Real Time Detection of Speed Breakers andWarning System for on-road Drivers

Mahbuba Afrin, Md. Redowan Mahmud and Md. Abdur RazzaqueGreen Networking Research Group, Department of Computer Science and Engineering

University of Dhaka, Dhaka - 1000, BangladeshEmail: {m.afrin.ritu, md.redowan.mahmud}@gmail.com, razzaque@du.ac.bd

Abstract—The excessive use of speed breakers on nationalhighways distracts vehicle drivers. In addition to that, driversoften can’t recognize the appearance of unmarked speed breakersand loose control of the vehicle, causing serious accidents andloss of lives. In the literature, there exist a few methods towarn on-road drivers about the upcoming speed breakers whichare highly error-prone and time consuming. Moreover, none ofthem pay any heed to track the information of infringing speedbreakers. In this paper, we come forward with a system thatfacilitates autonomous speed breaker data collection, dynamicspeed breaker detection and warning generation for the on-roaddrivers. Our system also incorporates real-time tracking of driver,vehicle and timing information for speed breaker rule violations.The proposed system outperforms the state-of-the-art works withwhich it is compared to in terms of response time and accuracy.

I. INTRODUCTION

Road traffic accidents have been increasing dramaticallyworldwide, which have become the leading cause of death byinjury and the tenth-leading cause of all deaths [1]. The Globalstatus report on road safety 2013 indicates that worldwide thetotal number of road traffic deaths remains unacceptably highat 1.24 million per year [2].

Human error (e.g. user behavior, speed limit violation, driverfatigue etc.), faulty transportation and highway infrastructuredefects (e.g. patholes, sharp bends, poor road conditions, etc.)are the three major factors that lead to traffic crashes and fatali-ties [3]. Driver-related factors include impaired, aggressive anddistracted driving. A 5% increase in average speed leads to anapproximately 10% increase in crashes that cause injuries anda 20% increase in fatal crashes [2]. Safe speed thresholds varyaccording to different types of road, different types of collisionand road users, with their inherent vulnerability [2]. Motoristsshould drive within speed limits and with a speed consistentwith road conditions. Speed breakers (speed humps, bumps,table, rumble strips) are traffic calming devices constructed tomake an intention of slowing down or reducing speed of avehicle as well as to improve safety for pedestrians [4].

Ignoring speed breaker on highway roads is one of themajor causes for occurring road accidents. According to theRoad Accident Report 2014 published by the road transportand highways ministry in India, while 4726 lives were lost incrashes due to humps, 6672 people died in accidents causeddue to potholes and speed breakers [5]. Speed-breakers areunobtrusive due to low visibility conditions, like when there

is snow, fog, rain or at night. Lack of warning system for theupcoming speed breakers on highways is one of the majorcauses for road accidents due to not controlling speed limit.Real time detection of speed breakers and warning system foron-road drivers in any kind of road condition can be a solutionto lessen the road accident rate due to over speeding.

In the state-of-the-art, a very few number of research workshave been done addressing this problem. In [6], a machinelearning based detection of vehicle braking and road bumpshas been proposed using accelerometer and magnetometer ofmobile phones. A similar work is a smartphone based speedbreaker early warning system using accelerometer [7]. How-ever, smartphone accelerometers are not competent enough totrack existence of speed breakers when the vehicle is runningat high speed. Furthermore, the magnetometer is not presentin all mobiles and [6] requires GPS for orientation, whichincreases overall complexity. Besides, running simultaneouslythe accelerometer and GPS devices in a smartphone is quiteenergy hungry and may produce erroneous results. Moreover,the need to store information regarding violation of alertsystem of speed breakers is not focused in the state of theart works.

In this paper, we bring forward a dynamic route basedreal time detection and warning system for imminent highwayspeed breakers. To collect the information about speed breaker,an autonomous data collection module has been developedwhere the iterative speed behavioral patterns of vehicles areanalyzed and probable location for the speed breakers areidentified. Here, speed breakers are clustered into routesaccording to their longitude and latitude values. Based onthe collected data, speed breakers are detected and warningmessage is generated using Haversine distance formula. Thesystem is designed in such a way, that extraneous databasesearching as well as access to cloud server each time isnot mandatory. Moreover, as GPS provided by smartphonesdoes not give the unerring result always and it quests forextensive access to internet, as a solution external GPS deviceis used for our proposed system which comes up with betterperformance. Therefore, the key contributions of this work canbe summarized as follows:

• Development of an autonomous system for speed breakerdata collection, analysis and route construction.

• A real-time and light-weight mechanism for detectingspeed breakers ahead that is runnable as a Smartphone ap-

2015 IEEE International WIE Conference on Electrical and Computer Engineering (WIECON-ECE)19-20 December 2015, BUET, Dhaka, Bangladesh

978-1-4673-8786-6/15/$31.00 ©2015 IEEE 495

plication. The detection system is smart enough to handlemultiple road-intersections, U-turns and can dynamicallydetermine a new route and its speed breakers.

• Development of an accurate and time-efficient warningsystem so that the driver can control vehicle speed inahead of time. Notification to vehicle owner/manager isan added feature of the system in the case the driverviolates a speed breaker warning.

• Finally, the proposed system works in adverse environ-ments as well, e.g., night, foggy or rainy weather, etc.

The remaining of the paper is organized as follows. TheSection II contains a study on the related work and SectionIII gives an explicit insight of our proposed system. TheSection IV and V elaborate the prototype implementationand performance evaluation, respectively. Finally, the work isrecapitulated in Section VI.

II. RELATED WORK

A machine learning technique to detect road anomalies andbraking events from accelerometer and magnetometer data isproposed in [6]. The method requires magnetometer for reori-entation but magnetometer is not present in all phones whichis susceptible to magnetic interference and increases batteryconsumption as well. Another method for detecting speedbumps and braking events was also proposed in [8] whichdid not differentiate between potholes and speed-breakers. Itrequires GPS for reorientation, increasing overall complexityand battery consumption. To avoid reorientation complexity,mobile phone crowd-sourcing based pothole detection algo-rithm is developed in [9].

An early warning system that uses a smartphone basedapplication to alert the driver in advance when the vehicleis approaching a speed breaker, is being developed in [7]. Theapplication constantly monitors the smartphone accelerometerto detect previously unknown speed-breakers [7]. However, theauthors did not describe in detail the methodology of warningsystem such as how to calculate distance between runningvehicles and upcoming speed breakers. Their system dependsonly on accelerometer and built-in GPS sensor in smartphones.But sometimes the GPS service on smartphones takes a longtime to get a fix in adverse environment and it does not alwaysgive the actual accuracy.

In our proposed system, an autonomous system is unfoldedfor speed breaker data collection, detection as well as to warnthe vehicle-drivers and notify the owners which is efficient interms of time and cost and requires less bandwidth as well aslimited access to database.

III. PROPOSED SYSTEM MODEL

Our proposed system is comprised of three inter-dependentmodules; Autonomous Data Collection of Speed Breaker,Dynamic Detection of Speed Breaker, Warning Generation andNotification as shown in Figure 1. Moreover, here we assumethat each vehicle is provided with a GPS tracker and enabledwith such functionality that helps to communicate with a cloudserver through Internet and to execute small scale computation.

Fig. 1. Proposed System Model

A. Autonomous Data Collection of Speed Breaker

To collect the information about speed breaker location,some volunteer vehicles are configured with speed breakerdata collection module. In general it is seen that just beforepassing a speed breaker drivers slow down the speed of theirvehicles and as soon as they pass the speed breaker, theyincrease the speed as shown in Fig. 2. Taking cognizance of thefact, this module gathers a volunteer vehicle’s speed behavioralpattern against location and sends it to the cloud server. Incloud server all volunteer vehicles speed behavioral pattern areanalyzed and probable location for speed breaker are identifiedwhere vehicles satisfy the breaker speed threshold. To performthis operation, cloud server conducts analysis of variance(ANOVA) on the provided speed behavioral pattern [10] andbreaker speed threshold is set by system administrator. Assimilar characteristic may be shown by the vehicles in othercases, this procedure is highly suspected to create false positiveproblems. Data collection through several iterative runs ondifferent times of a day can made this procedure trustworthy.

1 2 3 4 5

16

18

20

22

24

26

28

30

23.728100 N, 90.394078 E

23.728035 N, 90.394009 E

23.727957 N, 90.393916 E

23.727998 N, 90.393959 E

Spee

d Km/

hour

Time ( 5 second period)

Volunteer vehicle's speed behhavioral pattern

23.728065 N, 90.394049 E

Fig. 2. Volunteer vehicle’s speed behavioral pattern

However, after identifying the probable location, the speedbreakers are clustered according to the changing rate of their

496

TABLE INOTATIONS

Symbol Definition∆ϕ Difference between latitude of current position of vehicle and

routes or speed breakers position∆λ Difference between longitude of current position of vehicle

and routes or speed breakers positionR Earth’s radiusa Haversine factor varribalesc Haversine factor varribalesd Distance between two pointsn Number of routes

longitude and latitude values. If between two speed breakerstheir longitude and latitude are changed in a linear order, theyare assumed to be within a same route. Each route is markedwith the location of two end points. All route and their in liedspeed breaker location related information are then feed intoa cloud storage for further use.

B. Dynamic Detection of Speed Breaker

Dynamic Speed Breaker detection module enables a vehi-cle to detect speed breakers dynamically based on the dataprovided by the volunteer vehicles. This module directlycommunicates with the cloud storage that contains the routewise location related information of speed breakers. As soonas this module commences, it starts to find out the nearestroute from the vehicle’s current location using equation 1. Therelevant notations and definitions for this module are listed inTable I.

mini

n∑i=1

di (1)

Here, di refers to the Haversine distance between ith route andvehicles current location. Haversine distance can be formulatedas follows [11],

a = sin2(∆ϕ

2) + cosϕ1 × cosϕ2 × sin2(

∆λ

2) (2)

c = 2× atan2(√a,√(1− a))

d = R× c

Determining the probable route, this module loads the corre-sponding list of the speed breakers to the database of vehicle’slocal computational unit in a suitable order. It then began tocalculate the Haversine distance between the current vehicle’slocation and the speed breakers. If the distance of the vehiclefrom a speed breaker becomes smaller than a threshold value,Warning and Notification module will generate a warningmessage.

However, when a vehicle is in a particular route, it is surelyapproaching towards one of the route end points or towardsthe next speed breaker. If it is not so, the vehicle is assumed tobe getting out from the route. In this case, the module destroysthe local database and initiates the search for the next nearbyroute. Moreover, it might happen that, the vehicle enters intoa route whose corresponding speed breaker information is not

TABLE IILIST OF ROUTES

Route StartingLatitude

StartingLongitude

EndingLatitude

EndingLongitude

ShibBari-TSC 23.728250 90.395390 23.732106 90.395186Polashi-ShahidMinar 23.727579 90.389736 23.727962 90.395747TSC-Shahbag 23.732805 90.391044 23.728981 90.392342

TABLE IIISPEED BREAKERS OF POLASHI-SHAHID MINAR ROUTE

Serial Latitude Longitude01 23.728035 90.39400902 23.728114 90.39368903 23.728398 90.39239704 23.727945 90.390547

available. In this scenario, the vehicles will act like volunteervehicles and contribute to collect speed breaker related datathrough speed breaker data collection module.

C. Warning Generation and Notification

This module is responsible for alarming the on-road vehicledrivers about the forthcoming speed breaker. For each speedbreaker, a voice recorded alarm will be generated as warningmessage which is played for two times so that the driversdo not get irritated with the continuous alarm while driving.Moreover, it is expected that while passing a speed breaker,vehicles speed should be slowed down. If any driver does notobey this, a notification will be sent to the vehicle owner withproper information. This information might contain the speedcomparison of the vehicle just before and after passing thespeed breaker.

IV. PROTOTYPE IMPLEMENTATION

In order to implement the proposed system, we haveconsidered a smart mobile handset as the vehicle’s localcomputational unit with internet connection. A Garmin GLOBluetooth GPS is used [12] which is wirelessly paired withthe mobile device and provides location and speed relateddata. These data are more accurate compared to the mobiledevice GPS given data. Moreover, a mobile application isdeveloped that combines all the three aforementioned modulesalong with a Java Web Application and JDBC on GoogleCloud Platform to perform cloud assisted computations. Thisdeveloped mobile application is vehicle and device(android)independent.

However, for Speed Breaker Data Collection the mobileapplication gathers mobile device’s accelerometer and GarminGLO provided speed and location related data and sent tocloud server where the speed behavioral pattern of vehiclesare analyzed and probable location of speed breakers are iden-tified. Besides, here the nearby speed breakers are clusteredwithin a single routes and corresponding information are feedinto database. Table II and III partially represent our collectedroute and speed breaker related data respectively.

497

When the application is configured with Dynamic SpeedBreaker Detection module, the route wise speed breaker listis loaded into local mobile device from the centralized clouddatabase. Then, the Garmin provided vehicle’s current locationis compared with the routes and speed breakers location fromthe local database following the method mentioned in sectionIII.

If the next speed breakers is found within 100m distancefrom the vehicle’s current location a audio warning messagewill be played. Moreover, if the driver disobeys the rulesregarding speed breaker a notification will be sent to theowner account hosted in the cloud server along with the speedrelated data of the vehicle. Thus it performs the functionalityof Warning and Notification module.

V. PERFORMANCE EVALUATION

To evaluate the performance of our proposed system wehave voluntarily collected location of 100 speed breakers under35 different routes of Dhaka, Bangladesh. Among them only3% of data gave false positive. And almost no error hasbeen recorded to detect those speed breakers and to generatewarning message and notification. Besides, a significant im-provement in response time is observed when the nearby speedbreakers are assumed to be in lied a single route and instead ofdistant cloud storage, local route and speed breaker databaseis taken into account. Figure 3 demonstrates the response timefor varying number of speed breakers in both route and non-route wise clusterization of speed breakers. Figure 4 depictsthe comparison between local and cloud storage searchingfor speed breakers in terms of response time. In both casesour proposed concept outperforms the remaining. Comparedto mobile device GPS, the use of external GPS gives betteraccuracy (approximately 15% improvement).

10 20 30 40 50 60 70 80 90 100

0.04

0.06

0.08

0.10

0.12

0.14

0.16

Resp

onse

time i

n milis

econ

ds

Number of Speed Breakers

With Route Based clustering Without Route Based clustering

Fig. 3. Response time required in Route and Non-route wise clustering forsearching speed breakers

VI. CONCLUSION

This paper develops a system that can dynamically collect,detect and generate warning for upcoming speed breaker tothe on-road vehicle drivers which is more efficient in terms ofresponse time as well as bandwidth usage. We plan to bringforth more competent system in future to overcome the false

10 20 30 40 50 60 70 80 90 1000.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

Resp

onse

time i

n milis

econ

ds

Number of Speed Breakers

Using local database Using cloud storage

Fig. 4. Response time required for Local and Cloud storage searching forSpeed breakers

positive condition in identification of speed breaker to ensurebetter quality of service.

ACKNOWLEDGMENT

This work was supported by ”Grants for InnovativeProjects” by Information Communication and TechnologyDepartment of the Government of Bangladesh. Dr. Md. AbdurRazzaque is the corresponding author of this work.

REFERENCES

[1] D. Mohan, “Road safety in less-motorized environments: Future con-cerns,” International Journal of Epidemiology, vol. 31 (3), pp. 527–532,2002.

[2] D. M. Chan, “Who global status report on road safety 2013,” WorldHealth Organization, Tech. Rep., 2013.

[3] D. A. Aderamo, “Spatial pattern of road traffic accident casualties innigeria,” Mediterranean Journal of Social Science, vol. 3 (2), May 2012.

[4] D. Bowrey, R. Thomas, R. Evans, and P. Richmond, “Road humps:accident prevention or hazard?” Journal of accident & emergencymedicine, vol. 13, no. 4, p. 288289, July 1996. [Online]. Available:http://europepmc.org/articles/PMC1342738

[5] “Road accident report.” [Online]. Avail-able: http://timesofindia.indiatimes.com/india/Over-11000-people-killed-by-potholes-speed-breakers-last-year/articleshow/48950267.cms

[6] R. Bhoraskar, N. Vankadhara, B. Raman, and P. Kulkarni, “Wolverine:Traffic and road condition estimation using smartphone sensors,” inCOMSNETS’12, 2012, pp. 1–6.

[7] M. Jain, A. P. Singh, S. Bali, and S. Kaul, “Speed-Breaker EarlyWarning System,” in Proceedings of the 6th USENIX/ACM Workshopon Networked Systems for Developing Regions. New York, NY, USA:ACM, 2012.

[8] P. Mohan, V. N. Padmanabhan, and R. Ramjee, “Nericell: Rich mon-itoring of road and traffic conditions using mobile smartphones,” inProceedings of the 6th ACM Conference on Embedded Network SensorSystems, ser. SenSys ’08. New York, NY, USA: ACM, 2008, pp. 323–336.

[9] J. Eriksson, L. Girod, B. Hull, R. Newton, S. Madden, and H. Balakr-ishnan, “The pothole patrol: Using a mobile sensor network for roadsurface monitoring,” in Proceedings of the 6th International Conferenceon Mobile Systems, Applications, and Services, ser. MobiSys ’08. NewYork, NY, USA: ACM, 2008, pp. 29–39.

[10] D. Kim and Y. I. Lee, “Characteristics of driving behavior according toaccumulative task loading time of intra-city bus driver,” in Proceedingsof the Eastern Asia Society for Transportation Studies, vol. 9, 2013.

[11] R. C. Nitin and K. Mangesh, “Landmark based shortest path detectionby using a* and haversine formula,” International Journal of InnovativeResearch in Computer and Communication Engineering, April 2013.

[12] “Garmin glo bluetooth gps.” [Online]. Available: https://buy.garmin.com/en-US/US/oem/sensors-and-boards/glo-/prod109827.html

498