CA) describes a
system in which there is communication between the interction detectors and the vehicle. In theory, this can
prevent a number of longitudinal conflicts (but not all, e.g. pedestrians suddenly crossing the street).
We have coupled the functionality to a number of maneuvers and it turned out that 52% of all accidents on interctions can be avoided.
5. Perception
• State of the road surface / Low friction warning can prevent one-sided conflicts and applies respectively 7% and 3% of all accidents on ctions and interctions.
The impacts of speed limiting measures such as ISA cannot be coupled adequately to either conflicts or accident caus. As a result, it appears to be difficult to correctly asss the effects of ISA on the number of accidents. To asss the safety impacts anyway, Oei [6] propos an alternative approach that is bad on estimating the global reduction in the average driving speed and subquently using this relative decrea to asss the impacts on safety using the approach of Nilsson [7].
For the different analysis prented in the previous ctions, we may conclude that the systems with the largest safety potential on motorways are ACC, Platooning and Rear-end CA. For rural and urban highways, we may conclude that the safety potential of ADAS is much different and that the different analys do not provide completely consistent results. By taking the average outcomes of b
oth approaches, an overall indication of ADAS safety potential is shown in Figure 2 (per ADA category) and Figure 3. The latter shows the combined results of the different analys prented in the preceding ctions, where the results for ctions and interctions as well as for the two kinds of analys have been combined for lected ADAS.
Potential effect ADAS(maneuvers & caus)
ADAS- category
b e
Figure 2. ADAS safety potential per ADA category
Total potential effect of ADA on rural and
urban highw ays
0%
2%4%6%8%10%12%14%16%ISA
ACC
Platooning
Rear-end
Pedestrian
I-CA
ADA sy stem
P e r c e n t a g e o f a c c i d e n t s t h a t c a n
b e p r e v e n t e d b y A D A s y s t e m
Figure 3. ADA safety analysis results for ADA systems with highest safety potential
Let us finally mention that on motorways it turns out that the ADA systems that all systems that somehow reduce the amount of head-to-tail collisions (such as ACC, Platooning and Rear-End CA) score well. To a lesr extent this holds for urban and rural highways. However, in the latter ca, other systems also perform well, such as Obstacles / Pedestrian detection, ISA and Interction CA. The latter appear to have an especially high potential: from our analysis, it turns out that after implementation of ISA and Interction CA, respectively 18% and 16% of all accidents may be prevented. Since ISA is more likely to be implemented in the near future, it is considered as the ADA system to be considered for further analysis.
To gain better insights into expected effects of ISA and to come up with hypothesis concerning its effects, an impact pre-asssment was performed. In a comprehensive analysis, we could distinguish direct safety benefits (such as enhanced driving performance and mitigation of crash conquences) and indirect safety benefits (e.g. reduced exposure, reduced driver stress and fatigue, reduced conflicts and variance in behavior). Also, direct safety risks (driver distraction, overlo
ad, reduced situation awareness) and indirect safety risks (behavioral adaptation, loss of skill, etc.) can be distinguished [8-13].
VI. I MPACT A SSESMENT B Y M ICRO -S IMULATION This ction discuss the predicted impacts of ISA by micro-simulation. The considered situation pertains to a prototypical urban highway with controlled interctions. The study area is an urban highway in southern Tehran, known as the “Besat highway”. It is about 3 km in length, connecting the motorway from the west of highway with the motorway to the east of highway. The highway has four controlled interctions, indicated by red circles in Figure 4, which were modeled in VISSIM. Impacts on traffic efficiency and safety were quantified for a particular test ca pertaining to a controlled urban arterial. The results of the simulation study showed that the decrea in the total number of accidents ranges from 4% to 19%.
For application of the model, data collected from the traffic controllers have been ud to determine traffic demands during the simulated periods. Furthermore, the following traffic class have been considered: ISA equipped person-cars, regular person-cars, trucks, and buss. Their shares have been determined from available traffic data, as have the parameters describing their behavior (vehicle length, free speeds, etc.). The parameters that could not be established directly from the me
asurements were changed during the calibration pha such that the model predictions match available measurements as cloly as possible. More details on the calibration process will be discusd in the final paper. N ote that different scenario’s have been considered (different ISA penetration levels).
A. Impacts on flow characteristics and safety
Bad on micro simulation results, ISA reduces average speed. We could not establish statistically significant changes in the speed variance. According to the method of
