Advanced OR and AI Methods in Transportation OBSERVATIONS OVERTAKING MANOEUVRES ON BI-DIRECTIONAL ROADS Geertje HEGEMAN*,1, Serge HOOGENDOORN1, Karel BROOKHUIS2 Abstract. Observations of overtaking manoeuvres on two-lane rural roads were carried out to enhance the understanding of driver behaviour prior to, during and after an overtaking manoeuvre. An instrumented vehicle was driven with different speeds while other vehicles’ overtaking manoeuvres were recorded and analysed afterwards. The differences in duration of overtaking manoeuvres between different overtaking strategies and different speeds of the vehicle that was overtaken, turned out to be small. Fairly short perception-reaction times were observed, indicating that the decision to perform an overtaking manoeuvre is made before an appropriate gap in the oncoming traffic stream is available. 1. Introduction Driving consists of a large amount of subtasks [1], among which overtaking. Overtaking is considered to be a hazardous subtask: transportation experts estimate that lane change crashes, including overtaking and merging, account for 4 to 10% of all crashes [2, 3]. In the Netherlands around 26 traffic participants are killed each year because of overtaking failures, i.e. about 3% of all fatalities [4]. Overtaking is particularly difficult on two-lane rural roads, with oncoming traffic and relatively high speeds. On the Dutch ‘flow-roads’ passenger cars’ speed limit is 100 km/h while trucks, buses and vehicles with trailers face a limit of 80 km/h, which provokes overtake demand [5]. Possible solutions to make overtaking manoeuvres safer are overtaking lanes and intelligent systems in the vehicles that can assist drivers with the overtaking manoeuvre. To be able to design or develop these solutions, accurate overtaking manoeuvre data are required. These data are also useful for the development of traffic micro simulations models that include many assumptions to model overtaking manoeuvres [6]. To this end, an observation study on a two-lane rural road, with an instrumented vehicle is carried out. This study aims at acquiring qualitative empirical insights into overtaking behaviour on two-lane rural roads. * Faculties of Civil Engineering and Geosciences1 (Transport and Planning section), Technology, Policy and Management2, Delft University of Technology, P.O. Box 5048, 2600 GA Delft, The Netherlands, g.hegeman@ct.tudelft.nl 506 G. Hegeman et al. 2. Useful overtaking manoeuvre literature findings Existing literature is studied to find out what is already known about overtaking on bidirectional roads. The information found is used for the design of the present observation study. Already in 1963 a test track study on overtaking behaviour was performed [7]. The average duration of an overtaking manoeuvre with opposing traffic was found to be 6.7s. With another observation study along a road [8] the undesired overtaking strategies ‘lane sharing’, ‘cutting in’ and ‘braking to follow’ were observed. Three other strategies were defined: ‘Normal’, ‘Flying overtaking’ and ‘piggy backing’. ‘Normal’ or often called ‘accelerative’ overtaking, means that the overtaker follows a vehicle and waits for a sufficient gap to perform an overtaking manoeuvre. As soon as this gap is available, the overtaker accelerates, and performs the manoeuvre. ‘Flying overtaking’ is when the overtaking vehicle does not adjust its speed to the overtaken vehicle but continues its current speed during the overtaking manoeuvre. Finally ‘Piggy backing’ in case the overtaking vehicle follows another vehicle that overtakes a slower vehicle. In a more recent simulator study on the reliability of standards for safe overtaking [6], four important time durations related to an overtaking manoeuvre were defined: Perception-reaction time; The time for occupying the left lane; The time during which the passing car drives on the left lane; The clearance time distance between passing and oncoming vehicle at the end of the overtaking (time to collision). Both the overtaking strategies and the four time durations related to an overtaking manoeuvre are used to structure the analyses of the results of this observation study. 3. Applied Method to study overtaking behaviour Out of the many methods to study overtaking manoeuvres on roads with opposing traffic, i.e. accident analysis, driving simulator, test track observations, instrumented vehicle and observation along the road, using an instrumented vehicle is chosen. An instrumented vehicle provides the opportunity to observe naturalistic overtaking manoeuvres. Especially, if the vehicle is used passively, i.e. the researcher drives the vehicle him/her self, while other, unknown drivers perform overtaking manoeuvres. Speed, acceleration and distance information from all surrounding vehicles, including the overtaker, are recorded with unobtrusive cameras at the front and the rear of the car. Figure 1 shows a schematic picture of an overtaking manoeuvre to be observed. At the right, the instrumented vehicle is shown. 3 1 1 2 1 Figure 1. Schematic picture of a accelerative overtaking manoeuvre The variables that are observed in this overtaking manoeuvre observation study are: • What is the distance prior to and after the overtaking manoeuvre between vehicle 1 and vehicle 2? Observations overtaking manoeuvres on bi-directional roads • • • • 3.1. 507 Do the overtakers use their indicator? What is the duration of the total overtaking manoeuvre? Is this duration dependent on the speed of vehicle 2? How many seconds are left after the overtaking manoeuvre, before the first oncoming vehicle (vehicle 3) passes? (time to collision (TTC) between vehicle 1 and vehicle 3) Data collection overtaking manoeuvres The observations took place on a 5 km long road section of a two lane rural road with a speed limit of 100 km/h (N305, The Netherlands). Two researchers drove this road section up and down, with constant speeds of 70, 80, 90 and 100 km/h. Respectively 13, 24, 11 and zero overtaking manoeuvres were observed From the literature, three overtaking manoeuvre strategies were distinguished (8): accelerative, flying and piggy backing. These three strategies were observed during this observation, plus another strategy called ‘2+’: the overtaker overtakes one or more vehicles behind the overtaken vehicle and the overtaken vehicle itself in the same move. So in that case minimally two vehicles were overtaken. For each overtaking manoeuvre, the important moments, such as the moment the overtaker starts to move to the left lane, are determined. With these, the important time durations related to an overtaking manoeuvre, as defined in the section above, are calculated. The distance headway between vehicle 1 and vehicle 2 at the moment the overtaking manoeuvre starts and at the moment the manoeuvre ends are established too. 4. Results overtaking manoeuvre observation The first action to take when performing an overtaking manoeuvre is to accept an appropriate gap. As soon as the gap is accepted, the manoeuvre starts. The time to do this is called the perception-reaction time, being the time between the last oncoming vehicle has passed the overtaker and the moment the movement to the left starts. The start of the move to the left is defined as the moment the left front wheel of the overtaker touches the centre line. Most observed perception-reaction times in this observation are less than one second, indicating that a driver observes an appropriate gap in the oncoming stream in advance. As soon as the gap arrives, the overtaking manoeuvre can start. The use of the indicator prior to the overtaking manoeuvre is almost twice as much as after the manoeuvre. Nine drivers certainly did not use the indicator at the start of the manoeuvre, which is obliged according to Dutch traffic law. Figure 2 shows the observed distance prior to and after the overtaking manoeuvres. For almost one fifth of the overtaking manoeuvres, the prior distance is less than 10 meters. For some manoeuvres, the overtaker not even crossed the centre line totally at the moment the front of the vehicle is next to the overtaken vehicle (lane sharing). The shortest measured distance is 7.7 meters (accelerative strategy, speed of overtaken vehicle: 80 km/h), corresponding with 0.35s. Although none of the manoeuvres led to an accident, such short headways seem rather dangerous for two vehicles driving at more then 70 km/h. Mind that the in the Netherlands generally recommended safe headway is 2s.The mean headway 508 G. Hegeman et al. between vehicle 1 and 2 at the start of the manoeuvre is 17.8 m ± 9.8 m. ‘Flying overtaking’ and ‘Piggy backing’ have significantly longer headways (t = 2.503, t = 2.805, p< 0.05) then the ‘accelerative’ strategy. Frequency [#] 20 14 15 10 Headway prior to manoeuvre 0 18 17 Headway after manoeuvre 0 10 9 5 5 5 4 2 1 30-40 40-50 Bins [m] 50-60 1 0 <10 10-20 20-30 60-70 >70 Figure 2. Distribution of distance prior to and after the overtaking manoeuvre The distance between the overtaken vehicle and the overtaker after the overtaking manoeuvre is bigger than at the start. The mean headway at the end of the manoeuvre is 32.5 m ± 12.2 m. There is no significant difference between the four manoeuvre strategies or speed driven by the overtaken vehicle. Short distances at the end of an overtaking manoeuvre are less dangerous than at the start, since the speed of the overtaker is then (assumed to be) higher than the speed of the vehicle that is overtaken. Figure 3 shows a box plot of the duration of the overtaking per strategy. The mean duration of all overtaking manoeuvres in the collected data is 7.8 s ± 1.9 s. Whether the duration of the overtaking manoeuvre is dependent on either speed of the overtaken vehicle or on the overtaking strategy is studied simultaneously with linear regression. Compared to a accelerative overtaking manoeuvre and speed of the vehicle that is overtaken equal to 80 km/h, both a variation in speed between 70 and 90 km/h and a variation in strategy shows no significant difference in the duration of the overtaking manoeuvre at the 95% confidence level. However, the duration of an overtaking manoeuvre for the overtake strategy ‘Piggy backing’ turned out to be significantly shorter than ‘2+’ (t = 2.494, p<0.05). To enhance insight in separate parts of the overtaking manoeuvre, the duration of the move to the left, the time spent on the left lane and the duration of the move to the right are compared as well. Again, no significant differences between the speed driven by the vehicle that is overtaken or overtake strategy for all distinguished parts were found. The final observed measure is the time to collision (TTC), defined as the time between the moment the manoeuvre is completed and the passage of the first oncoming vehicle. For most observed manoeuvres in this study, it took more than 10 seconds before the first oncoming vehicle arrived. Five overtaking manoeuvres were performed with a TTC less than three seconds left before the first oncoming vehicle passed, with the lowest measured TTC 1.2 seconds. The size of a gap in the oncoming traffic stream could have influenced the duration of the overtaking manoeuvre. However, the observed overtaking manoeuvres with an oncoming vehicle within 5 seconds after the overtaker has moved back to the right lane were not significantly shorter than other manoeuvres, at the 95% confidence level. Observations overtaking manoeuvres on bi-directional roads 25 Total overtake time (s) 14 12 12 10 8 6 14 4 normal flying piggy backing 2+ 509 Duration total overtake manoeuvre (s) v_veh2 (km/h) N (#) Mean (s) S.D. (s) 70 12 7.7 1.9 80 20 7.8 2.2 90 11 8.1 1.3 Total 43 7.8 1.9 Duration total overtake manoeuvre (s) Strategy N (#) Mean (s) S.D. (s) normal 26 7.9 1.5 flying 6 6.8 1.4 piggy backing 6 7.8 2.0 2+ 5 9.1 3.4 Total 43 7.8 1.9 Overtake manoeuvre type Figure 3. Box plot duration of overtaking per strategy. Number of overtakings mean and standard deviation are shown per driven speed and per strategy. 5. Conclusions and discussion observation overtaking manoeuvres Almost fifty overtaking manoeuvres on two-lane rural roads with opposing traffic and a speed limit of 100 km/h were observed in great detail by means of an instrumented vehicle. At least nine drivers did not use their indicator at the start of the overtaking manoeuvre. This finding has implications for design possibilities of overtaking assistance systems. Such a system cannot rely on indicator use, since not all drivers use their indicator. An existing Japanese lane change assistant, available off-the-shelf, is based on indicator use [9]. From our observation results it should be recommended not to rely on indicator use, as was recommended by [10], who also observed drivers not using the indicator. The observed headways prior to an overtaking manoeuvre were a lot smaller than the generally recommended safe time distance of 2s. If a vehicle is equipped with e.g. an adaptive cruise control or with a collision avoidance system, these systems will either adjust the distance or give a danger warning if drivers perform an overtake manoeuvre as they do now. This implies that either such systems should be switched off during overtaking manoeuvres (undesired) or the performance of an overtaking manoeuvre has to be adjusted. If an overtaking assistant is to be developed, we recommend integrating such a system with existing distance keeping systems as mentioned. The observed average duration of an overtaking manoeuvre is 7.8s ± 1.9s. This is somewhat longer then the 6.7s found in [7], but in their experiment, drivers were always confronted with an opposing vehicle and were therefore ‘forced’ to complete the manoeuvre as fast as possible. For this study, unobtrusive overtaking manoeuvres are used. The duration of an overtaking manoeuvre becomes longer if the speed of the vehicle that is 510 G. Hegeman et al. overtaken is higher, but this difference turned out to be non-significant. Also, there is no significant difference in duration of the manoeuvres between the four overtake strategies: accelerative, flying, piggy backing and 2+. This results makes it feasible to develop one overtaking assistant, serving all overtaking strategies and a wide ranges of speed driven by the overtaken vehicle. Finally, the observed overtaking manoeuvres with an opposing vehicle within five seconds after the completion of the manoeuvre did not result in shorter duration of the manoeuvre. Since the number of overtakings with an opposing vehicle within 10 seconds after the completion of the manoeuvre was low, more research into the effects of short TTCs (time–to-collisions) is recommended. References [1] J. McKnight and B.B. Adams, Driver Education Task Analysis. Volume I: Task Descriptions. 1970, Human Resourches Research Organization (HumPRO): Alexandria. [2] L. Barr and W.G. Najm. Crash problem characteristics for the intelligent vehicle initiative. in TRB 80th Annual Meeting. 2001. Washington DC. [3] J.S. Wang and R.S. Knipling, Lane change/merge crashes: problem size assessment and statistical description, final report. 1994, US Department of Transport: Washington. p. 61. [4] SWOV, Ongelukkendatabase [Accident database]. 2003, Leidschendam. [5] G. Hegeman. Overtaking frequency on two-lane rural roads. Safety possibilities of ADAS. in TRAIL congress. 2004. Rotterdam. [6] A. Benedetto, C. Benedetto, and M.R. Blasis. Reliability of standards for safe overtaking: advances using real time interactive simulation in Virtual reality. in TRB 2004, 83rd Annual meeting. 2004. Washington DC: National Research Council. [7] A. Crawford, 'The overtaking driver'. Ergonomics, 1963. 6(2): p. 153-169. [8] T. Wilson and W. Best, 'Driving strategies in overtaking'. Accident Analysis & Prevention, 1982. 14(3): p. 179-185. [9] eSafetyWorkingGroup, Behavioural Adaptation to an Advanced Driver Support System, in Final report. 2000. [10] S.E. Lee, E.C.B. Olsen, and W.W. Wierwille, A Comprehensive Examination of Naturalistic Lane-Changes. 2004, National Highway Transportation Safety Administration: Washington D.C.