Investigation of the impact of a wireless Fog Warning System with respect to road traffic on a highway


Sudden visibility reductions on highways due to foggy weather conditions often lead to a drastic increase in car crash risks. Indeed, fog formation distorts drivers’ perception and judgment of inter-vehicular distances, vehicles’ speeds, and braking distances. In order to support drivers in dealing with the impact of fog, various on-board warning systems are being deployed today. Despite their added value, these systems are still in need of efficient solutions supporting smooth vehicle’s acceleration/deceleration profiles. This is to avoid sudden braking (hence, higher car crash risks) incurred by sensor technologies restricted to line of sight measurements. To meet this goal, we advocate in this paper a Wireless Fog Warning System (WFWS) where cooperative awareness messages are disseminated and used for calculating acceleration/deceleration activities. Without loss of generality, we build on IEEE 802.11p WLAN as a basis technology. Using simulations on the open-source vehicular network simulation framework Veins, we demonstrate both the potential of such a system for increasing safety and smoothing traffic flow—as well as of computer simulation as a means of its evaluation.


In spite of recent advances in roadside traffic management systems as well as onboard vehicular technologies, road accidents still result in serious injuries, permanent disabilities, and loss of lives [1], leading thereby to considerable economic losses and drastic social impact. Statistics are showing that in 2016 approximately 25,500 people were killed in road accidents across Europe. On average, about 8% of road fatalities occurred on motorways, 37% happened in urban areas, and most (55%) occurred on rural roads [2]. In India, nearly 150,785 persons were killed in 2016 as against 146,133 in 2015, according to the local Ministry of Road Transport and Highways [3]. The same source is highlighting that in every hour, 55 accidents take place and 17 persons are killed on Indian roads. Several reasons for car accidents have been reported, of which weather conditions are major ones. In this regard, statistics showed that on average, there are over 5,748,000 vehicle crashes each year in the USA, among which approximately 22% are weather-related [4]. Indeed, according to the US Department of Transportation’s Federal Highway Administration (FHA) (e.g., [1, 5]), foggy weather is the most dangerous type of weather condition for motorists. It reduces visibility; distorts drivers’ perceptions of depth, distance, and speed; and it alters their driving behavior. Some research works (e.g., [5, 6]) have revealed that the majority of fog-related vehicle crashes tend to occur in rural areas in the overnight (midnight to 5:59 AM) when lighting is usually limited and visibility is already compromised. Fog also deprives motorists from using high light beams under low visibility conditions as the reflected light (backscatter) from suspended water droplets forms a glare that further blurs the driver’s visibility.

Several research works (e.g., [1, 5, 7, 8]) have revealed that although the number of reported car accidents related to stressful foggy weather conditions is not substantial, the number of associated injuries and fatalities as well as the number of colliding vehicles are much higher than average. In order to deal with these considerable impacts of fog, emerging sensor technologies are being employed in traditional roadway systems to provide real-time traffic services to drivers through Telematics and Intelligent Transport Systems (ITSs). These systems are defined as in-vehicle systems that offer active safety (i.e., fog warning messages) and infotainment services as well as location and traffic information via wireless communication technologies [9]. Several on-board devices, such as Adaptive Cruise Control (ACC) systems, are being used. However, because of their high costs, their widespread adoption is still limited. As alternative, on-board Autonomous Emergency Braking (AEB) systems are being deployed. These systems, which are integrating environment-recognition sensors (e.g., short- and long-range radars, Lidar, and cameras) to activate automatic brake control in risky conditions, are commonly ineffective (accuracy drops considerably) particularly under severe weather conditions (such as dense fog situations). Self-driving technologies and collision mitigation systems (e.g., [10, 11]) are attracting increasing attention. However, they still need advanced predictive capabilities as they are currently reactive by nature. In fact, they can best reduce the force of crash impact under foggy conditions, without allowing enough maneuver time in mitigating the risk. Solutions relying on mounted cameras can become quickly blinded due to fog or low-visibility weather conditions.

The abovementioned shortcomings have attracted researchers to explore cooperative and proactive collision avoidance systems based on Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) communications. These systems are basically allowing vehicles to talk to each other and make collaborative timely decisions to deal with sudden events that would lead to road accident risks. They also include warning capabilities that can significantly reduce drivers’ reaction time to unexpected events. In this regard, Posner and Snyder [12] reported that a warning signal with a lead time of 200 ms would reduce the reaction time of the driver by nearly 50 ms. Based on this statement, we aim in this paper to demonstrate how connected vehicular systems can increase driver awareness and reduce the likelihood and severity of fog-related crashes on a highway through V2V and V2I communications.

We are proposing a Wireless Fog Warning System (WFWS) which transmits warning messages at variable frequencies under the following conditions: (1) no fog, (2) visibility limited to 100 m, and (3) visibility limited to 50 m. In addition to studying the impact of a WFWS under different visibility conditions, we investigated its impact for different acceleration/deceleration parameters. Compared to the existing literature on WFWS, our main contribution is a new solution supporting acceleration/deceleration vehicular profiles under various foggy conditions. This solution is integrated in a Wireless Fog Warning System that disseminates warning messages to neighboring vehicles in order to prevent possible crashes.

The remainder of this paper is structured as follows: Section 2 outlines the existing works that have proposed WFWS solutions. Section 3 elaborates on the fundamentals of our proposed solution and describes the design of our new WFWS. Section 4 presents the implementation in computer simulation and describes experiment setups for performance evaluation. Section 5 presents our results. Section 6 concludes the paper and discusses potential future work.

Related work

In order to deal with the persistent road traffic problems caused by fog, several fog recognition approaches have been proposed. In this context, Mori et al. [13] have described a method to compute fog density using multiple in-vehicle sensors. The approach is based on the usage of in-vehicle cameras and millimeter-wave radar devices. Hautiére et al. [14] have presented a technique to measure visibility distances under daytime foggy weather conditions using a camera mounted onboard a moving vehicle. Ronald et al. [15] have proposed the usage of LED-based optical transceivers as fog detectors by exploiting the backscattered property of emitted light. Other traffic management and incident detection and avoidance systems (e.g., [16]) are using variable speed limit boards depending on weather conditions. Furthermore, several systems for traffic crash avoidance are deploying additional techniques, including Forward Collision Avoidance Systems (FCAS) [17], Reverse collision warning systems [18], and Adaptive Cruise Control Systems (ACC). The ACC technique, which is available on recent vehicles, is based on maintaining automatically records of distances between the current vehicle and neighboring vehicles. FCASs are being used to provide a positive and beneficial influence towards the reduction of potential crashes. However, it is still critical for such systems to collectively identify the other vehicles in the host vehicle’s path. On-board vehicular FCW system use RADAR (Radio Detection and Ranging) [19] to measure the distance from the vehicles moving in front. This is, in fact, due to the FCW system’s ability to estimate the relative inter-vehicular motion path (i.e., range, relative speed, radius-of-curvature) and the system’s ability to predict the mutual intersection of these motion paths.

All the abovementioned systems are not as accurate and reliable in low visibility conditions. Even at perfect visibility, the in-path target identification and selection problem is technically very complicated and challenging in such systems. There was always a need to design more accurate wireless warning systems based on efficient algorithm for smoother vehicular acceleration/deceleration in order to apply on-time smooth brakes to avoid the sudden crashes.

Towards a wireless Fog Warning System

Without loss of generality, our WFWS prototype is based on IEEE 802.11p WLAN as a PHY/MAC technology. It employs BSM-like messages (that is, messages comparable to the periodic Basic Safety Messages specified in SAE J2945.1) that are generated periodically (e.g., at 1 Hz) by every vehicle. These awareness messages are broadcast on the DSRC/WAVE Control Channel (CCH), that is, using a 10-MHz wide channel on WLAN channel 178 (5.890 GHz). While the simple Fog Warning System was designed to emit such messages itself, it should be pointed out that future C-ITS systems are likely to use very similar messages for a wide range of purposes (many IEEE DSRC/WAVE system designs rely on such BSMs and European ETSI ITS-G5 systems mandating the sending of very similar messages called Cooperative Awareness Messages, CAMs). Of particular interest are two data fields which transmit the following information of a vehicle: (1) Vehicle’s current position and (2) Vehicle’s current speed and direction. These messages can be used by any vehicle in the vicinity to detect the sender’s presence (regardless of the visibility conditions, e.g., fog).

A vehicle that receives an awareness message can use its content to adapt its speed, if needed. As a proof of concept, the following simple algorithm is employed on message reception:

  • Ignore this message if the sending vehicle is not in front.

  • Ignore this message if the sending vehicle is not going slower.

  • Ignore this message if another message yielded a lower time to collision t.

  • Ignore this message if linearly extrapolating own position and speed vs. the sending vehicle’s position and speed yields a time to collision tc of more than 10 s.

Otherwise, decelerate to the sending vehicle’s speed, but limit deceleration by calculating the maximum deceleration as

$$ {a}_{\mathrm{max}}=\left(10\frac{m}{s^2}-1\frac{m}{s^3}\times {t}_{\mathrm{c}}\right). $$

Our algorithm thus yields a smooth deceleration profile, starting with 0 m/s2 when the sender is more than 10 s away from the lead vehicle and linearly increasing maximum deceleration (if needed, to the physical limit of modern vehicles) as the time to collision decreases.

Implementation and simulation experiments

The described algorithm was evaluated through computer simulations using Veins 4.6 [20], a popular open-source vehicular network simulation framework. Veins employs the SUMO road traffic simulator for modeling the movement of cars on streets. SUMO is a time-discrete microscopic simulator that applies well-known car following models such as the Intelligent Driver Model (IDM) or Krauss models and well-known lane change models such as MOBIL to an agent-based representation of individual cars.

At their core, traditional car following models are designed to result in traffic patterns that are inherently collision free. For this, individual agents are assumed to be controlled by ideal, fully predictable drivers. Roads are annotated with properties such as maximum speed or right-of-way and drivers are assumed to always obey these properties. Conceptually, drivers are assumed to rigidly follow traffic rules without lapses in judgment and without impairments (e.g., due to bad visibility conditions).

For the envisioned study of a fog warning system, however, these restrictions had to be relaxed: SUMO simulator [21], version 0.30.0, needed to be extended because the driver model implemented therein are operating on assumptions of perfect visibility and error-free operation. These hard-coded assumptions needed to be modified. The microscopic mobility model had to be extended with notions of driver impairment. To keep the model of driver impairment generic as possible, a new property of roads was introduced that reflects visibility conditions on this road as the distance at which other vehicles can still be perceived. Any vehicle further away than this visibility distance is simply ignored in the context of the car following model.

The SUMO simulator was modified to include these changes. In detail, the road network format was extended with a new visibility distance parameter and this property introduced into both the toolchain supplying the road network (that is, the network importer) and into the toolchain using the road network (that is, the microsimulation model). The car following meta-model was amended to ignore any vehicle further than visibility in meters away during car following. As this results in cars suddenly “appearing” in model, which causes SUMO to adjust the error on its own and to “fix” the speed of the following car instantaneously, a further change was introduced into the car following model to limit the maximum deceleration to a configured emergency deceleration value. This modified SUMO simulator could then be used as a basis for simulating various scenarios.

Without loss of generality, we are assuming that a number of vehicles are traveling on a given highway section as follows: seven vehicles drive at a speed of 26 m/s (approx. 130 km/h) with an initial time headway of 3 s. A lead vehicle drives at a much lower speed of 22 m/s (approx. 80 km/h), with an initial headway of 10 s (that is, 220 m) in front of the other cars. Table 1 summarizes these parameters.

Table 1 Vehicle configuration

Simulations are conducted at a time step length of 0.1 s. Vehicle movement is modeled using the IDM (Intelligent Driver Model) car following model [22], (patched, as mentioned, to encompass visibility impairments) with the following parameters: tau 1, delta 1, vehicle length 5 m, minimum gap 2 m, desired acceleration 2.6 m/s2, desired deceleration 4.5 m/s2, and emergency deceleration 9 m/s2.

For studying the simple fog warning application, visibility conditions on the road are varied between three different scenarios. Poor visibility can be classified as follows: fog is usually defined as a state of atmospheric obscurity in which visibility falls below 1 km. Atmospheric obscurity with visibility greater than 1 km is known as mist. If the visibility drops below 200 m, the fog is quantified as being thick fog and if it falls below 40 m it is described as dense fog [23]. As we seek to study the acceleration/deceleration profiles of a vehicle under such different visibility conditions, so configured the following three scenarios: no fog, visibility limited to 100 m, and visibility limited to 50 m.

Each car is (optionally) running the aforementioned simple fog warning system, sending awareness messages at fixed intervals. The fog warning application was configured to send messages with 88 bit header and 256 bit payload at user priority 7. The physical layer was configured to employ a modulation and coding scheme for 6 Mbit/s and transmit at a power of 20 mW.


In order to avoid introducing artifacts, all results presented in the paper as time series plots show how the respective metric has evolved over time in a single, representative simulation run. Fundamental observations and generalizations have been cross-validated against 50 independently seeded replications of simulations. A variety of metrics were explored for each of the configured scenarios. In order to remove any ambiguity on the use of some concepts, we briefly define in what follows some terms that will be recurrent throughout the remainder of the paper.

  1. a.

    Location: It refers to a particular place/location on the highway of a vehicle at a given time. In the results section, we will study this issue using Time vs Position plot, where we will define the time each vehicle will take to cover 1.5-km distance with respect to the position of each other vehicle on the highway.

  2. b.

    Velocity: The velocity of an object is the rate of change of its position with respect to a reference point in simulation we studied this parameter using Time vs Acceleration plot, where it defines the time each vehicle will take to Accelerate/Decelerate in order to keep the safe distance from each other.

  3. c.

    Relative Position: It is a point defined with reference to another position. In our case, we study the behavior of vehicles using Time vs Relative Position plot, where this metric will define the Distance between the Lead Vehicle and following Vehicles with respect to time.

  4. d.

    Speed: This parameter will tell us the rate at which an object moves, in order to study the acceleration/deceleration profile of vehicles. We analyze the timely response of following vehicles when they approach the lead vehicle. To adjust their speed smoothly and well before the crash point, we will study this parameter using a Time vs Speed plot.

The following figures illustrate how these metrics developed over time for the first 68 s of each simulation. Figures are arranged in a grid: different visibility conditions are arranged as columns of the grid while different Fog Warning System configurations are arranged as rows.

Figure 1 illustrates how each of the eight vehicles position evolved over time. As can be seen, each scenario starts with only the lead vehicle (car 0) on the road. Its position changes commensurate to it traveling at a fixed speed of 22 m/s, so that it reaches the end of the highway after approximately 68 s in each scenario. Ten seconds later, the first of the fast cars enters the highway under investigation, followed by six more cars at intervals of 3 s. These cars quickly close the gap to the first car and at the end of each simulation, the cars travel in a tight pack of roughly equal headway. The differences between scenarios can be explained as differences in the shape of the time/position trace: with the fog warning application turned off (first row), increasingly bad visibility conditions cause the following cars to start braking suddenly near to crash point. With the fog warning application turned on (second and third row), we observed much smoother braking profiles for all the vehicles.

Fig. 1

Time/position trace of simulated cars

To further investigate the impact of the fog warning application in more detail, we turn towards more fine-grained metrics. Figure 2 illustrates what the presence of visibility impairments means for the acceleration profile of cars on the highway. Only moderate braking can be observed on a highway with no visibility impairments (first column): cars are able to smoothly adapt their speed to the slow vehicle in front. As a side effect, enabling the fog warning application in scenarios without visibility impairments causes no discernible change in vehicle behavior (as desirable). Things change when visibility impairments are introduced (second column): The first of the fast cars moving at 130 km/h can be seen to brake sharply and suddenly and without warning (first row) when it reaches close to lead vehicle moving at 80 km/h in front resulting in crash situation. Enabling the fog warning application (second and third row) can be seen to substantially reduce the severity of deceleration. The same effect is even more pronounced if visibility impairments are further increased (third column). More so, without a fog warning application, the simulated drivers are so late in starting to brake that they seem to trigger a cascade reaction.

Fig. 2

Acceleration profile of simulated cars

To investigate this effect in more detail, we turn towards a dedicated metric which is illustrated in Fig. 3: changes in the relative position of cars (that is, the distance between a car and the leading vehicle) with respect to time. As can be seen, when driving in heavily impaired visibility conditions (third column) without a fog warning application (first row), even emergency braking cannot keep the distance between cars from falling below a safe following distance. Only after finishing the braking maneuver are cars able to re-gain a safe following distance. This effect all but disappears if the simple fog warning application is switched on, though the gaps cannot be made to close as smoothly as they would without the presence of visibility impairments.

Fig. 3

Relative position (distance to leader) of simulated cars

This can be seen in more detail by investigating the speed profile of cars in the scenarios under consideration. Figure 4 illustrates this speed profile of simulated cars. As expected, car 0 (the lead vehicle) regardless of visibility conditions can be seen to drive, unhindered, at a constant speed of 22 m/s in all scenarios. To avoid a collision with the lead car, all other cars (1–7) have to adapt and decelerate in reaction to vehicle 0 (lead vehicle), as soon as it is visible to the drivers. If visibility permits (first column), this transition can be seen to be very smooth throughout the whole convoy. Worse visibility conditions (second column) cause the transition to become less smooth: cars brake later, but more sharply. Worst visibility conditions (third column) even require cars to over-compensate for the late onset of braking: for a while, their speed dips noticeable below the speed of the lead vehicle in order to allow them to re-gain a safe following distance. This effect can be seen to be cumulative through the convoy: while the first vehicle only has to slightly over-compensate, the last vehicle of the convoy has to over-compensate substantially. While turning on the simple fog warning application (second and third row) cannot completely alleviate the effect of over-compensation, it can be seen to be successful at reducing it greatly and—more importantly—alleviating the effect of worsening over-compensation with increasing convoy length.

Fig. 4

Speed profile of simulated cars

Conclusion and future work

We advocated in this paper the implementation of a Wireless Fog Warning System (WFWS) for improving deceleration profiles under visibility limitations. Our simulation results demonstrated that a WFWS can, to a large degree, compensate for the negative effects of visibility impairments by adapting vehicles’ speeds and deceleration profile in response to received position/speed periodical broadcasts, thus both increasing the safety of drivers as well as smoothing traffic flow. In our future works, we aim to expand our study by cross-validation against field operational tests as well as implementing different scenarios, such as Fog Warning V2V and V2I, Safety Distance Warning, and Lane Change Warning, respectively.


  1. 1.

    B. Hamilon, B. Tefft, L. Arnold and J Grabowski (2014) Hidden highways: fog and traffic crashes on America’s roads. 2014 AAA Foundation for Traffic Safety Report Pp. 1–24

  2. 2.

    European Commission (2017) 2016 road safety statistics: what is behind the figures, available at:

  3. 3.

    Press Information Bureau Government of India Ministry of Road Transport & Highways, available at:

  4. 4.

    US Department of Transportation Federal Highway Administration (2018) How do weather events impact roads, Available at:

  5. 5.

    Abdel-Aty M, Ekram A-A, Huang H, Choi K (2011) A study on crashes related to visibility obstruction due to fog and smoke. Accid Anal Prev 43:1730–1737

    Article  Google Scholar 

  6. 6.

    Kang D, Park D, Cho S (2008) A new fog detection technology by using a current (pA~nA) path on the interior windshield of a vehicle and application, SAE Technical Paper 2008-01-0923

  7. 7.

    CHO H-J, KIM K-S (2005) Development of hazardous road fog index and its application. J East Asia Soc Transp Stud 6:3357–3371

    Google Scholar 

  8. 8.

    Creech MF (1972) Fog: a review of the literature pertaining to highway problems and possible solutions, Virginia Highway Research Council Report, pp. 1–10

  9. 9.

    Jeong Ah Jang, Hyun Suk Kim, Han Byeog Cho (2011) Smart roadside system for driver assistance and safety warnings. framework and applications

  10. 10.

    Victoria Transport Policy Institute, Autonomous vehicle implementation predictions , 26 January 2018 , Available at :

  11. 11.

    C Grover, I Knight, F Okoro, I Simmons, G Couper, P Massie and B Smith Published project report on automated emergency brake systems, Available at: en.pdf

  12. 12.

    Posner MI, Synder CRR (1975) In: Rabbitt PMA, Dornic S (eds) Facilitation and inhibition in the processing of signals, attention and performance. Academic Press, London

    Google Scholar 

  13. 13.

    Mori K, Takahashi T, Ide I, Murase H, Miyahara T, Tamatsu Y, Fog density recognition by in-vehicle camera and millimeter wave radar (2007) Int J Innov Computing Inf Control ICIC Int 3(5):1173–1182

    Google Scholar 

  14. 14.

    Hautier N, Tarel JP, Lavenant J, Aubert D (2006) Automatic fog detection and estimation of visibility distance through use of an onboard camera. Mach Vis Appl 17(1):8–20

    Article  Google Scholar 

  15. 15.

    Mallant RKAM, Kos GPA (1990) An optical device for the detection of clouds and fog. Aerosol Sci Technol 13(2):196–202

    Article  Google Scholar 

  16. 16.

    Cheryl Lynn (2002) Christopher Schreiner, Ross Campbell, Published project report on reducing fog related crashes. Available at: /online_reports/pdf/03-cr2.pdf

  17. 17.

    Zador PL, Krawchuk SA, Voas RB (2000) Report on automative collision avoidance systems (ACAS) program. US Department of Transportation

  18. 18.

    Mobility and Transport Road Safety, European Commision, Available at:

  19. 19.

    Sivakumar R, Mangalam H (2014) RADAR based vehicle collision avoidance system used in four wheeler auomobile segments, International Journal of Scientific & Engineering Research, volume 5(1)

  20. 20.

    Sommer C, German R, Dressler F (2011) Bidirectionally coupled network and road traffic simulation for improved IVC analysis. IEEE Trans Mob Comput 10(1):3–15

    Article  Google Scholar 

  21. 21.

    Krajzewicz D, Erdmann J, Behrisch M, Bieker L (2012) Recent development and applications of SUMO—simulation of urban mobility. Int J Adv Syst Meas 5(3&4):128–138

    Google Scholar 

  22. 22.

    Treiber, Martin and Hennecke, Ansgar and Helbing, Dirk (2000) Microscopic simulation of congested traffic, Proceedings of Traffic and Granular Flow ‘99, Heidelberg

  23. 23.

    Meteorological Office (1969) Observers handbook. HMSO ISBN 0-11-400329-7

  24. 24.

    BonGiovanni T, Shah J (2002) Fog Sensor / ITS Integration, Final Report, FHWA-NJ-2003-031, Department of Civil & Environmental Engineering Rutgers, The State University of New Jersey, pp. 1–67

Download references

Author information



Corresponding author

Correspondence to Fatma Outay.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Outay, F., Ahmar, AUH., Kamoun, F. et al. Investigation of the impact of a wireless Fog Warning System with respect to road traffic on a highway. Pers Ubiquit Comput 23, 893–899 (2019).

Download citation


  • Total travel time
  • Time to collision
  • Dedicated short-range communication
  • Cooperative awareness messages
  • Vehicle to vehicle
  • Vehicle to infrastructure
  • Wireless Fog Warning System