Traffic Flow Sensing Using Wireless Signals

Abstract

As an essential part of the urban transportation system, precise perception of the traffic flow parameters at the traffic signal intersection ensures traffic safety and fully improves the intersection's capacity. Traditional detection methods of road traffic flow parameter can be divided into the micro and the macro. The microscopic detection methods include geomagnetic induction coil technology, aerial detection technology based on the unmanned aerial vehicles (UAV) and camera video detection technology based on the fixed scene. The macroscopic detection methods include floating car data analysis technology. All the above methods have their advantages and disadvantages. Recently, indoor location methods based on wireless signals have attracted wide attention due to their applicability and low cost. This paper extends the wireless signal indoor location method to the outdoor intersection scene for traffic flow parameter estimation. In this paper, the detection scene is constructed at the intersection based on the received signal strength indication (RSSI) ranging technology extracted from the wireless signal. We extracted the RSSI data from the wireless signals sent to the road side unit (RSU) by the vehicle nodes, calibrated the RSSI ranging model, and finally obtained the traffic flow parameters of the intersection entrance road. We measured the average speed of traffic flow through multiple simulation experiments, the trajectory of traffic flow, and the spatiotemporal map at a single intersection inlet. Finally, we obtained the queue length of the inlet lane at the intersection. The simulation results of the experiment show that the RSSI ranging positioning method based on wireless signals can accurately estimate the traffic flow parameters at the intersection, which also provides a foundation for accurately estimating the traffic flow state in the future era of the Internet of Vehicles.

1. Introduction

The intelligent transportation system (ITS) is currently the frontier research field of world transportation. Cooperative vehicle infrastructure system (CVIS) is an essential part of the intelligenttransportation system. CVIS mainly uses wireless communication andsensor detectionto realize intelligent coordination and cooperation between vehicles tovehicles(V2V) andvehicles to infrastructures (V2I), fully utilize the road network resources, improveroad networktraffic safety and efficiency, ease traffic congestion, and optimize the overallroad networkgoals [1,2]. High-precision positioning technology is one of the criticaltechnologies ofCVIS [3,4]. It is an essential guarantee for vehicles' safe driving and the key torealizing mutualperception between people and vehicles. The intelligent intersection is animportant researchdirection of CVIS. The intersection's traffic flow state can be effectivelyintroduced bycollecting the passing vehicles' positioning data. It facilitates intelligent regulation ofthe intersectionsignal lights. Also, collecting the vehicle's positioning data is beneficial tothe autonomousvehicle's formation, which improves the traffic efficiency of thefuture intersection[5]. A coordinated signal control system for urban ring roads based on avehicle infrastructureinterconnection environment is developed [6-8] to improve urban ringroads' operationalefficiency and reduce traffic delays. The system can adjust the signaltiming parameterssuch as cycle length, green segmentation, offset, etc., improving roadnetwork efficiency.The main road's signalized intersection is divided into several subgroups, andthe signalis optimized within each subset [9,10]. The model used a mixed-integerlinear programmingtechnique to ensure an optimal global solution and got the ideal trunkgreen wavebelt. In [11], the authors propose a network aided traffic steering technique in 5Gmobile networkarchitecture. The 5G mobile systems can monitor network conditions and learnwith networkdata with the machine learning algorithms. In [12], the authors use the EBFsketches methodto combine Bloom Filter with an exponential histogram to query streams in thesliding windowto identify heavy vehicles in the speed traffic streams. In [13], the authors proposean analysismethod based on the priority of the wireless sensor networks (WSNs) inHighway Trafficto get the accurate location of the node. This approach also works well forV2X networkenvironments.

The positioning problem can be divided into indoor positioning and outdoorpositioning. Atpresent, there have been a lot of researches on indoor positioning. Most indoorpositioning systemsuse the RSSI based on the Wi-Fi signal [14-16]. There are usually twoprocessing methods.According to the propagation loss model, one is to obtain the distance betweenthe receiverand the transmitter by measuring the signal strength received at the receiver.Another wayis to use the fingerprint information of the RSSI of Wi-Fi to establish themapping relationshipbetween the RSSI and the location in the current scene, and the positioningsystem istrained through the offline data [17].

Both of the methods have limitations. The limitation of the positioning method basedon thesignal propagation loss model is that the loss model cannot accurately describe thewireless signal'sattenuation in various complex scenarios. And because of the multipathinterference incomplex scenes, it is difficult to model the signal attenuation accurately. On the otherhand, theWi-Fi receiving signal strength fingerprint positioning method's limitation is thatthe previousdata collection and offline training cost is relatively high. And the scheme isonly applicableto the static environment. After the environment changes significantly, theoriginal RSSIlocation database needs to be updated [18,19].

Recently, a lot of research on positioning focused on the channel state information(CSI) ofWi-Fi. CSI is different from RSSI in that it is a channel response from the physicallayer. Itdescribes the amplitude and phase of each subcarrier in the frequency domain [20,21].We canget multiple carrier responses relative to RSSI. More importantly, CSI is notsusceptible tomultipath effects, making it possible to achieve precise positioning using CSI. Adevice freestate detection and localization algorithm based on Wi-Fi CSI and support vectormachine (SVM)are proposed [22]. An indoor fingerprint recognition system based on CSI basedon deeplearning is proposed. DeepFi can effectively reduce the positioning error [23]. In[24], theauthors convert the CSI measurements of multiple channels into a radio image,extract colorand texture features from the radio image, learn the optimized depth features fromthe imagefeatures using a deep learning network, and estimate the position of a personusing machinelearning methods. In [25], the authors use the CSI information to achievesmoke detection.A channel state information component reconstruction (CSI-CR) algorithmis proposedto estimate the angle of arrival (AoA) of the human body reflected signals ina through-wall(TTW) scenario [26]. In [27,28], the detection of indoor human activitiesis achievedusing wireless signals. In [29,30], CSI is used for achieving accurateindoor positioningwith an accuracy of decimeters or even centimetres. In [31], the authors usethe Dopplerspread in the envelope of the received signal to estimate the mobile user's speed.CSI technologycan also be used for network security protection on the Internet of Vehicles.The authorspropose a reinforcement learning algorithm to identify rogue nodes using thechannel stateinformation of the communication link between vehicle-to-vehicle andvehicle-to- infrastructureto figure out a rogue attack problem [32]. In [33-35], the authorsestimate unreliablechannel information for capture-aware identification of mobile RFID tags, forthe sakeof traffic flow sensing in the future.

Compared with indoor positioning, the typical traffic scene is an environment thatis relativelyopen outdoors. The outdoor multipath effect is relatively small, but the spatialrange isrelatively wide. The signal is mainly affected by the increase of the distance andthe attenuationduring the propagation process. At present, most of the research is focusedon indoorpositioning, and outdoor positioning technology mainly relies on globalsatellite positioningtechnology and base station positioning technology [36]. The former uses atleast fourspace satellites at the same time to accurately locate objects. The latter is generallyapplied tomobile phone users. The mobile phone base station location service is also calledlocation basedservice (LBS), which acquires the location information of the mobile terminaluser throughthe network of the mobile telecommunication operator [37]. Our mainresearch scenariois a typical urban cross signal control intersection. The scene radius is within300 meters,and the Wi-Fi AP access point is set on the traffic signal. The vehicle on the laneis locatedby collecting the received signal strength RSSI value of the incoming andoutgoing vehicles,thereby deriving the traffic flow's status information.

This paper proposes a sensing method for traffic flow state information of WLANs'cross signalintersections based on the RSSI signal attenuation model. And a simulationexperiment wascarried out to verify.

In summary, the main contributions of this paper are as follows:

Firstly, we designed a positioning scheme based on the RSSI-ranging model applied toa typicalurban cross signal intersection scene. The Wi-Fi signal strength value RSSIreceived bythe vehicle node is measured. The RSSI ranging model is used to obtain the distanced betweenthe vehicle and the traffic light, which is applied to derive the traffic flowparameters.

Secondly,we built a simulation platform for the Internet of Vehicles throughsumo, OMNeT++,and veins. First, we built a simulation scenario of a typical crosssignal intersection.Then we built a car-network communication environment at thesignalized intersectionand collected the signal strength indicator RSSI received by the vehicle nodeto locatethe vehicle. Finally, we derive the traffic flow parameters based on thevehicle's positioningdata.

Finally, we show the experimental results of the simulation experiment and analyzethe results.We summarized the shortcomings of this work and looked forward to futurework.

2.Related work

Traffic flow detection plays an important role in ITS. Real-time road trafficparameter collectionand accurate road congestion evaluation are the prerequisites for applyingbetter trafficcongestion avoidance strategies to improve the traffic flow [39,40]. Thetraditional trafficflow detection methods mainly include ultrasonic detection, induction coildetection andgeomagnetic detection. The magneto-electric induction detecting method, in whichloop detectorsacquire the section flow data or geomagnetic detectors placed underground ata certainroad section [41,42]. The floating vehicle method extracts traffic parameters fromthe trajectorydata by intelligent onboard devices, such as bus floating data and taxi floatingdata [43].For example, the video image detecting methods include the traffic parametercollection byelectronic-police camera or the UAV [44] and the radar detecting methods. Thefloating vehiclemethod can be applied to micro and macro traffic flow estimation. Themagneto electricinduction detecting method can only be applied to microscopic traffic flowestimation. Thevideo image detecting methods can also be divided into traditional visualdetection methodsand deep learning-based methods. For traditional visual methods, in [45], theauthors combinehistogram of oriented gradient (HOG) extracting features with SVMtraining detectionmodel as the vehicle detection method. In [46], Gaussian mixture model (GMM)is employedto model the dynamic background of the traffic scene. Then, the binaryforeground contoursare extracted by image subtraction. The real and complete vehicles are obtainedto detectand monitor the binary vehicle contours' location and the current frame. In [47],the authorsproposed a pyramid-YOLO network for detecting vehicles in dense scenes, whichcan effectivelydetect small-size and occluded vehicles. And based on the detection andcounting results,an estimation model is proposed to estimate traffic flow parameters of volume,speed anddensity. In [48], the authors collect the vehicle parameters by the radar data whosefeatures areanalyzed from the dimensions of single parameter sampling characteristics andmulti parameterrelationships. Further, the correlations of different traffic flow parameters aregiven usingthe grey correlation analysis method.

This paper proposes a traffic flow parameter estimation method based on wirelesssignal RSSIdata to extend indoor positioning technology to outdoor intersection scenes.This providesa practical basis for traffic flow estimation in the IoV environment.

In conclusion, the method based on deep learning vision is more expandable andaccurate toestimate traffic flow compared with traditional visual methods. However, this methodrelies ondata set support and is also influenced by environmental factors such as illumination.The estimationmethod of traffic flow parameter based on the millimetre-wave radar is lessaffected byenvironmental factors. The device is easy to deploy and maneuverable. Theestimation methodof traffic flow parameter based on wireless signal RSSI data is suitable for thenetwork environmentof the Internet of Vehicles under multi-scale conditions. This can provide abasis fortraffic flow estimation in the IoV environment.

3. System Architecture and Problem Formulation

Our main work is shown in Fig. 1. First of all, we use the SUMO, OMNeT++, and Veinsto builda network simulation platform, which realizes the construction of the single pointof intersectionand the configuration of the network environment. Then we collect thevehicle originalRSSI trajectory data required by the simulation experiment and process itwith Gaussianfilter. Then we calibrate the parameters of the ranging model. Finally, wecollect RSSItime track data, set the simulation step size, and then obtain the vehicle'sspace-time trackmap to calculate the traffic flow parameters.

Fig. 1.System flowchart.

3.1 THE RSSI RANGING MODEL

In an actual outdoor environment, the intersection of the cross signal is not as small asthe indoorspace. The radio waves do not have too much refraction and reflection duringthe transmissionprocess. Instead, they are on the sides of the road, tall buildings, orapartments. Andthe multipath effect is caused by ground reflections, but there is not muchshadow obscurationin the lane. Besides, it should be accompanied by large-scale andmedium-scale propagationfading of electric waves. We choose the logarithmic distanceattenuation model[16]to characterize the Wi-Fi signal's attenuation at the intersection.

\(\operatorname{RSSI}(d)=\operatorname{RSSI}\left(d_{0}\right)+10 \operatorname{nlog} \frac{d}{d_{0}}+X_{0}\)         (1)

Where d is the interval between the wireless signal receiving terminal and the transmitting terminal, is expressed as the distance of the reference point, generally selected as 1m. n is a fixed value coefficient determined by the surrounding environment, indicating the degree to which the radio wave's attenuation increases with the distance d, and n is the signal attenuation factor.X₀is a Gaussian random distribution whose variance is σ, where σ is 3 to 14 dBm.

We can get a ranging model based on RSSI signal attenuation.

\(\mathrm{d}=10^{\frac{A_{0}-R S S I}{10 n}}\)         (2)

Therefore, according to the RSSI ranging model, we can calibrate the RSSIranging modelparameters by collecting the corresponding RSSI value and the distance d betweenthe correspondingRSU and the vehicle receiving node in the simulation experiment. Thefirst calibratedparameter is K: environmental attenuation factor, and\(A_{0}\)is RSSI value whenthe referencedistance is 1 meter.

3.2 BUILDING A SIMULATIONPLATFORM

3.2.1Simulation experiment research scene

The research scene is mainly the cross signal intersection. The experimental studyscenario isshown in Fig. 2. We deploy the roadside unit RSU on the traffic signal at thecross-signal intersection,and the roadside unit RSU emits radio waves outward at the centre frequencyof 5.89GHz. The vehicle node enters the cross-signal intersection and receives thewireless signalfrom the roadside unit RSU. We collect the RSSI data received by each vehiclenode persecond. We can obtain the RSSI trajectory data of the cross-signal intersection'sentrance laneto obtain the microscopic traffic parameters of the intersection (traffic flow rate,speed, anddensity). The research scene is a typical single-point crossroad intersection scene.This typeof scene is mainly the most common single-point intersection scene in the city.Such scenesare prevalent in small cities and large cities, so they have the research's significance.

Fig. 2. Simulation scene.

3.2.2 Establish a signal intersection model

In [38], the authors propose a model for simulating real-world transportation networksby representingthe irregular road networks with static and dynamic attributes. We use theabove methodto construct the simulation model of regular cross-shaped intersections. First,we mainlyestablish and use a single-point cross signal intersection model and then builda complexdouble cross signal intersection model. For common intersections, the distributionof trafficflow is different at simple intersections. The traffic flow distribution ofcommon intersectionsis shown in Table 1.

Table 1. The traffic flow distribution

(1) Single-point cross signal intersection traffic signal timing setting

From the above traffic flow distribution table, the optimum signal period C canbe calculated.The timing scheme is designed as follows.n>

We first set the yellow light time A to 3 seconds, the start loss to 3 seconds, and thefull redtime to 2 seconds.

Since the four import lanes have left more than 200 vehicles per hour, they are all setto turnleft. Therefore, the phase of the signal light is controlled by four phases.Simultaneously, thesaturated flow rate S of each lane is set to 1400 vehicles per hour.

The key flow rate ratios of the four phases calculated by (3) are 0.237, 0.118, 0.178,0.118. Therefore,the total critical flow rate ratio is Y = 0.65 and less than 0.9, indicating thatthe settingis reasonable.

\(y_{i}=\frac{Q_{i}}{S}\)         (3)

Wherey_i representsthe critical flow rate ratio of the ith phase,Q_irepresents the phase critical flow, and 𝑆represents the saturated flow rate of the lane.

And then, we can calculate the total loss duration L = 20 seconds by (4)

\(\mathrm{L}=\sum_{k} l+I-A\)         (4)

Where L is the total loss duration, l is the start loss, I is the green light interval, and A is the yellow light duration.

Then we calculate the optimal period is C = 100 seconds by (5)

\(\mathrm{C}=\frac{1.5 L+5}{1-Y}\)        (5)

The total effective green timeG_E= 80 seconds is calculated by (6). The effective green time for each phase calculated by (7) is 35s, 15s, 15s, 15s.

\(G_{E}=C-L\)         (6)

WhereG_Eis the total effective green time.

\(g_{i e}=G_{E} \cdot \frac{y_{i}}{Y}\)         (7)

Whereg_ieis the effective green time of the ith phase.

The actual display green time of each phase is calculated by (8), and the start loss is exactly equal to the yellow light duration, so the effective green time is equal to the actual display green time.

\(g_{i}=g_{i e}+l_{i}-A\)        (8)

Whereg_ieis the actual green light display time of the ith phase.

(2) Single point cross signal intersectio n environment configuration

In the actual cross signal intersection scene, there are vehicles in the lane and buildingson bothsides of the lane, which causes multipath effects such as radio waves to occurduring transmission.Therefore, when the sumo simulation software models the commonintersection scene,it is necessary to establish the intersection model and establish the architecturalmodel aroundthe intersection. In the sumo software, the polygon.xml file is used to edit therequired buildingtype, shape, size, and position. The building model established by the experimenthas 52symmetrical forms. The actual single point cross signal intersection is as shown inFig. 3

Fig. 3. Signal intersection model.

3.3 BUILDING A SIMULATION PLATFORM

We mainly realize real-time communication of network nodes through the followingseveral modules.The Annotation Manager module mainly implements simulation visualization.The obstaclecontrol module is mainly used to realize signal attenuation caused by radiowaves encounteringobstacles during transmission. The shadow barrier control module in Veinsis providedby the environment's model around the intersection of the cross signal establishedby sumo.The Base World Utility module provides the simulation scenario's size fornetwork simulationin the veins simulation. The Connection Manager module can determine therange ofcommunication and the interference range of nodes in the wireless network according tothe powerP_tof the signal transmitting terminal, the wavelength 𝜆 of the wireless signal, thepath losscoefficient n in the signal propagation loss model, and the minimum received power P_(r)of the signal receiving terminal. The RSU module mainly realizes setting the roadside unit RSU in the intersection model and establishing a network communication model. Thespecific designand composition of the RSU module are shown in Fig. 4.

Fig. 4. RSU model.

The RSU comprises three modules: appl (RSU network application layer), nic, andmobility. Thenic module uses the nic module based on the IEEE802.11P protocol in the vehiclenetwork architecture,as shown inFig. 5. The nic module's physical layer adopts thephy802.11p communicationmodule, while the MAC layer uses the communication module basedon IEEE1609.4.ThePhy802.11p module uses a bidirectional connection, first connected tothe antennamodule to obtain messages sent from the antenna. This is followed by anopposite connectionto the MAC1609.4 module. The MAC layer is also responsible forcommunicating withthe appl (network application layer) as the nic module's upper module. Themobility moduleis independent of the other two modules in the RSU module. The mobilitymodule mainlysets the starting position of the RSU and mobility characteristics. In this article,the locationof the RSU is set at the east entrance of the intersection.

Fig. 5. V2X wireless communication model.

The vehicle nodes are constantly moving, so the Internet of Vehicles' networktopology isconstantly changing. The wireless communication methods between nodes are asfollows: First,the application layer of the RSU node generates Beacon type data afterinitialization activationand then periodically sends a message to itself (RSU). The applicationlayer processesany messages sent to itself to achieve the purpose of timed loops. Theapplication layerthen transmits the beacon message to the MAC layer. After receiving the beaconmessage, theMAC layer encapsulates it into a MAC frame. It then uses the CSMA/CA mechanismfor channelcontention to obtain a nextmacevent, the MAC layer converts the transceiver modeof thePHY layer. The physical layer then sends a control message to the MAC layer, andthen theMAC layer sends the beacon message to the physical layer. After receiving thebeacon framesent by the MAC layer, the PHY layer encapsulates it into a data frame (Airframe)of thePHY layer and then sends the frame to the channel; then broadcasts the Airframedata frameto the antenna. After receiving the Airframe sent by the RSU, the basic physicallayer inthe vehicle node module processes the frame through the handle airframe function.The handleairframe function mainly uses three states for message processing. The firststate mainlyuses the filter signal function to obtain the signal's attenuation value by thelogarithmic distanceattenuation model. The second state then obtains the received signal powerthrough theprocess signal function and determines whether to accept the signal data pack. Thelast thirdstate is to use sendup to transfer the data up to the MAC layer. The mac layer inthe vehiclenode module receives the data sent by the PHY layer. It then sends it up tothe applicationlayer after unpacking, thereby completing the communication process betweenthe RSUnode and the mobile vehicle node. The vehicle node RSSI is acquired in thephysical layer.

4. SIMULATION RESULTS ANDDISCUSSION

4.1Calibration measuring distance model

We choose the log path shadow attenuation model as the ranging model. Thecarrier's centrefrequency needs to be set to 5.89Ghz, and the environmental impact factor n is setto 3.0.In the sumo simulation, we used a car to calibrate the west entrance parametersthrough theintersection.

4.1.1 Preprocess raw data

First, multiple original RSSI data at different distances are collected in thesimulation, andthen the original data is preprocessed. Multiple measurements of the RSSI at thesame locationwill have Gaussian random interference. The Gaussian filter is used to select theRSSI valueof the high probability range, and then the average is used to reduce the Gaussianrandom interference.

4.1.2 Calibration model parameters

The position d of the vehicle node collected in the simulation process andthe correspondingRSSI data are plotted by MATLAB as shown inFig. 6. This showsthat therelationship between distance d and RSSI is logarithmic attenuation, which isin linewith experimental expectations.

Fig. 6. Relationship between distance d and RSSI.

We can get (9) from the (2) so that lgd is x, the actual measured RSSI is y, andA₀ isthe RSSIwhen the distance between RSU and the mobile node of the vehicle is 1m. So the fit ofA₀and n can be obtained by linear fitting through MATLAB. The fitting curve is asshown inFig. 7.

\(RSSI = A_{0} - 10nlgd\) (9)

Fig. 7. The fitting curveofA₀and n

4.2 Detecting traffic flow parameters of straight traffic flow at the westentrance ofthe actual single-point signal intersection

The experimental scene is mainly that the phase of the signal light at the intersectionis inthe first phase, and the traffic in the east-west direction is straight through, the trafficvolume inthe west import is 900 hours per hour, and the east entrance is 600 vehicles per hour.And thevehicle travels at a constant speed. The terminal collects the Wi-Fi signal strengthreceived bythe west-imported vehicle, obtains the vehicle's distance by the ranging model, anddraws thespace-time trajectory image of each vehicle in the measurement time throughthe MATLAB,as shown inFig. 8. We use MATLAB to linearly fit it to obtain the slope valueof thecurve. Each curve's average slope was averaged to obtain a vehicle traffic flow at aconstant speedof 13.13 m/s. And the relative error was 5% compared with the actual 12.5 m/s.The averageheadway distance is 4.16s. Then, from the front of the car, the westbound directtraffic flowQ is 865 vehicles per hour, and the relative error is 3.8%. Then, the relationshipbetween theflow, velocity, and density of the traffic flow is available. The density of thestraight westboundtraffic is 66 vehicles per kilometre.

Fig. 8The space-time trajectory of each vehicle.

4.3 Detectthe queue length of left-turn traffic

Dedicated left-turn lanes appear in the queue. The west-imported left-turn trafficvolume is200 vehicles per hour, and the traffic flow is also at a constant speed. During the test'sred lighttime, collect the vehicle's RSSI value entering the special left-turn lane of thewest entranceand then obtain the vehicle's corresponding distance. The time and space diagramof eachcar drawn by MATLAB is shown inFig. 9. When the red light is cut off, thedistance betweenthe first car and the distance signal light is 13.5 meters. The last car distance signalis 42.6meters, so the distance between the first and the last one is 29.1 meters, and thedistance betweenthe vehicles when the queue is 5 meters, so you can get a queue length of 6 cars.

Fig. 9. The time and space diagram of eachcar.

4.4Discussion of the results

First, we calibrate the RSSI based ranging model. The model is transformed fromthe logarithmicdecay model. As shown inFig. 6, we obtained the distance relationshipbetween vehiclewireless signal RSSI data and RSU through measurement. The relationshipis logarithmicdecay. By fitting the logarithmic decay function, we obtained the parametersof theRSSI ranging model A₀= -38.3361 dBm, 𝑛 = 2.8876. This fits the assumptions ofthe simulationscenario. Then we set the average speed of the traffic flow as v = 12.5m/s. Therefore,we measured the spatiotemporal trajectory diagram of the inlet channel, and the blue pointis thediscretized distance point obtained by the RSSI ranging model. We fit the blue dot toget thespace-time curve of each car. The slope of each space-time line is the average speed ofthe vehicle.The average speed of the vehicle is 13.13m/s, and the relative error is 5%. Thismethod hasgood measurement accuracy.Fig. 9is the continuous space-time graph obtainedfrom discretepoints. The average direct width between each space-time curve is calculated, andthe averagetime headway is 4.16s. Therefore, the experimental calculation shows that thequeue lengthof vehicles on the entrance road during the red light period is 6 vehicles. To sumup, thismethod can better measure the queue length during the intersection red light.

5. CONCLUSIONS

Based on the ranging model based on the RSSI logarithmic distance attenuation model,we studythe traffic state sensing method based on V2X wireless signals and validate thescheme's feasibilityand accuracy through simulation. The main work of this paper is summarizedas follows:

(1) By analyzing the wireless signal's outdoor propagation environment at thesingle-point intersection,combined with the transmission mode of the radio wave and the mainfactors affectingthe attenuation, the wireless signal attenuation model is required for theexperiment isdetermined as a logarithmic path attenuation model. An RSSI-based ranging modelis obtained.

(2) The experiment is carried out by the simulation method. Firstly, the model ofthe intersectionand its surrounding environment is established by the sumo simulationsoftware. Thenthe OMNeT++ and veins are used to build the simulation platform of the vehiclenetwork. Inthe simulation process, the original RSSI data of the imported road is collected first,and thenthe data is preprocessed by Gaussian filtering and culling abnormal data. Theparameters ofthe ranging model are then calibrated. The relative error of the parameter is3.75%.

(3)Simulation experiment to measure the speed, traffic flow, traffic density ofthe westbounddirect traffic at the common intersection, and the dedicated left-turn lane'squeue lengthduring the red light. The measurement results are close to the simulationsetup's parameters,indicating the method feasibility and accuracy.

Our approach is done through simulation experiments, so there will certainly besome deviationsfrom the actual case. We also take this issue into consideration. So we willexplore thisproblem further in the next study. Our next step is to extend the simulationenvironment topractical application scenarios.