PDMS-FBG-Based Fiber Optic System for Traffic Monitoring in Urban Areas

This article introduces fiber Bragg grating (FBG) system for monitoring selected traffic flow parameters in urban areas. The system is able to monitor traffic density or speed of cars driving in single lane. The proposed system consists of five Bragg gratings encapsulated in Polydimethylsiloxane polymer and is characterized by easy and inexpensive installation in the milled upper layer of the road covered with an asphalt mixture. The results of first operational tests presented in this study include 3978 passing vehicles of various specifications. The system’s vehicle detection success rate is 99.62%. And further, a mean absolute error 1.35 kph and a relative error 2.62% when measuring vehicle speed.


I. INTRODUCTION
Traffic sensors are devices that provide input data for traffic management systems. These sensors are installed primarily on and above the roadway or within the road surface. The basic information and traffic parameters provided by the sensors include vehicle speed, vehicle classification, wheelbase and number of axles, vehicle weight, axle load of the road, stopped vehicle detection, lane or direction change detection and traffic density, which indicates the quality of traffic flow. The traffic sensors are primarily divided into two basic groups, intrusive and non-intrusive (i.e. destructive or non-destructive to the road and its surroundings).

A. STATE-OF-THE-ART SENSORS
The conventional traffic sensors include inductive loops [1], [2] and magnetic field sensors [3]. These are among the most commonly used sensors for their simplicity and reliability.
The associate editor coordinating the review of this manuscript and approving it for publication was Zinan Wang . On the other hand, they also have some disadvantages such as limited usage in reinforced concrete structures, higher costs associated with their installation in the road structures and the susceptibility to breakage due to heavy traffic.
Pneumatic sensors represent a portable technology with very easy installation. The principle of their operation is based on measuring the change in pressure in a pneumatic road tube [4], [5]. The drawback of the technology is that standing or slow-moving vehicles cannot be easily identified which causes inaccurate axle counts at higher traffic intensities. Moreover, this type of sensor is not used in adverse weather conditions, such as snow or rain.
Piezoelectric sensors [6]- [8] work on the principle of compression of a measuring probe placed in a milled in longitudinal groove. The voltage generated by the sensor is proportional to the force or weight (wheels or axles) of the passing car.
Microwave motion sensors [9], [10] uses electro-magnetic radiation; the vehicle speed is determined using micro-waves reflected back to the receiver based on the Doppler effect.
The disadvantage of these sensors is their susceptibility to electromagnetic interference, radio waves are reflected from metal parts of car body and other obstacles, such as barriers, street lighting poles, iron or concrete structures.
The principle of operation of Closed Circuit Television (CCTV) technology is based on the digitization of a static image, where the passage of the vehicle changes the values of color and brightness. This technology is based on a set of one or more interconnected video cameras and a powerful computer for data processing and analysis. Using CCTV, it is also possible to register state license plates [11] and calculate the vehicle speed using the total distance travelled and the time taken from start to finish of selected section. CCTV technology is not affected by vehicle changing lanes, because the cameras monitor all lanes and are interconnected [12], [13].

B. FIBER OPTIC AND OPTICAL SENSORS
Fiber optic and optical sensors have different principles of operation and offer several significant advantages, such as reliability, accuracy, high sensitivity, electromagnetic and chemical resistance or a wide operating temperature range. These sensors are used in many areas, such as automobile traffic monitoring.
Active infrared sensor gates work on the principle of transmitting and receiving an optical beam (laser or infrared, respectively) between the transmitter and receiver. [14], [15]. This kind of measurement is of a portable or stationary nature and is designed for speed and traffic density monitoring, or excessive trucks measurement.
Another type of sensors is based on microbending effect [16], where the attenuation of the optical fiber changes as the car passes. The fibers are stored in special protective housing and installed into the road structure. This type of sensors is rather experimental and is widely used in practice. In addition to micro-bending technologies, interferometric systems have also been used among the literature, mostly to count automobile axles [17], [18].
A special type of fiber optic sensors are Bragg gratings. To detect vehicle passage and measure vehicle speed, these sensors can be installed in the road itself (invasively) [19] or they can be placed on the road surface for short-term noninvasive use [20]. Weigh-in-Motion (WIM) sensors are other widely used sensors in traffic flow monitoring. This type of sensor is mainly based on Bragg gratings, which are installed in the road in various forms of encapsulation [21]. Portable pads based on Bragg grating are also being developed for weighing cars, but their use is limited by the maximum speed of the monitored cars [22]. In general, WIM are used in conjunction with other systems, so that the overall system can, in addition to the vehicle weight, provide information on speed, license plates, driver's appearance, vehicle dimensions, etc.
Finally, Rayleigh scattering based distributed acoustic sensing (DAS) has emerged as a powerful and rapidly adopted sensing technology in the recent years. The DAS technology evaluates Rayleigh scattering in the fiber optic cables, where acoustic activity generates acoustic energy that is absorbed by molecules in the fiber core. This causes a phase shift in backscattered light, which is detected by the evaluation unit and subsequently analyzed. The application of this technology in automobile traffic has been demonstrated in [23], [24], the authors state that it is possible to monitor more traffic parameters (such as vehicle detection, direction, speed) using a single optical fiber.

II. FIBER BRAGG GRATINGS
Fiber-optic Bragg gratings (FBGs) are a rapidly evolving group of fiber-optic single-point sensors [25]. Bragg gratings are formed by a structure with periodic variation in the index of the fiber core. In this structure, the refractive index of the core n 1 alternates with an increased refractive index n 3 = n 1 +δ n , where δ n is the induced refractive index which results from exposure of the photosensitive optical fiber to UV light. Fig. 1 shows the structure and the operational principle of the Bragg grating.
The structure of the Bragg grating reflects a narrow frequency band of the broadband light and transmits the others. The central part of the reflected spectrum is called the Bragg wavelength λ B and is given by: where n eff is the effective refractive index of the grating in the fiber core is the grating period. Other wavelengths are transmitted without attenuation. The Bragg wavelength is the geometric and optical properties that change under the influence of mechanical and thermal stresses. The relation of the Bragg wavelength change, the relative deformation and temperature is expressed by the following relation: where λ B is Bragg wavelength shift, k is deformation coefficient, ε is deformation, α is the coefficient of thermal expansion, α n is the thermo-optic coefficient and T is change of temperature [26].

III. DESIGN AND IMPLEMENTATION OF PDMS-FBG SENSOR
Road traffic monitoring using FBG sensors can be implemented using two approaches. In first, non intrusive approach, the sensors can be mounted overhead, to the side of VOLUME 8, 2020 the roadway or on the road surface non-invasively. The second, intrusive approach, requires the sensors being mounted directly into the road structure (invasively). The non-invasive installation approach is only suitable for temporary use. This approach is characterized by a simpler implementation, but the life of these sensors is reduced, and it is also necessary to remove the sensors during winter road maintenance. The intrusive approach requires cutting into the road for installation and maintenance. Although this approach is more costly, its service life is longer and it is capable of continuous operation. This article describes an intrusive system that was invasively installed in the upper layer of the road structure. Bragg gratings in a single-mode optical fiber with primary acrylate or polyimide protection have tensile strength but very susceptible to mechanical damage or breakage. For this reason, it is necessary to protect the optical fiber in sensory applications, especially when installed in the road, where the fibers must be invariant to chemical and thermal changes in concrete and also resistant to high pressure caused by from passing vehicles.
Different requirements are placed on the encapsulation of the Bragg grating. On the one hand, the sensor must be robust and resistant to the negative effects of the environment, but on the other hand it must have a sufficient sensitivity. Polymeric materials have been selected for encapsulating the Bragg gratings since they are sufficiently flexible for deformation measurements yet hard to break. Based on the initial tests and also previous experience [27]- [29]., the authors selected polydimethylsiloxane (PDMS), a polymeric two-component potting material.
PDMS is a non-toxic and non-flammable polymeric organosilicon compound that is resistant to chemicals, compressive stress, UV radiation. At the same time, it does not affect the function of FBG, see [30].

A. PDMS-FBG SENSOR INSTALLATION TO ROADWAY
For the first phase of the experiments, we used a prototype of the PDMS-FBG sensor with dimensions of 60 × 30×7 mm. The sensor was made using single-mode optical fiber G.657. A with acrylate primary protection and Bragg grating (Bragg wavelength of 1548.388 nm, reflectance 93.3%, spectral width of 198 pm (full width at half maximum, FWHM)). The FBG fiber was placed in a mold printed on a 3D printer and subsequently encapsulated with PDMS Sylgard 184 polymer. SYLGARD 184 is a silicone elastomer kit which contains two chemicals polymeric base and a curing agent which cross-links with the polymeric matrix. Curing was carried out for 90 min at 80 • C. Figure 2(a) shows a diagram illustrating the deployment of individual parts in the PDMS-FBG sensor, Figure 2(b) shows the sensor prototype. The fiber was protected by a PVC (Polyvinylchloride) tube with an outer dimension of 3 mm (PVC protection is visible in Fig. 3

(a))
The PDMS-FBG sensor was embedded across in the monitored road surface, please see Fig. 3. In the first step, a longitudinal groove (2 m long, 10 mm wide,15 mm deep) was milled into the road. The PDMS-FBG sensor was placed in a part of the groove located 1 meter from the curb that was widened to the dimensions of about 80 × 120 mm, please see Fig. 3(a). In the second step, the groove with the sensor was filled with Biolast 2K asphalt (Nadler Strasentechnik, Barleben, Germany), see Fig. 3(b). The cold asphalt curing took place at an outdoors (temperature of 23 ± 1 • C) and took 90 min.
The effect of the curing process of the selected type of cold asphalt on the PDMS-FBG sensor Bragg wavelength is shown in Figure 4. A sharp increase of 411 pm (yellow area in Figure 4) related to asphalt-pouring is followed by a gradual decline of the wavelength (200 pm in 1.5 hours) associated with the curing process (blue area). The process of installing the sensor into the road surface caused an increase in the Bragg wavelength by about 190 pm. This small Bragg wavelength change during the installation and curing process does not lead to significant spectral requirements in the design of the sensor branch with five PDMS-FBG sensors described below. The embedding of the FBGs in PDMS and asphalt curing process also does not affect the shape of the reflection spectrum, and thus the functionality of the PDMS-FBG sensor.

B. PDMS-FBG SENSOR SPATIAL SENSITIVITY
Traffic density and speed monitoring is conditioned by accurate vehicle detection. Therefore, it is necessary to ensure the sensitivity of the sensor element to car passages along the entire width of the road lane (i.e. width of approx. 3.2 m -which is an average width of the roads in the Czech cities). The designed PDMS-FBG sensor is a single-point sensor; to cover the entire lane it is necessary to  use n PDMS-FBG sensors connected in series. To determine the optimal number of sensors needed, it is necessary to know the spatial sensitivity of the PDMS-FBG sensor.
The spatial sensitivity of the sensor was determined by analyzing the magnitude of the sensor's response to the axle of the vehicle passage at various distances from the PDMS-FBG sensor. For this purpose, a grid was drawn on the road on both sides of the sensor with a spatial resolution of 5 cm (see Fig. 5). Subsequently, a test was carried out based on 642 passages of vehicles with various specifications (605 cars, 31 cars with a trailer, 6 trucks). The distance between the axle and the sensor was then determined for each vehicle using a video recording. Figure 6 shows the responses of PDMS-FBG sensor to vehicle passage depending on the axle distance of the vehicle from the sensor. The axle passes directly via the PDMS-FBG  sensor correspond to the highest response, which corresponds to an absolute value of 600 µm. As the crossing distance increases, the response decreases; for example, at a distance of 30 cm, absolute value of the response is 20 pm. This is the value that can be certainly detected by the evaluation unit since the minimum resolution is 1 pm. Based on this analysis, the maximum distance between PDMS-FBG sensors was set as 60 cm, i.e. the system is able to detect axle that passes between two sensors 60 cm apart.

C. DESIGN AND IMPLEMENTATION OF PDMS-FBG SENSOR-CHAIN FOR SINGLE-LINE TRAFFIC MONITORING
Based on the above analysis, the spatial sensitivity of the PDMS-FBG sensor for vehicle detection was defined as ±30 cm. To cover the entire width of the 3.2 m lane, a sensor-chain with five PDMS-FBG sensors was designed, see Figure 7.
In the next phase, it is also necessary to determine the spectral spacings of individual PDMS-FBG sensors. These spacings must be defined by the maximum possible sensor deflection caused by the car passage. Spacings must prevent overlapping of adjacent spectra, i.e. signal crosstalk.  In the pilot tests with a limited number of 642 passages, the maximum signal amplitude ranged from 583 pm to 631 pm. A spectral window must be assigned to each PDMS-FBG sensor and the individual spectral windows must not overlap. Individual PDMS-FBG sensors can only operate in their spectral windows, otherwise a measurement error, so-called first order crosstalk, can occur.
Although the pilot test performed involved vehicle passages of various specifications and weights, some margin is considered, which takes into account, for example, the passage of large goods vehicles (trucks), which may elicit an even greater response. Therefore, we selected the spectral windows size of 2 nm.
When modeling the sensory branch, it is necessary to consider other factors such as manufacturing tolerance of Bragg wavelength in FBG production, effect of Bragg grating encapsulation in PDMS polymer, effect of asphalt curing, uneven distribution of road temperature changes, etc. Based on analysis [31], which beyond the scope of this article, a 2 nm margin was defined. For this reason, 4 nm (2 nm for measuring and 2 nm for other factors described above) is available for each measuring channel. Table 1 shows the Bragg wavelengths for individual phases, beginning with design of the sensory branch and culminating with the installation of the system in the roadway, and the spectral parameters of the measuring channels. In Table 1, 'Design' column includes the proposed Bragg wavelengths, whereas 'Used FBG' column shows the wavelengths of the FBGs produced. Moreover, the 'PDMS-FBG sensor' column includes the wavelengths of the implemented PDMS-FBG sensors and 'PDMS-FBG sensor in the road' column includes the wavelengths of PDMS-FBG sensors installed in the road after curing process. The last three columns include the maximum (λ max ) and minimum (λ min ) wavelengths defining the spectral window and its width. Figure 8 illustrates the spectral distribution of the sensory branch. The Bragg wavelength is given for each reflection spectrum, and the width for each spectral window is shown in the lower part of the graph. Since the response to vehicle passage may be bipolar (positive and negative peaks, see Fig. 9), the spectral windows are defined to be symmetrical around the Bragg wavelength.
The PDMS-FBG sensor-chain with five PDMS-FBG sensors covers 20 nm of spectral space. When using two measuring chains, which will not overlap spectrally, it is possible to implement a single-line speed measurement with a one evaluation unit. The first measuring system can use a spectral space from 1510 to 1530 nm, the second from 1530 to 1550 nm. If speed measurement is required in two lanes, additional (third and fourth) measuring chains can be used and cover the spectral space from 1550 to 1570 nm and from 1570 to 1590 nm, respectively. All 4 chains will require 80 nm spectral space, which can be evaluated by one FBG evaluation unit FBGuard FAST 1550 (Safibra, s.r.o., Prague, Czech republic). This FBG unit consist of a wide-spectral light source with FWHM (Full Width at Half Maximum) 80 nm and an output power of 1 mW, and CCD (Charge-coupled device) spectrometer with a sampling frequency of 1 kS/s. For speed measurement, detecting the direction of the vehicle passage or two-lane monitoring, two PDMS-FBG chain sensors can be used connected to the FBG unit via a coupler. The individual sensors and the asphalt have a certain degree of flexibility; each passage is thus manifested by a positive peak followed by negative one, which is caused by the sensor stress release. The signal contains a low-frequency component corresponding to the memory effect and relaxation of the PDMS-FBG sensor and temperature change. These unwanted components are removed by a 4-order Butterworth high-pass 127652 VOLUME 8, 2020  filter with a cut-off frequency of 0.2 Hz. The filtered signal is shown in Figure 9(b).

D. PDMS-FBG CHAIN SENSOR SIGNAL PROCESSING
The PDMS-FBG chain sensor has been designed so that each axle is detected even if the passage takes place between two adjacent PDMS-FBG sensors. As a result, each passage stresses at least two sensors (unless the axle passes between the curb and the first sensor). When considering the axle load on several sensors, it is possible to sum signals from five PDMS-FBG sensors and obtain a more distinctive sum signal for vehicle passage detection. The individual responses of FBG sensors are typically manifested by a positive leading edge with a peak followed by a negative drop, caused by the elasticity of the PDMS material and asphalt at the moment of release of the axle load on the sensor branch. For this reason, only the positive parts of the response are taken into account in the signal processing. Figure 10 shows the sum of the positive parts of the filtered signals from 5 sensors. The evaluation unit sampling frequency 1000 samples/s was selected as minimum sampling frequency that ensures reliable detection of individual axles at speeds in normal city traffic. This fact along with fast responses of the PDMS-FBG chain sensor allows us to differentiate the individual passing axles, please see Fig. 10. The above signal processing procedure has been applied to all data presented in this publication. Figure 11 shows an example of a direct detection of 33 passing vehicles of various parameters. These vehicles' speed ranged from 32 to 87 kph with an overall response ranging from 33.36 to 339.55 pm. Table 2 shows the responses of the PDMS-FBG sensors and the sum signal for 3 selected passes of the same car. The first passage was realized in the middle part of the lane (Fig. 12(a)). The design of the sensor branch allows these passages to be detected with all PDMS-FBG sensors. The second crossing was carried out in the middle of the road, and therefore the PDMS-FBG sensor close to the curb did not show any response (Fig. 12(b)). Similarly during the third passage, the sensor located closest to the center of the road did not detect the crossing of the car at the roadside (Fig. 12(c)).   The response of individual PDMS-FBG sensors to car crossings is illustrated in Figure 13. The blue bar represents the response of the PDMS-FBG sensor located at the roadside, the green bar corresponds to the PDMS-FBG sensor located in the middle of the road.

IV. LONG-TERM MEASUREMENT AND EXPERIMENTAL VERIFICATION IN REAL OPERATION
This chapter describes and summarizes the results of the long-term traffic density and vehicle speed monitoring. The proposed measuring system using the PDMS-FBG chain sensor was tested in 18 all-day measurements taking place from January to November 2019 in city Ostrava (Part Vresina, GPS: 49 • 50'03.3''N 18 • 07'25.5''E). The measurements were carried out in various weather conditions (including adverse weather conditions such as rain, light snow or temperatures below freezing). These conditions had a negligible effect on the function (sensitivity) of the system and individual PDMS-FBG sensors. Nevertheless, the design of the measuring window of the sensors is designed in a way to encounter these affects by means of a so-called protection zone.
Along with the PDMS-FBG system measurement, the team also took a video recording of the individual car passages, which provided following information: time of the passes, direction, type of vehicles. The reference speed of the vehicles was obtained using optical gates, see subchapter 4.2 for more details. As part of a long-term measurement, 16 motorcycle and 27 cyclist passes were detected, and were removed from the dataset. Figure 14 shows the results of the single-lane vehicle detection. The results display the traffic density for all 18 measuring days realized from January to November 2019.

A. TRAFFIC DENSITY MONITORING
During the measurement, a total of 3,978 vehicles (3,927 cars and 51 trucks) passed through the measuring area.
The camera system captured all 3,978 vehicles while FBG measuring system was able to detect 3,963 vehicles. Summary of the detection accuracy is given in Table 3. The overall  accuracy of vehicle detection using the proposed measuring system was 99.62%.
The measurements were analyzed and classified; the significant points were categorized as True Positive -TP, False Positive -FP or False Negative -FN) [32]. Based on that, the detection quality of the proposed system can be evaluated using parameters such as Sensitivity (SE), Accuracy (ACC), Positive Predictive Value (PPV), Harmonic mean of precision and sensitivity (F1) and False Negative Rate (FNR) [33].
B. VEHICLE SPEED MEASUREMENTS Speed measurement requires placing at least two PDMS-FBG chain sensors in a given distance. The chain sensors were built into the road at a distance of 2 m from each other. For the reference speed measurement, a pair of optical gates was  used. The gates were installed at the locations of individual chain sensors as displayed in Figure 15. Figure 16 shows an example of vehicle speed measurement using signals from two PDMS-FBG chain systems corresponding to the vehicle passing at a speed of 63.33 kph. This reference was recorded using the optical gates. The speed can be calculated as v = L/ t, where L is the distance between the individual chain sensors (2 m) and t is the time difference between the peaks corresponding to the passage of the first axle (respectively second axle) over PDMS-FBG A and PDMS-FBG B sensor system. The resulting speed for first and second axle is v 1axle = 62.46 kph and v 2axle = 62.56 kph, respectively. By averaging these two values, we obtain a more accurate estimate of the speed v = 62.51 kph. This way, we determined a speed measurement error of 0.82 kph with a relative error of 1.29% related to the actual speed of the passing car. Figure 17(a) shows the real speeds of 3,963 vehicles obtained by the optical gates. The average speed of the passages was 50.93 kph, the lowest speed was 38.00 kph, and the maximum speed was 63.99 kph. The speed measurement error is shown in Fig. 17(b), and the relative error in Fig. 17(c).

V. CONCLUSION
The article presented the design, development and implementation of an optical fiber FBG system for monitoring selected traffic flow parameters in urban areas. The pilot study introduced herein included analysis of 3978 passing vehicles. The results revealed high accuracy of vehicle detection (99.62%) and speed measurements (mean absolute error 1.35 kph, mean relative error 2.65%). The advantages of the presented system include easy installation in the upper part of the road surface, possible connection to the existing dark (unused) fibers along city roads and implementation of remote evaluation within the SMART cities concept.
This pilot study presents the results a long-term research, in which 18 full-day measurements were performed from January to November 2019 in different weather conditions. Despite extensive research and development, there are still several issues and uncertainties that need to be addressed in follow-up research. This is primarily the effect of adverse weather conditions on the sensitivity and accuracy of the measurement and the overall life of the PDMS-FBG chain sensor in the roadway. He has been as an Associate Professor with the VSB-Technical University of Ostrava, since 2017. His current research interests include digital signal processing (linear and adaptive filtering, soft computing-artificial intelligence and adaptive fuzzy systems, non-adaptive methods, biological signal processing, digital processing of speech signals), wireless communications (software-defined radio), and power quality improvement. He has more than 200 journal and conference articles in his research areas.
EMIL BEDNAR was born in Ilava, Slovakia, in 1995. He received the bachelor's and master's degrees in information technologies from the Department of Information and Communications Technologies, Faculty of Electrical Engineering and Computer Science, VSB-Technical University of Ostrava, in 2017 and 2019, respectively, where he is currently pursuing the Ph.D. degree. His current research interest includes fiber-optic sensors predominantly in traffic and civil engineering.
JAKUB KOLARIK was born in Ostrava, Czech Republic, in 1992. He received the bachelor's degree from the Department of Cybernetics and Biomedical Engineering, VŠB-Technical University of Ostrava, in 2014, and the master's degree in control and information systems from the Department of Cybernetics and Biomedical Engineering. He is currently pursuing the Ph.D. degree in technical cybernetics.