Corresponding Symposium Track: 7- Municipal applications of GIS

Portable Geospatial Pavement Surface Assessment Model Wael M. ElDessouki, Ph.D. Department of Civil Engineering, Jazan University Contact address P.O. Box 2108, Jazan 45142, Saudi Arabia Phone: (+966 542328296) E-mail: wael_eldessouki@yahoo.com Corresponding Symposium Track: 7- Municipal applications of GIS ABSTRACT Assessment of pavement surface condition is a crucial part for pavement maintenance and management systems. A comprehensive assessment for the pavement condition in a metropolitan area will result in an efficient resource allocation and a successful pavement maintenance plan. Pavement assessment is typically performed either by manual inspection, which is highly demanding in terms of man power, or automated and usually outsourced to specialized companies that use highly sophisticated and expensive equipments. Large cities and municipalities can afford to outsource such tasks. However, small size cities and municipalities cannot afford either the automated and/or the manual method due to limited resources. Hence, in this paper, a portable low cost system for geospatial assessment of pavement surface condition that can be used by small cities and municipalities will be implemented. System design and specification will be presented and discussed. Samples of the currently available results will be presented and discussed as well. Finally we will conclude with recommendations for future work and implementation strategies. Keyword(s): Pavement Maintenance, Pavement Surface, Computer Imagery, GPS Introduction and Background Construction and maintenance of road networks are major investments for socioeconomic developments of nations as well as for small communities. Proper and timely maintenance of paved roads is crucial to preserve such capital. Fund allocation and prioritization of pavement maintenance projects is a challenging task, to say the least, which faces almost all decision makers. That can be rationalized to the nature of such problem which is a multi-objective optimization problem. The lack of quantitative assessment data for roads condition makes the decision making in such a problem even worse. It is clear by now, that quantitative assessment data for the road network condition will improve the decision making process. That will lead us to the question of which data and parameters are needed to assess pavement condition? The needed parameters in assessing pavement condition are derived from pavement characteristics which can be divided into five main categories according the World Bank report [1]. These categories are: 1- Roughness, which can be defined as “the deviations of a pavement surface from a true planar surface with characteristic dimensions that affect vehicle 1 2345- dynamics, ride quality, dynamic loads, and pavement drainage”, according to ASTM-E867[2] definition in Sayers et al[3]. In simple terms, roughness is a measure for the unevenness of road surface. It is a key measure for rid comfort and the overall usability of the road. Texture, which can be defined as the microscopic unevenness in pavement surface. It plays a major role in safety, noise, and rid comfort. Skid Resistance, it can be defined as friction level between the tires and the pavement surface. It is usually a function of pavement texture and usually associated with safety analysis of roads Mechanical/Structural Properties, which deals with the bearing capacity that the pavement can stand without permanent damage. Surface Distresses, which is associated with pavement deterioration, performance and aging. Surface cracks, surface defects, and lateral and longitudinal deformation can be identified as the major indicators for surface distress. On the network level maintenance, it would be irrational and expensive to collect a complete pavement characteristics data, as listed above, for all roads in the network. This would need almost half of the maintenance budget and, depending on available manpower and network size; it might take a year or may be more to analyze that volume of data. For network level analysis, a subset of pavement characteristics is sufficient. Even with that reduction, the data collection process is still expensive and costly such that small municipalities cannot afford to carry out and instead relay on expert judgment and qualitative assessment. Hence, the objective of this research was to develop a system with minimal cost and simple interface for pavement condition assessment, yet without compromising the quality of data. The results of the system can then be easily viewed with almost any Geographical Information System (GIS) software. The main pavement characteristics data that will be collected in the proposed system are pavement roughness and surface cracks. Both parameters are widely accepted as reasonable indicators for road serviceability and performance. In this paper we will present an System Design Road roughness has been measured since the 1920’s using different techniques. With the recent technology development, automation for such systems has evolved. The main measuring techniques can be classified as manual and automated. The automated techniques started with the South Dakota Roughness index that was based on using ultrasound to measure pavement surface deviations from true plan. Then, industrial laser technologies were introduced into this arena and added more to the cost of such systems. Despite the various claims that laser profilers have a higher precision and can take readings more frequently than ultrasonic based profilers, a comparative study by Ksaibati et al [4] concluded that the results using both technologies were strongly correlated and produced almost similar roughness indices for the tested samples. Given the fact that the proposed system here is intended only for usage on network level assessment, the conclusion of Ksaibati et al[4] led us to comfortably select ultrasonic sensors to be used for roughness measurements in the current system. 2 The overall system design is shown as a block diagram in Fig. 1. The system comprises of the following modules: 1- Roughness Measurement Module 2- Spatial Referencing Module 3- Pre-Processing and Data Storage Module 4- Surface Distress Module 5- Collective Data Storage Module Collective Data Storage Module: User – Input GUI Surface Distress Module: Notebook Computer Serial Connection Roughness Measurement Module: Pre-Processing and Data Storage Module: User – Input Via Push Button Digital Camera Microcontroller Unit Ultrasonic Sensor #1 Ultrasonic Sensor #2 Ultrasonic Sensor #3 Ultrasonic Sensor #4 Spatial Referencing Module GPS Receiver (TTL) Figure 1 System Block Diagram In the following sections we will limit our discussion to only the first three modules, and it will be premature at this stage to discuss the last two modules due to the fact that they are still underdevelopment and experimentations. 1- Roughness Measurement Module: This module consists of 4 ultrasonic Ping)) sensors by Parallax Inc.[5 ] that were temporarily mounted on the rear bumper of a regular passenger car using suction cups as illustrated in Fig 2. The measurement range for these sensors is from 2 cm to 3 meters with a precision of +/- 0.1 cm, or +/- 1 mm. The frequency of measurement was set in the firmware embedded in MCU to be 40Hz, i.e. 40 readings per sec., which means that if the test vehicle was driving at 40 km/hr, a reading will be taken every 27.75 cm (10 in). 2- Spatial Referencing Module: Global Positioning System (GPS) receiver was used for spatially referencing the roughness measurements. The GPS receiver used was BR355 by GlobalSat [6]. The GPS data records that was parsed and saved were the GPRMC records which included the basic referencing needed for this task. 3- Pre-Processing and Data Storage Module: The pre-processing module comprised of a multi-cog microcontroller 32 bit unit (Prop 3 8x32) that comprised of 8 cogs ran in parallel at 80MHz. One cog for the main program, 4 cogs were devoted to controlling the ultrasonic sensors, 2 cogs for receiving and parsing GPS data stream, one cog that can be used for serial communication with PC to store data, or running a local SD storage card. The operator has indicators when a GPS signal is available and a push button to start/stop recording data. In addition, a set of LED were included to give indications to the operator about the current state of the system. As this paper is being written, the roughness data is being preprocessed and stored along with GPS records locally on SD card integrated to the system. Figure 2 Ultrasonic Distance Sensors Mounting on the Rear Bumper of a Passenger Car Roughness Measurement Methodology: As mentioned before, roughness is a measure for the unevenness of the pavement surface. Highway engineers started in measuring road roughness and rid comfort as early as the 1920’s. Road roughness measurement methods can be divided into static and dynamic. Fig. 3 illustrates the DipstickTM device that is used to obtain road profile and for measuring roughness. It is obvious that such system static methods are highly demanding in terms of manpower and time. The dynamic methods can be also divided into response type systems and rolling profilers. The Bureau of Public Roads (BPR) introduced the Roughometer, which was a response type profiler that record the strokes in the vehicle suspension and accumulate it over a distance, Fig. 4. The results of such systems that followed the same concept were dependent on vehicle characteristics and not unstable. The second category of profilers was based on rolling an apparatus, similar in concept to the Dipstick, and measuring variations in the mid span. 4 Figure 3: The Dipstick TM Device for Estimating Road Profile and Roughness (Sayers [3]) Figure 4: The BPR Roughometer (Sayers [3] ) Then, the current roughness index (RI) measurement method consists of the following steps: Step 1: Take a sample of 8 elevation measurements over a 200 mSecs time period, i.e. 25 mSec apart, that implies that the distance between two successive road elevation readings ranged between (20 cm – 30 cm) if the test vehicle was driving in the range between (28.8 km/hr – 43. Km/hr). According to the ATSM E1926-98[7], for calculating IRI (International Roughness Index) for reliable precision in the results, the distance between successive sample readings should be at most 0.30 meter or less. Then we calculate the average road deviation in the sample as following: 5 n  d i 1 _ i d Equation (1) (n  1) Where, _ d - is the average elevation for the sample (stored by the MCU on local SD), d i - is the elevation measurement (i), n – is the number of readings in the sample, in this case n=8,  - The average road elevation deviations for the sample, the units will be in (mm/sample), (stored by the MCU on local SD), By subtracting 1 from the number of readings in the sample, we account for the average point is considered as the true reference elevation point. This step is carried out within the Pre-Processing and Data storage module. The data stored from each sample are then spatially referenced with a RMC record obtained from the GPS data. The following steps are done in the post processing phase for the results: Step 2: The values obtained using Eq.(1) have units of (mm/sample), which appears at first glance meaningless and irrelevant. The roughness measurements as defined by Sayers [8], is basically an elevation slope, elevation/linear distance. Hence, the value obtained in Eq.(1) are then normalization by dividing each value by the distance spanned by the sample, and the results will be (mm/meter). In order to make that normalization, the distances between each successive GPS positions were calculated using the spherical Earth projected to a plane method. However, it is important to mention that the GPS receiver had an update rate of 1 Hz and the samples were taken at 5 Hz. Therefore, the value for each sample was divided over only 1/5 of the distance obtained from the GPS data. That implies an inherent assumption that the distances between samples within 1 second were equal, which is acceptable since the driver did not apply excessive accelerations or decelerations during the test runs. Then, the normalized deviations in road elevation value for the sample becomes the Roughness Index for the sample, RI j    * f RI  f GPS * Rearth dLat 2  cos( Lat ) * dLong  2 Equation (2) Where, RI j - The normalized sum of road elevation deviations for the sample (mm/meter), which is the Roughness Index for sample (j), f GPS - Frequency of GPS data updates (Hz), f RI - Frequency of roughness sampling updates (Hz), or the number of samples/sec,  - The sum of road elevation deviations for the sample (mm/sample), Rearth – The spherical radius for the Earth (6371009 meters), dLat –the difference between the latitudes of the two points (in radians), 6 Lat - the mean latitude for the two points (in radians), dLong – the difference between the longitudes of the two points(in radians), It is important to note here that the above equation assumes that the Earth is a perfect sphere, which is definitely not the case. However, it is fairly acceptable to make such assumption in the cases of short distances. The maximum distance between any two successive GPS data records was function of vehicle speed and did not exceed 17 meters, which means that the error would be trivial and negligible. Step 3: Roughness Index aggregation over segments of preset length: In this step, we have designated a 40 meters segment length to aggregate the data and to calculate average RI values. It is clear from steps 1 & 2 that our Roughness Index calculation methodology is based on time, and the number of samples is inversely proportional to the speed of the test vehicle. This fact will give bias to the sections that was driven at lower speeds. Hence, in order to compensate for that bias the following calculation was adopted: ml RI 40 l   RI j 1 j Equation(3) ml Where, RI l40 - The average roughness index for the lth 40 meter segment, RIj – The roughness index for sample (j) in the lth 40 meter segment, ml – The number of RI samples in the lth 40 meter segment. It is important to mention that the presented methodology here for measuring road roughness index does not match that for the International Roughness Index (IRI) as described by Sayers et al [8]. Due to the fact that the current system takes sample measurements on time bases which makes the distance between samples vary with the speed of the vehicle, the IRI calculation method by Sayers[8] cannot be applied to the raw data. Hence, we refrain from labeling our results as IRI ; however, it is still a reasonable indication for road roughness as we will demonstrate in the following section. Results and Discussion Two test cases were made for initial evaluation for the performance of the system. Both tests were carried over on streets and major roads in the city of Jazan. It is important to emphasize that these test runs are for demonstration purpose only and cannot be considered as validation and verification of the system. Test Case 1: This test case included 4 road segments with different age and pavement surface condition. Fig. 5 illustrates those segments and subjective remarks regarding both age and condition. Fig. 6 shows the obtained RI using the developed system. The roughness results were aggregated over 40 meters sub segments, classified and color coded into three categories, Green (RI <3 mm/m), Yellow (3 mm/m <RI<6 mm/m), and Red ( 6 mm/m < RI), which corresponded to pavement condition good, moderate, and poor respectively. The limits for this classification were adopted from pavement 7 roughness interpretations a set by the World Bank in [8]. A quick comparison between Fig. 5 and Fig. 6 shows that the obtained by the current system matches reality. Note, the location of the speed humps on segment 2, the system was able to detect then almost in exact locations. Segment 1: Age: New Condition: Good Segment 2: Age: Relatively New Condition: Moderate Segment 3: Age: Old Condition: poor Speed humps Segment 4: Age: Old Condition: poor Figure 5 Roads Covered in Test Case 1 with Age and Condition Remarks (Google Earth) Roughness(mm/m) 0 – 3 (Good) 3 – 6 (Mod.) 6 – 20 (Poor) Figure 6 Roughness Results Obtained in Test Case 1 8 Test Case 2: In this test, a set of main roads and streets in the city of Jazan was driven and the results are shown in Fig. 7. In this test case, the aggregation length for averaging roughness results was selected to be 100 meter due to the size of the test case. Minor streets pavement condition matched reality for most part. However, major road with minor, but still apparent, defects, the system failed to make a good representation for them. This bias can be concluded due to the extended aggregation length of 100 meter which caused severe suppression for damaged short lengths of the segment. That is an indication that the model needs further calibration and more testing. Figure 7 Roughness Results for Test Case 2 (Data aggregation @ 100 meters) Conclusion and Future Work In this paper a low budget system for pavement condition assessment that is currently under development at Jazan University has been presented. The current total cost of the system modules, that have been developed and presented in this paper, is estimated to be less $200. System component and methodology for measuring roughness index has been also presented. The early results from the developed system were good and reflected actual pavement conditions. The current system included only roughness measurement module, it is intended to integrate surface distress and cracks classification modules in the next stage. In addition, the roughness measurement module needs further modifications and fine tuning before being considered as a final stage deliverable. The current roughness results need to be validated and verified using manual methods for roughness measurement. 9 Acknowledgements This work was partially sponsored by Jazan University and SABIC under Deanship of Scientific Research grant number 33/4/27. Also, the author would like to thank Dr. Muhammad Ali Mubaraki for his valuable comments and feedbacks on this work. List of References [1] C Bennett, A Chamorro, C Chen, H Solminihac, and G. Flintsch “Data Collection Technologies for Road Management”, version 2.0, World Bank Report- 2007 [2] ASTM E867 “Standard Terminology Relating to Vehicle-Pavement System” [3] M. Sayers and S. Karamihas, “The Little Book of Profiling: Basic Information about Measuring and Interpreting Road Profiles”, The Regent of University of Michigan, 1998 [4] K. Ksaibati,R. McNamara, and J. Armaghani, “A Comparison of Roughness Measurements from Laser and Ultrasonic Road Profilers”, Research Report, FLDOT, 1998 [5]http://www.parallax.com/Portals/0/Downloads/docs/prod/acc/28015-PING-v1.6.pdf [6] http://www.usglobalsat.com/store/download/57/br355_ds_ug.pdf [7] ASTM E1926 – 98 “Standard Practice for Computing International Roughness Index of Roads from Longitudinal Profile Measurements” [8] M. Sayers, T. Gillespie, and W. Paterson “Guidelines for Conducting and Calibrating Road Roughness Measurements”, World Bank Technical Paper # 46 – 1986 [9] “Technical Guidance for Deploying National Level Performance Measurements”, Final Draft Report, NCHRP 20-24(37)G- 2011 10