the potential impact on traffic safety of lateral support

0 downloads 0 Views 80KB Size Report
due to the position of the target vehicle and brakes with optimised capability). ..... (3) Bauer H. (ed) (1996) Automotive Handbook. Robert Bosch Gnbh, 4 th.
THE POTENTIAL IMPACT ON TRAFFIC SAFETY OF LATERAL SUPPORT SYSTEMS Gianguido Sala Centro Studi sui Sistemi di Trasporto SpA Corso Re Umberto 30 - 10128 - Torino - Italy Tel 0039-011-5513811 - Fax 0039-011-5513821 E-Mail : [email protected] Lorenzo Mussone Dipartimento di Idraulica, Trasporti e Infrastrutture Civili, Politecnico di Torino C.so Duca degli Abruzzi, 24, I-10124 Torino, Italy +39-11-5645601, [email protected]

SUMMARY The paper reports the simulation activities planned in the first six months of 2000 aiming at investigating the possibility of reducing the number of accidents in crossroads through the use of on board systems. This research, lead within the European Project LACOS, the fourth framework research programme project funded by DG XIIIC, investigates on the potential benefit which can be introduced by the use of electronic on board devices. These devices could be derived as extension from those already developed within the project LACOS. Some hypotheses are done in order to describe the features of these systems, both considering automatic and warning activation. Their features are implemented in an already existing accident simulator and analyzed in different scenarios of traffic and equipment taking into account different vehicular speed, flow, flow composition, braking capability, road surface adherence and device capability.

EXTENSION OF LACOS APPLICATIONS, AN OVERVIEW The project LACOS developed three demonstrator vehicles equipped with distance detection both in longitudinal and lateral direction. The main goal of LACOS system is to provide a support for lateral areas in order to help the driver in the overtaking manoeuvres. At the moment is implemented the warning mode through an haptic human machine interface and information displayed in dedicated display located in the dashboard and in the lateral mirror. Through a simulation approach it is possible to extend the range of LACOS applications to detect vehicles running in crossing carriageways. In fact, previous studies carried out on accident analysis show that crossroad locations are one the most dangerous spots. For the aim of an experimentation by simulation, some assumptions on the features of a new system have been made (figure 1): a) The system can detect moving vehicles in the area limited by the road in front of the equipped vehicle (detection area 1); b) The system can detect moving vehicles in the road crossing that of the equipped vehicle (detection area 1 + detection area 2); c) The system reacts automatically (the system takes the control of the vehicle after a time due to the position of the target vehicle and brakes with optimised capability).

1

Intersection layout

y

B1

B

Dlat A

Xb

Detection area 1

Detection area 2

α Dvis = equipment range angle α Dlat = maximum lateral distance Dvis = scenario distance, maximum distance of visibility of B vehicle

A

Xa

0

ysup x yinf

Dlat xmin xmax

B2

Figure 1:Intersection layout and parameters for the scenario simulations.

SIMULATION CONCEPTS A simulator to study the effect of CAS system was already developed (8, 9, 10) in order to study the evolution of a perturbation within a platoon when unsafe manoeuvres occur. The paper includes a short description of the simulator features, in particular for those aspects that allow us to simulate some accident types in urban environment. The concern in LACOS applications is related to vehicular safety, mainly in motorway environment. The envisaged extension regards the possibility to use LACOS application in presence of crossroads even considering an urban environment. Unsafe manoeuvres are simulated by a platoon of vehicles (“A”) running toward a crossroads. When the platoon leader achieves a certain distance to the crossroads (scenario distance, Dvis figure 1), a car running on the crossing road (“B”) engages the same crossroad at constant speed without giving the right of way to the platoon leader. An algorithm based on “if-then” rules allows driver (or the system if the activation is automatic) to choose the best strategy among the following ones: a) disengagement the crossroad at constant speed b) soft brake c) hard brake The following vehicles in the platoon must adapt their behaviour to the new situation in two ways: the first by braking after a reaction time by using the maximum braking capability allowed by the vehicle, the second one by following the leader if it does not brake. In case of accident, a follower brakes by using the maximum braking capability and results are similar to those shown in (9, 10). The calculation of the secondary effects (or side effects), that is the number and gravity of crashes inside the platoon due to deceleration manoeuvres before the leader collided with a “B” vehicle, is also worked out. The simulator, by a statistical approach, calculates the probability of collision of vehicles in platoons. A different probability can be calculated whether we consider or less the possibility of disengagement of the crossroad at constant speed. A Monte Carlo procedure is applied to guarantee that the various combinations of vehicle features were actually used.

2

VEHICLE SIMULATION The overall vehicle set has been divided into vehicle classes (motorbikes, cars, lorries and vans) and when necessary, in engine size classes. Each class is characterized by typical lengths, widths and mass, deceleration capability, driver reaction time (without equipment) and the percentage of equipment. The proportion of vehicles in each class corresponds to the current distribution for the Italian scenario. The Module Of Vehicle Dynamic If we consider a flat road without additional hooked up weight, the predominant contribution to braking forces is due to the brake system limited by adherence between tires and asphalt (11). Besides adherence many other parameters affect and limit the braking force of a vehicle, they depend on the mechanical status of the brake system, tire condition, anti-look system availability, ability of driver to brake, and so on. An additional parameter “k” is inserted in the calculation of maximum brake force in order to introduce some perturbation in braking manoeuvres. This parameter is constant, related to the “quality” of driver and to the state of conventional mechanical equipment; it is assumed it can vary uniformly in the range 0.85 to 1 of its nominal value. A module simulates for each vehicle the dynamics of motion (braking on a straight or curved stretch) (4) according to the following system of two differential equations: dx2/dt = x1(t) 1) dx1/dt = a(t) 2) in which x2(t) indicates the position of the vehicle, x1(t) its speed and a(t) its acceleration which is a function of the coefficient of adherence µ = µ (x1) which is assumed function of speed. What's more, a generic relationship between adherence and non-constant velocity of a hyperbolic type is hypothesised: µ (x1) = (a x1 + b) / (c x1 +d)

3)

which, in the case of a dry surface, becomes a straight line and on a wet surface a hyperbole (3). The value of the coefficient of adherence is recalculated at every step of integration and set to 0.06 seconds. It must be remembered that higher resolutions would not necessary when speed is already reduced by braking manoeuvres. But, in order to increase precision of simulation results, the step of integration in the area of crossroads is reduced to 0.01 seconds just starting when the vehicle B is crossing the intersection in order to calculate the point of collision between vehicles with an error of few centimetres. Control Strategy Strategies developed in LACOS seem not useful to approach the problem of crossroads; but considering experiences worked out in previous TAP programme projects, in particular the project AC-Assist, a suitable control strategy for automatic intervention could follows the "latest as possible" intervention philosophy. Latest as possible means that the car must brake only at the last moment if the driver has not braked before. The law implemented in the simulator for platoon control works according to the following equation: 2 2 vf vl − + offset 2a f 2al

3

4)

where v and a are speed and acceleration respectively of the leader vehicle (subscript l) or of following one (subscript f). The addition of the constant term “offset” makes this formulation more conservative in order to increase safety of the system itself. The same formulation is used for the calculation of the intervention distance for the warning activation and for the driver alone activation. The calibration of the offset allows to the simulator to estimate a different level of risk perception of drivers. The control strategy is based on performance achievable by sensors installed on board vehicle and processing data autonomously. In the case of a Warning System, the driver increases its driver capability (the simulator assume the driver is paying more attention) and reduces its reaction time randomly (ranges are fixed by literature data), but in any case it keeps the control of the vehicle. In the case the system has the possibility to detect the detection area 2 (figure 1), the system reduces the speed (warns the driver) of the equipped car up to a safer speed (if applicable). The rules adopted for equipped vehicles are the following ones: • no action is undertaken if inequalities 5) and 6) are satisfied, that is no collision is predictable if vehicle A and B travel with the current constant speed; the coefficient of 0.9 increases the correctness of prediction: X b + Lb +

Wa

Vb X a + La + Va



2 < 0 .9

Wb

X a + La +

2 < 0 .9

Xb

Wb

5)

2

Va W − a 2 Vb

6)

if a collision is possible the vehicle starts braking with a constant decceleration of 0.5m/s2.

If a vehicle B is crossing the area 1 the driver of vehicle A starts braking at the maximum deceleration of the vehicle; in this case the difference between equipped and not equipped vehicles concerns their reaction time. The maximum lateral distance of visibility is related to the equipment considered but in the following simulations we consider that for urban networks with many buildings the maximum reasonable value is 20m. Only to investigate sensitiveness of the system higher values are tested. Reaction Times and System Delay The reaction time of drivers for not equipped vehicles was the subject of a particular survey (2, 5, 6, 7); these papers reports about driver reaction times and specifically in a braking manoeuvre but unfortunately only the paper of Gordon (5) proposes a distribution of reaction time according to driver characteristics. Because other papers don’t deal with this aspect in such a detail but they are consistent with the average values, we use data proposed by Gordon (5) as reported below. Percentile % 50 75 Braking reaction time [s] 0.85 1.11 Table 1: Braking reaction time distribution

90 1.24

95 1.42

97 1.63

99 2.16

In equipped vehicles the control strategy is assumed to work in the range 0.150 ÷ 0.250 seconds. This time consider also the time spent to activate the brake. Considering warning systems, the warning time is very reduced (sum of the time of data acquisition and the 4

calculation of the algorithm) and in the worst cases can be of 50-80 msec. As said above, this time must be added to the reaction time of the driver. Simulation scenarios By varying road conditions (a, b, c, d in eq. 3) and its geometry (Xmin, Xmax, Ysup, Ying in fig.1), the equipped vehicle rate, the composition of vehicular traffic (according to (1)), the initial conditions (Xa, Xb in fig. 1), we are able to define different scenarios. Platoon headway obeys to a negative exponential distribution and it ranges in the interval 3-18 seconds. The average speed for A vehicles is extracted randomly in the intervals 90±15km/h and 50±10km/h and for B vehicles in the intervals 90±15km/h, 50±10km/h and 15±5km/h, according to the scenario. In such a way flow is ranging between 200 and 1200 veh/h and density between and 12 and 40 veh/km. Scenarios take into account the other following main aspects: • the different braking levels of vehicles. • the different distance of visibility. • the different percentage of equipment. The different braking level concerns the capability and habit of drivers seldom accustomed to face a vehicle not giving the right way. For each scenario the percentage of equipment is varied with a distribution within classes according to the principle that luxury cars and heavy vehicles will be first equipped with the new devices. The total equipment rate is set to 0%, 10%, 25% and 50% in order to study the evolution of market introduction and to ensure a good sensitivity analysis. The combination of these scenarios covers the main part of cases in urban or sub-urban crossroads both for geometry and flow conditions. Number of A->B collisions "A" speed 90km/h scenario

Number of collision (%)

35

Percentage of equipment 30

0%

10%

25%

50%

25

20

15

10

5 90 / 800

50 / 800

15 / 800

"B" Speed [km/h]

/

90 / 1200

50 / 1200

15 / 1200

/

"A" Flow [veh/h]

Figure 2: Number of collisions “A against B” changing “B” speed, “A” flow and the percentage of equipment. Evaluation methodology The analysis aims at investigating the number of accidents calculated by simulations and carried out with and without equipped vehicles. The number of iteration for each scenario has been set to 10.000, a value singled out empirically which guarantees an estimated error less than 3 per thousand and it is not too much time consuming.

5

Parameters considered in the result analysis are: • Number of collisions • Equipment (Automatic/warning detection area 1, Automatic/warning detection area 1 + 2, no equipment) • Collision speed and mass of vehicle A, mass of vehicle B • Point of collision • Average speed and spacing of vehicles in platoon and side effects (crashes) • Type of collision, vehicle A against B (A->B) and vice versa (B->A). Results are recorded into three types of files with three different degree of detail: • with the total of cases; • with the description of A leader and B vehicle features (speed, mass, point of collision, type of collision, equipment) for each iteration; • with the description of features of all A vehicle in the platoon for each iteration.

RESULTS Obviously the number of collisions increases by decreasing speed because the intersection area remains busy for a longer time. With no equipped vehicles (0% percentage) the number of A->B and B->A collisions depends on speed difference and on flow. If A speeds is comparable to B speed the A->B and B->A number of collisions are similar; the higher the A speed the higher the A->B number of collisions and vice versa. By increasing flow the number of cases without collisions decrease with a ratio 7 to 1 (by changing flow from 200 to 1200 veh/h the relative number of non collisions decreases from 1 to 0.8); the number of B>A collisions increases with a ratio of about 1 to 1 instead the number of A->B collisions increases less with a ratio of about 1 to 0.7. This result can be attributed to the fact that A vehicles can try to avoid collision only by decelerating but this strategy is less effective for longer vehicles. The percentage of equipment doesn’t change the trend but it affects the reduction of collision number as shown in Table 2. The ranges are wide depending on “B” speed (from 90 to 15 km/h), the lower it is the higher the reduction. Reduction of the number of collisions (%) “A” speed Percentage of equipment 90 km/h 50 km/h 10% 0.2 – 3 0.1 – 8 25% 0.2 – 10 0.01 – 19 50% 0.3 – 14 0.05 – 31 Table 2: Reduction of the number of collisions by changing the percentage of equipment. Side effect (that is crashes inside the platoon ) is present only when A speed is much higher than B speed but it is irrelevant because in the worst case it arrives to the 1.5% of all cases. The collision speed of vehicles “A” is not greatly affected by equipment if the speed of vehicle “B” is greater than or equal to initial speed of vehicles “A” but instead it decreases significantly, also for “A” not equipped vehicles, if the speed of vehicle “B” is lower. For equipped vehicles the speed reduction depends on the value of the starting speed: the best cases reduce it of about 30% and 50% for a starting speed of respectively of 90 and 50 km/h (figure 3 shows cases for initial speed of vehicles “A” of 90 km/h).

6

When considering a very slow vehicle “B” (4 km/h, like a pedestrian) the equipment can reduces the collision speed no more than 10% but on the other hand the number of A->B collision becomes almost nil at 50% of equipment. Lateral distance of visibility can affect collisions only if the time to collision of “A” vehicle is not negligible to the time needed to “B” vehicle to cross the intersection. In fact, if “A” speed if much greater than “B” speed, “A” vehicles that can actual collide with the “B” vehicle can see it only when it is very close to the stop line. Therefore the additional braking time cannot be very significant. Equipment generally reduces the kinetic energy of collisions especially for equipped vehicles (figure 4). The degree of reduction depend firstly on “B” speed and secondly on the percentage of equipment. Considering a “B” speed of 15 km/h it ranges around 60% for an “A” speed of 90 km/h and around 70% an “A” speed of 50 km/h.

CONCLUSIONS The simulation programme has allowed us to investigate some traffic situations recurrent in urban networks in which an error in driver behaviour can create at risk situations or even an accident. The results show a reduction of the number of collisions increasing the percentage of equipment so as in the collision speed; the kinetic energy of collisions instead decreases much more and it is due to the fact that vehicles (particularly heavier ones) can brake more quickly (longitudinal control) and with more time (lateral control). But increasing the distance of visibility or the lateral distance of visibility doesn’t affect much performance because hard braking is reasonable only when a vehicle crosses the stop line without giving the right of way. This means that better results can be achieved only by controlling behaviour or decisional errors. 15 / 800

90km/h "A" speed scenario

26.00 24.00 22.00 20.00

Collision speed [m/s]

18.00 0%

16.00 14.00

10%

12.00

25% 50%

15 / 1200 15 / 800 15 / 400 15 / 200 50 / 1200 50 / 800 50 / 400 50 / 200 90 / 1200 90 / 800 90 / 400 90 / 200

Percentage of equipment

B vehicle speed [km/h] A flow [veh/h]

Figure 3: Collision speed for equipped vehicles changing “B” speed, “A” flow and the percentage of equipment.

7

Average kinetic energy of collision "A" speed 90km/h, "B" speed only equipped

Average kinetic energy of collision "A" speed 50km/h "B" speed 50km/h Kinetic Energy only equipped vehicles [joule]

Kinetic Energy

300000

[joule] 275000

85000 0 80000 0 75000 0 700000

of

equipment

10%

800

800 400 50 %

200

"A" Flow [veh/h]

Percentage of equipment

400 50%

10 % 25

Percentage %

200000 1200

65000 0

25%

1200

250000 225000

200

"A" Flow [veh/h]

a) b) Figure 4: Kinetic energy of collisions changing “A” flow and the percentage of equipped vehicles for “A” speed of 90 km/h (a) and 50 km/h (b).

AKNOWLEDGEMENTS This paper is supported by the project founded by the EC DGXIII LACOS. Project partners: Centro Ricerche Fiat (Coordinator), ADC, MAGNA, Magneti Marelli, TÜV, Volkswagen, Renault, CSST.

REFERENCES (1) (2)

AA. VV., Le auto in cifre 1998, ANFIA, Torino Allen R.W., Rosenthal T.J., Hogue J.R. (1996), Modeling and simulation of Driver/vehicle interaction. Society of Automotive Engineers, report 960177. (3) Bauer H. (ed) (1996) Automotive Handbook. Robert Bosch Gnbh, 4th edition, Stuttgart, 1996. (4) Gillespie T.D. (1992). Fundamentals of vehicle dynamics. Society of Automotive Engineers, Inc., USA, 1992. (5) Gordon D. A., McGee H. W., and Hooper K.G. (1984), Driver characteristics impacting highway design and operations, Public Roads, vol. 48, No.1, June 1984. (6) Palmertz C., Jacobsson L., and Karlsson A-S. (1998), Pedal use and foot positioning during emergency braking, IRCOBI conference, September 1998. (7) Ray L.R. (1996), Nonlinear tire force estimation and road friction identification: field test results. Society of Automotive Engineers, report 960177. (8) Sala G, Clarke N., Carrea., and Mussone L. (1997), Expected impacts of Anti-Collision Assist Applications, Fourth World Congress on ITS, Berlin (9) Sala G., Mussone L. (1999a) “A simulator for the evaluation of the impact on traffic safety of innovative collision avoidance systems”. 8th WCTR., Selected papers, Vol.3, 541-554, Elsevier Science, Oxford, 1999. (10) Sala G., Mussone L. (1999b) "The Evaluation Of Impact On Traffic Safety Of AntiCollision Assist Applications”, Proceedings of 5th World Congress On Intelligent Transport Systems, Ertico, 9-11 November 1999, Toronto, Canada, pp.1-10. (11) Warner C.Y., Smith G.C., James M. B., and Germane G.J. (1983), Friction application in accident reconstruction. International congress and exposition, Detroit, paper n. 830612.

8