Available online at www.sciencedirect.com

ScienceDirect

Procedía Technology 17 (2014) 640 - 649

Conference on Electronics, Telecommunications and Computers - CETC 2013

Algorithms for road safety based on GPS and communications

systems WAVE

Fábio Cardosoa*, Antonio Serradora, Tomé Canasb

aInstituto Superior de Engenharia de Lisboa, Rua Conselheiro Emídio Navarro, 1, 1959-007 Lisboa, Portugal bBrisa Inovagao e Tecnologia, SA Lagoas Park, Edificio 15, Piso 4, 2740-262 Porto Salvo, Portugal

Abstract

Road accidents are problems of great importance to all humanity. The objective is to develop a system, through WAVE communication, V2X, which act so as to alert the driver of collisions risk, implementing algorithms that uses data received by a GPS system. Highlighting three scenarios: crossing or intersection; sudden approach to the vehicle ahead and road curves. Results show that it is possible to avoid road accidents, obtaining the indication of a collision risk with crossing vehicle 15 meters before interception.

© 2014PublishedbyElsevier Ltd.This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/3.0/).

Peer-review under responsibility of ISEL - Instituto Superior de Engenharia de Lisboa, Lisbon, PORTUGAL.

Keywords: Road Safety Algorithms; Collision Warning; Vehicle-to-vehicle; WA VE communication; GPS system.

1. Introduction

In the future, communications among vehicles and road infrastructures will be a reality. For such communication systems IEEE as defined the 802.11p [1] standard, which, together with sensors or other systems like GPS and vehicle CAN bus, can generate applications with great interest to road safety as a preventive mechanism. Being active, for example, acting on vehicle's brakes or just warning drivers to take appropriate precautions. Such applications can therefore prevent road accidents.

* Corresponding author. Tel.: +351-912-823-970. E-mail address: 32883@alunos.isel.pt.

2212-0173 © 2014 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).

Peer-review under responsibility of ISEL - Instituto Superior de Engenharia de Lisboa, Lisbon, PORTUGAL. doi:10.1016/j.protcy.2014.10.187

There are two main devices on WAVE communications, the Road Side Unit (RSU) and the On Board Unit (OBU), two important definitions in the standards IEEE 1609 [2][3].

This paper presents several road safety algorithms based on WAVE communication, vehicle-to-vehicle, using GPS , which warns drivers when faced to risk of collisions, assuring a system as reliable as possible by minimizing situations of false warnings. It this paper several algorithms are implemented and compared and a new strategy is proposed and tested in a representative road scenarios, such crossing or intersection; sudden approach to the vehicle ahead and road curves.

This is a result of greater importance to humankind, because, road accidents were causing 32 000 deaths in the United States in 2010, according to The New York Times [4], more than 30 000 deaths per year in the European roads [5] and, according to Diario de Noticias, about 1000 people die each year in Portugal [6], is expected to reduce these values with the introduction of systems like the one developed in this project.

This paper is organized in order of the various phases of this project. The second chapter presents the study phase, which has a description of some most relevant algorithms published in literature. The development is detailed in Chapter 3, which presents new methods and algorithms to get a more reliable and complete system. Chapter 4 presents some results obtained. In Chapter 5, the most relevant results are presented.

2. State of the art

Several research and development companies of automobile brands, among others, develop algorithms for sudden approach and crossing scenarios, some of them mentioned throughout this chapter.

2.1. Sudden approach

The scenario of sudden approach to the vehicle ahead is analyzed by one dimension, as scheme present in Fig. 1. Where, based on vehicle speeds, v1 and v2, the distance between them, d, and other factors, it is intended to calculate the critical warning distance, dw.

Fig. 1 - One dimension analysis scheme.

The Mazda [7] uses a kinematic analysis to determine critical braking distance, that is given by

l(v2 (v1 - vrel)2\ t

dw[m] = -1---I + v1t1 + vrelx2 + d0 ,

2 \at a2

where vrel is the relative velocity between the following and leading vehicles, t1 the system delay, t2 is the driver delay and ax and a2 are the maximum decelerations of the following and leading vehicles, respectively. To make the algorithm more conservative, a headway offset d0 is added to the critical distance.

Honda's system [7] is based on experimental data, where critical warning distance dw is given by

dw[m] = 2,2vrel + 6,2 . (2)

The PATH algorithm [7] is a modified version of Mazda's algorithm. The critical warning distance is given by

W r 1 1(V* K - Vrel)2\ , _ ...

dw№]=2\—----J + v^ + do , (3)

where a is the maximum deceleration of the vehicles and t is the total delay for the system and driver.

The PATH algorithm aims to obtain a more conservative warning distance than Mazda. However, warning is scaled by a non-dimensional warning value w given by

d dfrr

w = d -d , (4)

where dbr is the critical braking distance which is calculated as 1 2

^br[m] = vrelT+2aT . (5)

Decreasing w shows increasing collision risk. w>1 means safe driving situation. The system gives a warning if w is smaller than the audio warning value. The system triggers brakes if w<0.

The algorithm presented in the literature of the IEEE [8], get the critical distance by converting the reaction time of the driver and the vehicle braking to distances, resulting

dw[m] = TRv1 +0,1v1 + 0,006v^ , (6)

where TR is the reaction time of the driver.

NHTSA algorithm is also based on kinematic analysis, including information of current vehicle acceleration, that results a consistent equations present in [9].

2.2. Crossing or intersection

The IET, Institution of Engineering and Technology, has developed algorithms for two dimensions, based on dynamic equations of vehicles motion, x(t) and y(t) [10].

Considering the vehicles as points in space, particle model, collision exist when

J(x1(t1) - x2(t2))2 + (y1(t1) - y2(t2)f = 0 . (7)

The collision happens when t1 and t2 are equal, this value is appointed TTC, Time to Collision. If there is no valid solution means that vehicles are not at collision risk.

To the circle model, example in Fig. 2, collision exist when

J(x1Ct1) - x2(t2))2 + (y1Ct1) - y2(t2))2 = 2R , (8)

R = VL2 + W2/2 , (9) being L and W the vehicle dimension, length and width, respectively.

Another variant of this algorithm, and closer to reality, is rectangle model, Fig. 3. Where Znm is the coordinate (xnm , ynm) of vertex m on vehicle n, and the vehicles are on a collision course if a vertex is inside some vehicle rectangle area, confirmed with condition

x2m < max(xlm,xlh},

where, if m=1 or m=3, so h=2 or h=4, and vice versa. In the example of Fig. 3, the condition that detects collision risk will be min(x11,x14) < x22 < max(xu,x14).

The Safety Zones algorithm [11] consists in the creation of a safety virtual rectangular zone around the vehicle and detects the overlap areas between two or more vehicles, which indicate collision danger.

The virtual rectangular zone vertices calculating method is identical to that performed in the IET algorithm, rectangle model, with the difference in rectangle length, replacing occurrences of L by 2F, where F, designated safety distance, is identical to dw.

Fig. 2- Two dimension analysis scheme, circle model.

Fig. 3 - Two dimension analysis scheme, rectangle model.

3. Devolopment

3.1. False alarms

The algorithms contained in the state of the art are very susceptible to generating false warnings. The reason is that we are facing the real world, which is quite different from the model used, not only by the vehicles do not follow perfect trajectories, but also due to errors of systems around. For example, a common GPS system has errors up to 10 meters [12].

It can be observed two cases of false warnings to sudden approach scenario in Fig. 4 a), parking situation and opposite direction circulation, this being easily eliminated by the vehicle azimuth.

In two dimensions analysis also are visible two scenarios for potential false warnings, presented in Fig. 4 b), which concerns the situation of curve or bridge, which are easily eliminated with maps inclusion, this is not the case of this project development, but that would be a better option in a future project, connecting the road safety algorithms with the common GPS paths planning.

Some solutions to these cases mentioned below.

Fig. 4 - False warnings on crossing.

3.2. Snake algorithm

The definition of snake algorithm consists in knowledge of vehicles tracks, geographic coordinates, velocities and directions, past and present, named this track to snake.

Adopting a solution that know the coming vehicles snake, can contribute to correcting false alarms, for example, knowing that vehicle ahead made a curve, not a change of direction (different radius of curvature and speed), we know that we are not facing a crossing. It giving birth to curve algorithm, which indicates the driver about speeding for approaching curve, specified below.

Algorithm utilities:

• Acceleration: based on algorithms studied is not considered if vehicle shows decreasing or increasing velocity. Information needed to prevent a false warning, since vehicle shows to decreasing velocity, it can omitting warning generated by algorithms. With the knowledge of snake easily gets current acceleration by averaging the last points of this.

• Stopping: Of the information obtained through the GPS system which, if it not contains a preprocessing, is false, is the azimuth when there is no movement, ie, a situation of stop at a stop sign, red light, walkway, parking, etc. Snake algorithm corrects this problem. When is given a zero speed, is saved the values in snake using the azimuth from the previous position, and back to register values when is noticed movement.

• Parking: Parking is one of false warning scenarios, which can also be eliminated using snake algorithm. It detects a vehicle leaving the parking lot when all points of their snake are inside a given area. To facilitate this detection considers a circular area, and vehicle is in parking lot if all positions of his snake complying the condition (x-a)2+(y-b)2<r2, where r is the radius of circular area, (x, y) is the snake point coordinate and (a, b) the coordinate of area center that corresponding to first snake position.

3.3. Curve algorithm

The Curve algorithm consists of alert the driver that circulates at excessive speed for the approaching curve, based on knowledge of coming vehicle snake. It is also the algorithm that it will distinguish between intersection/junction and curve, one aspect reported as a false warnings generator.

Is obtained the curve radius by two points from snake, start (x1, y1, a1) and end (x2, y2, a2) of curve, which is obtained traversing the snake points to detect a change in azimuth, we can put them on a circle and take the radius value, from which it deduces the following expression:

R=-,p = \a2-ai\, d = j(Xl-x2)2 + (y1-y2)2. (11)

Civil engineering have been studied algorithms for the curve circular arc radius projection as a function of maximum speed intended for that roadway, which, according to [13], has the relationship

R = 127 (ft+Se) , (12)

where V is the max velocity in km/h, R the curve circular arc radius, Se is the overflow in percent and f, the transverse attrition coefficient. In Europe, at this time the overflow is limited to a value of 7% and the transversal attrition coefficient depends on the speed, values at [13], to 50 km/h have the value 0,16. With this is obtained the maximum speed for that curve given by

vmox[km/h] = V« x 127 (ft + Se) = 5A4R . (13)

To warn the driver it is necessary a warning area, to easily way, a circular area, where the center (ac, bc) is the snake point that corresponding to begin of the curve. Also is registered azimuth of begin point, being the entry curve azimuth <pc. The radius Rw of this area must consider some aspects, width of roadway, sample rate and GPS system error. It was obtained the expression

R =—— + 15 (14)

w[m] 2 .

So, the driver is warned if the vehicle, situated in coordinate (x, y) with azimuth (p and velocity V:

• Is inside the warning area: (x - ac)2 + (y - bc)2 < R^;

• Runs in curve direction: ly - q>cl < 20°;

• Runs at higher speed: V > Vmax.

The distinction between curve and crossing is easily obtained by this algorithm. According to practical results, in a crossing, vehicles describe rays up to 12 meters, in order to fulfill the call perpendicular in the Highway Code. If curve approaching submit a radius greater than this value the collision detention algorithms will not run.

3.4. Implementation

It is intended the implementation of a complete system, from the physical layer to the application. Lower layers are guaranteed by a system developed to WAVE communications.

The top layers, application, have been implemented in this project in C programming language, over a single board computer, the module Raspberry Pi Model B [14], with the operating system Arch Linux ARM. Communication with the lower layers is done by USB.

The network layer provides the messaging WSMP, Wave Short Message Primitive, and has been developed according to the architecture in standard IEEE 1609.3 [2].

The main application consists of a simple graphical presentation, in a touch screen, with information of vehicles around, considered augmented reality. Each vehicle makes known its location to other through broadcast messages by WAVE communication. The application executes the algorithms selected through tests and generates a graphic warning when there is risk of collision. After several tests and analyses the algorithms implemented in the final system are: Safety Zones; Snake and Curve.

The GPS system used is the Motorola M12 Kit Evaluation Board with RS232 communication on the NMEA 0183 protocol. The screen used is TFTCar CTF1020-5 10.2", connected by RCA to the graphics and USB to the touch.

Was developed in C# a simulation software, which allows to test the algorithms without depending on hardware, and allows the simulation of physical vehicles.

The complete system consists of two devices, one RSU, which receives data from simulation software by Ethernet, and sends it via WAVE, simulating physical vehicles, and one OBU, which corresponds to the equipment present in vehicle, that contains the application described above, schematized in Fig. 5.

Fig. 5 - Complete system wiring diagram and the corresponding hardware photo.

4. Results

4.1. Sudden approach

Firstly, one dimension algorithms are compared, in order to select the most indicated. In Fig. 6 is an example of dw variation on v2 function, to 50 km/h of following vehicle speed. Mazda and PATH algorithms show to be the better solution, because IEEE algorithm has the drawback of not depending on coming vehicle and Honda's algorithm didn't have security distance to close velocities.

Warning Distence (v=50 km/h)

Mazda Honda _ • PATH • IEEE _

. - - , , * .'

v2 [km/h]

Fig. 6 - Sudden approach algorithms dw variation.

Fig. 7 - Sudden approach scenario test on highway.

Can be seen results in highway of sudden approach scenario, in Fig. 7, extracted from simulation software, where the blue vehicle has the OBU and circulating to 120 km/h when it was alerted for collision risk with the red vehicle, 60 m ahead circling to 40 km/h.

4.2. Crossing

In digital way, the IET algorithm proved to be very effective, which was not confirmed when it used real coordinates, due to errors from the GPS system and from imperfect vehicles trajectories. After some terrain experimentations, Safety Zones algorithm shows to be very consistent even with real coordinates, because do not expect a collision point but a safety area. It is visible his correct operation in examples of Fig. 8.

Another advantage of this algorithm is its operation identical to one dimension algorithms, converging also the sudden approach scenario, example in Fig. 9.

Fig. 9 - Safety Zones algorithm operation example for sudden Fig. 8 - Safety Zones algorithm operation examples. approach

In Fig. 8, it can seen both trails to simulate a real collision course in a junction, where the yellow vehicle implements the algorithm, and circulating at 30 km/h when it has been alerted to 15 m from the junction of collision risk with another vehicle, red colour, that circulating at 40 km/h. This example was obtained based on real trajectories previously saved, identical results were obtained on terrain tests facing a real scenario in the same crossing.

4.3. Curve

In Fig. 10 is some results of Curve Algorithm, where turns n° 2, 5, 6 and 7 are curves and the remaining are crosses, that confirms the radius and average speed differences between both.

A terrain test facing a real curve situation can be seen in Fig. 11, photograph of the real system. With data received from the vehicle ahead, green vector, was obtained curve data, with 33 meters radius and 31 km/h maximum speed, and driver was warned of speeding. The red circle represents the warning area.

5. Conclusions

By the results perceives the possibility to predict and avoid road accidents, main focus that makes this project so interesting.

The sudden approach scenario, which is the most studied in literature, has broad options to choose and showed a high efficiency. Electing Mazda and PATH algorithms as the most suitable, that, in highway, for 100 km/h are obtained warning of a vehicle to 20 km/h at a distance of 95 meters, ideal distance to take the proper precautions. Its result was also useful for two-dimensional algorithms, introducing in virtual rectangle size calculation of Safety Zones algorithm.

The warning on crossing situation also proved successful using Safety Zones algorithm. Where algorithms with very theoretical concept shown to be incapable facing a real situation, which is the case of IET algorithm, that appeared to be good in simulation. This scenario proved to be very fluid and useful in field tests, obtaining the indication of a crossing vehicle 15 meters before interception.

The development of Snake algorithm proved to be very useful and versatile, thus greatly improving the results of studied algorithms. For a first approach of the new concept, overspeed alert to a curve, showed good results, the automatic calculation of the maximum speed for a given curve is close to ideal, for a curve in a town with 33 meters radius was defined and warned a maximum speed of 31 km/h, on the field seemed adequate. However, even though Curve algorithm generates a correct warning, this largely depends on the proper driving of motorists, which is not the best way.

An improved version of this project would be the inclusion of maps, linking algorithms with a common GPS paths planning system, fixing many problems highlighted.

Acknowledgment

Authors acknowledge the HEADWAY project members at ISEL, GIEST, and Aveiro University, IT, Instituto de Telecomunicafdes, in collaboration with BIT, Brisa Inovagao e Tecnologia.

References

[1] IEEE Computer Society, IEEE Standard for Information Technology Telecommunications and Information Exchange Between SystemsVLocal and Metropolitan Area Networks-Specific Requirements Part 11, Wireless LAN Medium Access Control (MAC) andPhysicalLayer (PHY), IEEE Std 802.11p, NY USA, July, 2010.

[2] IEEE Vehicular Technology Society, IEEE Standardfor Wireless Access in Vehicular Environments (WAVE) - Networking

Services, IEEE Std 1609.3, NY USA, Dec., 2010.

[3] IEEE Vehicular Technology Society, IEEE Standard for Wireless Access in Vehicular Environments (WAVE) - Multi-

channel Operation, IEEE Std 1609.4, NY USA, Feb., 2011.

[4] The New York Times, Traffic Deaths Fall to Lowest Level Since 1949. Available in:

http://wheels.blogs.nytimes.com/2011/12/09/traffic-deaths-fall-to-lowest-level-since-1949/ [Accessed on 25-09-2013].

[5] European Commission, Mobility And Transport, Road Safety, Statistics - accidents data. Available in:

http://ec.europa.eu/transport/road_safety/specialist/statistics/ [Accessed on 25-09-2013].

[6] DN Portugal, 937 mortes em acidentes rodoviários no anopassado. Available in:

http://www.dn.pt/inicio/portugal/interior.aspx?content_id=1929570 [Accessed on 16-09-2013].

[7] Intelligent Vehicles Symposium, Development of a Collision Warning System for Adaptive Cruise Control Vehicles Using a

Comparison Analysis of Recent Algorithms, Tokyo, Japan, June, 2006.

[8] Paul Alexander, David Haley and Alex Grant, Cooperative Intelligent Transport Systems: 5.9-GHz Field Trials,

Proceedings of the IEEE, Vol. 99, No. 7.

[9] National Highway Traffic Safety Administration, Development of an FCW Algorithm Evaluation Methodology With

Evaluation of Three Alert AlgorithmsFinal, Final Report, Virginia, USA, June, 2009.

[10] Yunpeng Wang, Wenjuan E, Daxin Tian, Guangquan Lu, Guizhen Yu, Yifan Wang, Vehicle Collision Warning System

and Collision Detection Algorithm Based on Vehicle Infrastructure Integration, Beihang University, Jilin University, Jilin, China.

[11] Anurag D, Srideep Ghosh and Somprakash Bandyopadhyay, GPS based Vehicular Collision Warning System using IEEE

802.15.4 MAC/PHY Standard, Indian Institute of Management Calcutta, AKC School of Information Technology University of Calcutta, Kolkata, India, 2008.

[12] Bruno Moreno Stuck and Hélio Koiti Kuga, Medindo distancias através de um único recetor GPS, INPE-12331NTC/321,

Instituto Nacional de Pesquisas Espaciais, Ministério da Ciencia e Tecnologia, 2005.

[13] ENGIVIA, Norma de tragado, Revisao, Disposifoes Normativas, Instituto de Infra-Estruturas Rodoviárias IP, Lisboa,

Portugal, Nov., 2010.

[14] Element14, Quick Start Guide The Raspberry Pi - Single Board Computer.