In Indian four-lane express highway, millions of vehicles are travelling every day. Accidents are unfortunate and frequently occurring in these highways causing deaths, increase in death toll, and damage to infrastructure. A mechanism is required to avoid such road accidents at the maximum to reduce the death toll. An Emergency Situation Prediction Mechanism, a novel and proactive approach, is proposed in this paper for achieving the best of Intelligent Transportation System using Vehicular Ad Hoc Network. ESPM intends to predict the possibility of occurrence of an accident in an Indian four-lane express highway. In ESPM, the emergency situation prediction is done by the Road Side Unit based on (i) the Status Report sent by the vehicles in the range of RSU and (ii) the road traffic flow analysis done by the RSU. Once the emergency situation or accident is predicted in advance, an Emergency Warning Message is constructed and disseminated to all vehicles in the area of RSU to alert and prevent the vehicles from accidents. ESPM performs well in emergency situation prediction in advance to the occurrence of an accident. ESPM predicts the emergency situation within 0.20 seconds which is comparatively less than the statistical value. The prediction accuracy of ESPM against vehicle density is found better in different traffic scenarios.
This work is motivated by the statistics of increased highway accidents and rise in death toll every day. In India, 1, 37,423 people were killed and 4, 69,900 people were injured in road traffic crashes in 2013 (NCRB 2013). Traffic fatalities increased by about 54.3 percent during the period 2003–2013 [
Road traffic accidents have a high impact on the economy of the developing countries and affect the development of those countries [
The objective of this paper is to design a framework to implement Intelligent Transportation System (ITS) using Vehicular Ad Hoc Network (VANET) with an intention to reduce the number of road accidents and to decrease the death toll caused by road accidents. An Emergency Situation Prediction Mechanism (ESPM), a proactive framework for accident prediction and prevention, is proposed to meet the objective. ESPM is designed for a four-lane express highway of Indian traffic system. In the existing Indian four-lane express highway the vehicles move in two lanes in left-hand side in one direction against the two lanes of the other side as shown in Figure
Indian four-lane express highway.
ESPM attempts to predict the possibility of occurrence of an accident in advance before it occurs. An accident is said as an emergency situation in this paper. The Road Side Unit (RSU), the vehicles in the range of the RSU, and the four-lane express highway are considered in ESPM for emergency situation prediction and prevention.
Once the vehicles enter into the range of RSU, periodically the vehicles send a Status Report (SR) to RSU simultaneously the RSU performs traffic flow (TF) analysis. A Four-Lane Sensor Grid (FLSG) is designed to support ESPM for traffic flow analysis. ESPM performs prediction based on both SR and TF values. As the prediction process involves both the SR and TF values, the prediction accuracy (PA) is promisingly increased.
The highway road accidents can be categorized as head-on accidents, lateral impact, and rear-end accidents [
The rest of the paper is organized as follows. In Section
Intelligent Transportation System (ITS) is an emerging area of research over the past two decades. Transportation system becomes intelligent with emerging computing hardware, positioning system, sensor technologies, high speed telecommunication, and data processing.
ITS makes the travels of ease and comfort, it also provides reliable and improved travel safety. The environment gets benefit out of ITS because of low carbon emissions and air pollutions. Due to improvement in traffic efficiency and reduction of accident rate the economy of countries is improved [
VANET is a technology that uses moving vehicles as nodes and forms a Mobile Network. Vehicles communicate with each other via Vehicle to Vehicle communication (V2V) and with roadside base stations via Vehicle to Infrastructure communication (V2I) [
VANET Architecture.
ITS in VANET leads to the concept of Automated Highway System (AHS). AHS provides Safer Highway Transportation by making a vehicle to predict the actions of neighboring vehicles [
Caliendo et al. [
Hu et al. [
Sörstedt et al. [
Worrall et al. [
De Oña and Garach [
Garcia-Lozano et al. [
Rokade et al. [
Saravanan et al. [
It is learned from the literature that most of the existing models for accident prediction is based on either the vehicle parameters or the traffic parameters and most models are reactive and based on statistical data rather than real time data. Sometime there is a possibility of failure in the process of prediction because of not considering either of the above parameters. Hence in ESPM, both the parameters of vehicle status and traffic conditions are considered for emergency situation prediction process to ensure the performance.
The proposed Emergency Situation Prediction Mechanism (ESPM) is a framework for traffic accident prediction and prevention. Overall structure of ESPM is shown in Figure
ESPM framework.
ESPM performs prediction of emergency situation in three phases and prevention in the fourth phase. First, the participating vehicles will send Status Report (SR) to the near-by Road Side Unit (RSU). Second, RSU monitors the traffic flow in its range. Third, the RSU performs emergency situation prediction based on the SR and TF values and in fourth phase the RSU constructs and disseminates Emergency Warning Message (EWM) to all vehicles in its range and to the near-by RSUs. As a result the vehicles have a chance to take preventive action against accidents either by reducing the speed of the vehicle or taking an alternate route [
In this paper much attention is given to the first three phases and the fourth phase will be a future work.
The ESPM framework is broadly divided into four phases, namely, reporting, monitoring, prediction, and prevention. Each phase is assigned a distinct task. The pseudo code of ESPM is presented below followed by a detailed discussion.
Pos: position of the vehicle; Sp: speed of the vehicle; Yr: yaw rate (i.e., rotation angle); SR: Status Report; FLSG: Four-Lane Sensor Grid;
Dm: decision matrix;
RSU: Road Side Unit; Periodic _Timer: set as 100 milliseconds; Dp: decision parameter; Td: traffic density.
(1) Sleep until (1) For each vehicle ( (1) Read Pos from GPS Unit. (2) Read Sp from Speed Sensor. (3) Read Yr from Yaw-Rate Sensor. (4) Construct SR with (5) Send SR to RSU. (2) Return
(1) Sleep until (1) For each (1) Sense the presence of the Vehicle ( (1) If a vehicle is in (1) Set else (1) Set (2) Return (2) Send (2) Return
(1) Sleep until (2) For each vehicle ( (1) Analyze the Pos, Sp and Yr. (2) If any of the following case arises (1) Case (2) Case (3) Case then (1) Set Dp as Abnormal. else (1) Set Dp as Normal. (3) Return
(1) Sleep until (2) For each (1) Read (2) Construct Decision Matrix (Dm) with position values using (3) Analyze the Decision Matrix with Traffic Density (Td). (4) If (1) Case: Inner to Outer: (1) If (1) Lane change is Normal. (2) Set Dp as Normal. (2) else (1) Lane change is Abnormal. (2) Set Dp as Abnormal (2) Case: Outer to Inner (1) If (1) Lane change is Normal. (2) Set Dp as Normal. (2) else (1) Lane change is Abnormal. (2) Set Dp as Abnormal (3) Return
(1) Sleep until (2) Read Dp value in Module 1 and Module 2 (1) If Dp is Abnormal then an Emergency Situation is predicted and report for prevention. (3) Return
It is assumed that each vehicle (a car) is equipped with (i) sensors for measuring speed and yaw rate (ii) Global Positioning System (GPS) for finding the location of the vehicle and (iii) transceivers for transmitting and receiving data [
Once the vehicle enters into the range of the RSU, periodically (explained in Section
Components of Status Report (SR).
Where ID is used for the identification of the vehicle, position is the geographical position of the vehicle which is determined by the GPS device attached to the vehicle. Speed is determined by the speed sensor and yaw rate (i.e., rotation rate) is determined by yaw rate sensor. The location of the vehicle in a road lane can be calculated from the position and speed values of SR [
In this phase, the RSU performs traffic flow monitoring based on the traffic flow data. The traffic flow data can be obtained with the help of sensors embedded in the four-lane express highway called Four-Lane Sensor Grid (FLSG).
A Four-Lane Sensor Grid (FLSG) is a setup, where each lane is divided into blocks [
Organization of four-lane sensor grid (FLSG).
In FLSG, each block is denoted by
In real time, lanes 1 and 2 are fixed with travel speeds 60 km/h and 90 km/h, respectively, in Indian four-lane express highway. Lanes 1 and 2 are symmetric to lanes 4 and 3, respectively. The width of each lane in FLSG is 3.6 meters and the width of the four lanes is 17 meters including the highway divider.
The total number of blocks in FLSG can be computed using
The total number of Sensors in FLSG can be computed using
The length of blocks in each lane is assumed to be 6 meters and the minimum and maximum speed limits of the lanes are fixed to be 60 km/h and 90 km/h, respectively, in simulation. Therefore the period for sending the SR is fixed as 0.24 seconds that is the minimum period computed for the above speed limits. (The maximum period is 0.36 seconds).
In this paper, the size of FLSG is considered as 24 blocks (i.e.,
The RSU periodically monitors the traffic flow in its range. In VANET, traffic flow can be predicted using past and present traffic flow data (i.e., the location dynamics of the vehicles). Traffic flow prediction is not possible only by considering past traffic data due to on road traffic accidents, off road events, and unavailability of traffic data in all links of a traffic network because most roads/links are not equipped with traffic sensors [
RSU is programmed in such a way that it receives data from Sensys sensors [
The prediction phase is instrumental for emergency prediction based on the SR and TF data received from reporting and monitoring phases, respectively. If either SR or TF values are abnormal then RSU predicts an emergency situation (i.e., the possibility of occurrence of an accident).
Once the RSU receives the SR, it checks whether speed, position, and yaw rate are within the specified limits. If any of the following cases arise then RSU concludes that the SR is abnormal.
If the position of the vehicle is the same in two continuous time periods then the vehicle is in halt state as represented by the following equation:
If the difference in speed of the vehicle in two continuous time periods is greater than 30 km/h then the vehicle is with abnormal speed represented as
If the vehicle’s change in yaw rate exceeds 30 degrees then the vehicle has performed an abnormal lane change represented as
In (
The RSU periodically receives the traffic data (i.e., GP value) through each sensor in FLSG; it transforms the traffic data into a decision matrix (DM) for traffic flow analysis as shown in Figure
Transformation of four-lane sensor grid into decision matrix.
If all the positions of DM are filled with 1s, then it means that the highway is filled with vehicles and the Traffic Density (TD) is high. If all the positions of DM are filled with 0s then the highway is free and TD is low. Three scenarios of abnormality are identified and explained below.
In FLSG, the sensors are fixed at the entry points of all the blocks. If a vehicle enters the block, immediately the sensor senses and reports to the RSU. The RSU records this by placing a “1” in the respective position of the decision matrix. The block size of FLSG is designed in such a way that, once a vehicle enters into a block means it automatically leaves the previous block (i.e., the sensor area in the previous block). Hence, if a vehicle is partially in two blocks the preceding sensor alone reports to the RSU about the presence of the vehicle.
Once the speed of the moving vehicle decreases to 0 Km/h (i.e., deceleration to 0 Km/h) (or) the position of the moving vehicle remains the same for two continuous time periods then it is concluded that the vehicles resides in the same position (or) stops travelling.
If a set of vehicles travelling in a lane one after the other, sudden stoppage of a vehicle might cause rear end collision [
Scenario of abnormal deceleration with respective decision matrix.
A vehicle in
In DM, if a particular column is filled with continuous 1s then the probability of an emergency situation is much higher and the change in GP values of DM need to be closely monitored against abnormality.
If the moving vehicle suddenly accelerates its speed (or) the speed of the vehicle crosses the maximum lane speed limit then the vehicle might collide with the one before it (i.e., front end collision) [
In case vehicles are moving closely one after the other, if a following vehicle suddenly accelerates its speed abnormally then the following vehicle collides with the one in front (front end collision) and this might result in a chain of collisions as shown in Figure
Scenario of abnormal acceleration with respective decision matrix.
The lane change the vehicle is normal only if the following condition is satisfied. Otherwise, the lane change is abnormal:
The ESPM is designed for a four-lane express highway; the lane change is possible only between lane 1 and lane 2 in one direction and between lane 3 and lane 4 in the other direction (symmetric to the other direction). At par with expression (
Two types of lane changes are possible such as inner to outer (ITO) and outer to inner (OTI). Consider the present position of the vehicle in DM as
Abnormal inner to outer lane change with respective decision matrix.
Similarly, a smooth or normal OTI lane change is possible only if the position values (
Abnormal outer to inner lane change with respective decision matrix.
Otherwise, both the lane changes would be abnormal and this leads to an emergency situation.
If any of the above three scenarios arise then it could be concluded that there is a possibility of occurrence of an accident and the prediction is successful (i.e., an emergency situation is predicted).
This prevention phase is purely dependent on the prediction phase. Once the emergency situation is predicted by the prediction phase, then the RSU constructs an Emergency Warning Message (EWM). This EWM is then broadcasted by the RSU to all the vehicles and near-by RSUs in order to alert them about the possibility of occurrence of an accident. The near-by RSUs will broadcast this emergency message to the vehicles in their respective ranges. Upon reception of EWM, the vehicles might prevent themselves from accidents either by stopping or taking alternate lanes/routes.
A mobility scenario for highway traffic in Indian four-lane express highway has been developed using Freeway Mobility (FM) model. The FM model generates common scenarios such as stopping, lane changing, and overtaking in highways. The FM model injects vehicles in each lane at a specific traffic rate. As specified in Figure
The FM model does not allows lane change; hence (
Network Simulator (ns 2.34) is used for simulation. The mobility traces of the scenarios (i) sudden stoppage (deceleration), (ii) sudden increase in speed (acceleration), and (iii) abnormal lane change of vehicles are developed using the FM model.
The simulation is performed on two layers of nodes. To simulate the fixed RSU and FLSG sensors nodes one (fixed) layer of nodes are used. The second layer of nodes is the moving vehicles whose mobility traces are opted from the FM model. The main simulation parameters are listed in Table
Simulation parameters.
Parameter | Value |
---|---|
Highway scenario dimension | 1000 m, 2 lanes, and 1 direction |
Simulation time | 300 s |
Assumed vehicle length | 5 m |
Vehicle speed limits | 60–90 Km/h |
Average speed | 70 Km/h |
Vehicle density | 50–70 vehicles per 500 m |
Traffic rate/vehicle injection rate | (1/75, 1/60, 1/45, 1/15, 1/5) vehicles/sec/lane |
Vehicle transmission range | 500 m |
Sensing radius of sensors | 6 m |
Number of vehicles | 150 |
Number of road sensors | 200 |
Number of lanes | 2 lanes |
Distance between two sensors in a lane | 6 m |
Distance between two sensors in two neighboring lanes | 3.6 m |
Period of message exchange | 0.10 second (i.e., 100 milliseconds) |
The moving vehicles will send SR to the fixed RSU nodes once in every 100 milliseconds (lesser that the required 240 milliseconds as explained in Section
Based on both the SR and sensor data, the RSU node will analyze the abnormal behavior of vehicles and arrive at a conclusion about the possibility of occurrence of an accident (i.e., prediction). The vehicle might behave abnormally in any of the scenarios as mentioned in the previous Section
The results demonstrate the efficiency of the ESPM framework for emergency situation prediction. Prediction probability and vehicle density are selected as performance metrics for evaluating the performance of ESPM.
In order to demonstrate the accuracy of the ESPM approach the following three scenarios were simulated.
In this scenario, initially a moving vehicle (mobile node) in a lane has been forced to stop by making its speed reach 0 km/h instantly and the reaction in the system has been monitored. After receiving the SR of the respective vehicle, the RSU node has reported an abnormality within the specified time period and a successful prediction has been achieved.
The simulation was repeated for 10 runs by gradually increasing the vehicle density and randomly vehicles in the two lanes were made to stop suddenly. The reaction of both the vehicles and RSU has been monitored for accuracy and performance of ESPM. In Figure
Prediction accuracy versus vehicle density for Scenario 1.
In this scenario, among the moving vehicles (mobile nodes) in two lanes a vehicle was forced to increase its speed to cross the maximum limit of 90 km/h and the systems’ reaction has been monitored. The RSU node has reported an abnormality within the specified time period after reception of SR from the respective vehicle. Similarly the speed of the vehicles in two lanes was increased and the reaction of the RSU node has been monitored.
This simulation was repeated for 10 runs by gradually increasing the vehicle density (i.e., number of vehicles). The vehicle behavior and reaction of RSU was closely monitored for assessment of performance and the same is shown in Figure
Prediction accuracy versus vehicle density for Scenario 2.
The simulation result shows that the prediction accuracy is close and above 92 percent both abnormal behaviors (AB). Comparing both Figures
During simulation in this scenario, vehicles were made to change their lanes from inner to outer and outer to inner in different runs.
(a) A vehicle was allowed to perform a normal inner to outer (ITO) lane change and forced to perform an abnormal inner to outer lane change as explained in Section
(b) Similar to the previous case a vehicle was allowed to perform a normal outer to inner (OTI) lane change and forced to perform an abnormal outer to inner lane change as explained in Section
The prediction performance of ESPM in this scenario is promising and this is shown in Figures
Figure
Prediction accuracy versus vehicle density for Scenario
Prediction accuracy versus vehicle density for Scenario
The speed limit of inner lane is greater than the speed limit of outer lane. This leads to the increase of complexity in OTI lane change when comparing to ITO lane change. Due to this reason the prediction accuracy of OTI lane change might decrease.
A vehicle attempting for OTI lane change must increase its speed from 60 km/h range to 90 km/h range in order to meet the speed of vehicles in the inner lane. At par with (
In all the three scenarios, the simulation results show that the successful prediction achieved within 0.2 seconds. This is much better than the statistical requirements of 0.5 seconds [
In this paper, an Emergency Situation Prediction Mechanism (ESPM) is proposed to predict the possibility of occurrence of an accident in Indian four-lane express highway. The primary objective of ESPM is to predict an emergency situation in advance, thereby preventing accidents and reducing the death toll. ESPM is used to perform prediction of an emergency situation in four phases. The first three phases (reporting, monitoring and prediction phases) are used for prediction and the fourth phase is used for prevention. In ESPM, the prediction accuracy is computed against vehicle density for three different scenarios. In all the three scenarios, it is noticed that there is a small gap between the analytical evaluation and simulation results. The results show that the performance of ESPM is promisingly increased towards prediction. Prediction accuracy of ESPM against vehicle density is almost above 92 percent.
Implementation of ESPM for Indian four-lane express highways is expensive because of the development and deployment of infrastructures such as FLSG and RSU even though most of the present day vehicles are automated vehicles. But in near future ESPM has a high scope of implementation for improving ITS.
In future, attempts will be made to improve the performance of ESPM. Based on prediction of emergency situation the prevention of accidents would be done by disseminating distress beacons (i.e., Emergency Warning Message) to alert all vehicles in the range of the RSU and the vehicles in the range of the near-by RSUs to prevent from accidents.
The authors declare that there is no conflict of interests regarding the publication of this paper.