- Email: [email protected]
Embedded system to evaluate the passenger comfort in public transportation based on dynamical vehicle behavior with user’s feedback
Measurement 47 (2014) 442–451
Contents lists available at ScienceDirect
Measurement journal homepage: www.elsevier.com/locate/measurement
Embedded system to evaluate the passenger comfort in public transportation based on dynamical vehicle behavior with user’s feedback Juan C. Castellanos ⇑, Fabiano Fruett School of Electrical Engineering, University of Campinas, Av. Albert Einstein 400, CEP: 13083-852 Campinas, SP, Brazil
a r t i c l e
i n f o
Article history: Received 24 January 2013 Received in revised form 27 August 2013 Accepted 30 August 2013 Available online 19 September 2013 Keywords: Global Positioning Systems Accelerometer Comfort Intelligent transportation systems
a b s t r a c t This paper shows the development of a system (Hardware, Firmware and Software) focused to assess the dynamic motion factors that affect the comfort in public transportation systems. The data is collected, on-board processed and transported using the public transportation system vehicles as mobile smart sensors. Therefore, the acceleration measurement using a tri-axial accelerometer, the position detection using Global Positioning System (GPS) and the appropriate algorithms allow the system to detect rude driver styles and defects on the pavement. The firmware is composed by two algorithms. The first one is based on the detection of acceleration and Jerk magnitudes out of the comfort range, which is called Jerk-Acceleration Threshold Detection (JATD). An algorithm to compute the Jerk with comparable results to prior researches is proposed in this paper. The second algorithm, called Comfort Index with Acceleration Threshold Detection (CI-ATD), is based on the detection of acceleration values out of comfort range and the average ride comfort. The average ride comfort is supported by the recommendation of the international standard ISO2631-1. The comfort range or threshold values can be set using the user’s perception. A software developed in LabVIEW™ interface, visualizes discomfort event in online maps for geographic location of each event. Also, the software implements road unevenness detection, which is based on the collected data analysis. The system was successful tested in a conventional bus line on its daily ride, the results reveals that most of the events are due to vertical acceleration disturbances. Also, a preliminary test indicates higher sensibility for vertical than longitudinal or transversal accelerations. Ó 2013 Elsevier Ltd. All rights reserved.
1. Introduction Transport policies are important for society, because efficiency in the transportation system have a positive economic and social impact as better accessibility to markets, employment and additional investments [15]. Nowadays, one of the principal problems in big cities is the large fleet of vehicles [5,21]. Some policies such as improving the quality of public transportation would collaborate to turn
⇑ Corresponding author. Tel.: +55 (19)88410241. E-mail addresses: [email protected] (J.C. Castellanos), [email protected] (F. Fruett). 0263-2241/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.measurement.2013.08.068
car’s users into public service users; this change of roles will have a positive impact in decreasing traffic jams, atmospheric and noise pollution [5,11,18]. However, the efficiency of policies depends on the service quality. According to the Highway Capacity Manual 2010 [19], the quality of service can be influence from several factors as speed, travel time, reliability, convenience, maneuverability, cost, accessibility, safety, comfort, etc. Statistics from vehicles fails, accidents, number of complaints and specific surveys are used to assess the quality of the transportation service [16]. However, performing comfort statistics is an expensive task in terms of human resources and time, because of surveys and personal
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
interviews [12]. Hence, systems to collect objective data in aim of guarantee or assess safety and comfort are being proposed. Several applications have emerged to improve the quality of transportation system; those could be classified in four topics: mobility, information, security, and comfort. In the first topic, mobility solutions as Bus Rapid Transit (BRT) and subways use fleet management system, as the ITS4mobility [20], to improve the impact in urban mobility. Nowadays, information about buses routes, station localization, arrival and departure times can be accessed through web sites or phone applications, e.g. the Bus Catcher is an application developed to assists mobile users to plan their travels using a PDA, resulting in saving time [3]. Since the last decade, Global Positioning Systems (GPS) are most widely used; this technological advance became the Vehicle Tracking System in one of the most successful solution in the security field. The comfort measurement may be the most challenge topic to be measure, because it depends on the human perception and it is affected by several factors. Förstberg [6] specifies some types of human-vehicle interaction variables that influence the user’s comfort: Human Factors (as age and gender), environmental factors (as temperature, noise, and pressure), spatial factors (as workspace, leg room, seat shape, etc.) and dynamic motion factors. The ISO 5805 [9,17] defines comfort just as ‘‘subjective state of well-being or absence of mechanical disturbance in relation to the induced environment’’. In this paper, the comfort scopes will be limited to mechanical disturbances or dynamic motion factors. Usually, the quality of the comfort is evaluated using subjective rating tests [17], which may include several issues not coupled with the ride as mood, tiredness, etc. Also, the vocabulary used to rate may be different among individuals and changes in perception level or acceptance level may occur. Therefore, the objectivity of the evaluation of comfort is an interesting issue to be implemented, because, it also allows repeatability, comparison between competitors and creation of scales as mentioned by Strandemar [17]. Vehicles interaction with road will be reflected in terms of acceleration, Jerk (the rate of acceleration change) and angular motion, which depend on bus maintenance, driver behavior and road’s state. In elevators, Jerk is used to confirm motion control settings associated to ride quality [10]. Moreover, Andersson et al. [1] demonstrated the influence of the train lateral Jerk in walking, standing and sitting passenger. It is possible to assess the dynamic motion comfort using five methodologies: average ride comfort, estimated ride comfort, comfort disturbances, motion sickness and vibration effects in daily activities [6]. Reveriego [14] collects data using the acceleration, speed and position through a mobile telephone. Those data can be used to evaluate the fore-and-aft and lateral ride comfort using a methodology based on acceleration thresholds. In the same way, Zoysa et al. [22] proposes a system sensor network to monitor road surface condition. It can be used to detect vertical ride comfort using the comfort disturbance with
443
a vertical acceleration threshold. However, those solutions are not designed to process the acceleration data as average ride acceleration, which make difficult the comparison, e.g. between drivers, roads, etc. Lin et al. [12] uses average ride comfort methodology to assess the comfort. Using participatory phones to collect data, a ride index is computed in a server and matched with the transportation database. This system allowed them to compare different bus types in aim to determinate the most comfortable vehicle (bus type) and ride. However, the comfort evaluation depends on the availability voluntaries and their care for measurement, e.g. fixation technique. Also, the battery discharge and costs associated to data communication (GPRS) can discourage the voluntary users and limit the distance of the evaluation. The main contribution of this paper is the implementation of a complete system (Hardware, Firmware and Software) specially designed to assess the dynamic motion comfort in public transportation. The data reported by the system can be highly important for transportation agencies, road state surveillance authorities and efficiency of quality transport polices. The innovative work includes using buses as mobile sensors to collect, process and transport information related with comfort in public transportation (as driver’s behavior, road state, etc.). Also, previously works often used just one measurement methodology. The proposed system uses two complementary measurement methodology (average ride comfort, comfort disturbances) for an objectively comfort assessment without need additional data processing. The average ride comfort methodology follows the entire international standard ISO 2631-1 [8] for comfort measurement. The comfort disturbances detect sudden motion, which is measured through acceleration and Jerk signals above a defined threshold. The threshold value can be determined by the own system using passengers experience feedback hardware under controlled ride conditions. Those two methodologies are complementary, the first methodology provides an average measurement, which can be used to compare tracks, vehicles or drivers and it is in accordance with the standard ISO 2631-1. On other hand, the second methodology allows the source of discomfort events to be localized, geographically and temporally. In this way, algorithms using the two methodologies were created in this paper: Jerk and Acceleration Threshold Detection (JATD) and Comfort Index with Acceleration Threshold Detection (CI-ATD). These methodologies were both embedded in the system software. The hardware, described in Section 2, is based on tri-axial accelerometers, a Global Positioning System (GPS) module, wireless communication modules and user’s feedback module. The firmware, described in Section 3, is divided in the two algorithms JATD and CI-ATD. The software, described in Section 4, is composed by three parts: event’s visualization interface, set-up (configuration) interface and defects on the pavement detection interface. Also, Section 5 presents the results obtained using the system with the two algorithms, user’s feedback module and analysis of data to detect defects on the pavement. Finally, section 6 presents conclusions and discussion associated to this paper.
444
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451 Table 1 Technical specification of sensor node.
2. Hardware The system hardware was based on the former work previously described in [4]. The system of the current work also uses a sensor and base node including a new extension module called User’s Feedback Extension Module (UFE). The UFE module allows the operator to set the acceleration threshold with real user’s experience feedback for specific population e.g. elderly people, passengers with special needs, etc. The hardware description and electrical characteristics of the Embedded System for Comfort Measurement (ESCM) and UFE module are shown below.
Item
Description
Acceleration range Acceleration bandwidth Sample frequency Maximum position accuracy Wireless range Memory type Energy consumption Power supply range Size
±4 G 50 Hz, 100 Hz 500 Hz 5m 140 m (indoor) SD™ Card 250 mA 5–30 V 5 cm 9 cm
2.1. Sensor node
2.3. User’s feedback extension (UFE)
The ESCM sensor node type acquires, processes and transmits the acceleration data which is used to detect discomfort events. Also, the events generated are saved in a memory. At the end of the journey, the sensor node delivers the events data to the base node. The sensor node is installed on the vehicle with the device’s axes align with the motion axes. Fig. 1 shows the sensor node, which contains a 32-bit microcontroller, a GPS module, a digital tri-axial MEMS accelerometer, a temperature sensor, a wireless communication module to 900 MHz and finally a socket for a Secure Digital (SD) memory. Furthermore, the sensor node has two expansion ports. Table 1 describes some technical specifications of the sensor node.
The UFE module is used to determinate the acceleration threshold value according to the passenger’s perception. It is composed by a microcontroller and eight switches. Those modules are connected to the expansion port at the sensor nodes.
2.2. Base node The second ESCM device, the base node, is installed in the central station. This node transmits a broadcast signal to request the event’s transmission from the sensor node. This node is an interface between wireless module and a computer USB port, where the events are visualized and analyzed in the user’s interface based on LabVIEW™.
3. Firmware In this paper, three measurement techniques are used to assess the comfort disturbances or the comfort level. The first one uses the acceleration data to detect accelerations out of a range defined by a threshold value. The second one calculates the Jerk signal to detect Jerks values out of a range defined by a threshold value. The last technique is based on the international standard ISO 2631-1, which describes the vehicles vibration effects on the human body and it defines a comfort index (CI) based on the Root Mean Square (RMS) value from acceleration-weighted data. Using these measurement techniques, two algorithms were created to evaluate the comfort, both of them using acceleration measurement. The algorithms are: the Jerk-
Fig. 1. Sensor node.
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
Acceleration Threshold Detection (JATD) and Comfort Index with Acceleration Threshold Detection (CI-ATD). The three techniques are explained below:
3.1. Acceleration threshold detection with user’s feedback The acceleration threshold detection allows the system to detect disturbances in the comfort due to excessive acceleration. Here, the maximum absolute value of acceleration is computed each second. If this value is higher than the threshold, the firmware gets the position from the GPS module and an event structure is generated, after that the event is saved in the SD memory. The event’s structure is composed by a header, Bus Id, number, type, value and position of the event. Each event’s structure contains a field called ‘‘Event Type’’, which indicates the source of the disturbance. Table 2 shows the possible values for Event type field. The vertical, transversal and longitudinal acceleration signals from a bus journey of 21 min were measured. The signals show several disturbance events due to fast turns, abrupt starts, breaks maneuvers and imperfection or obstacles on the road. Fig. 2 shows an example of this
a
detection technique, where is detected a vertical disturbance event of two seconds produced by a speed bump. In this case, the acceleration has magnitudes higher than the threshold set to 0.5 G. 3.1.1. UFE module firmware The UFE module generates a passenger’s event each time that the user indicates a discomfort using a pushbutton. The firmware of the UFE module is based on a slave device, which holds in a register the information of the buttons pressed until the master device (sensor node) reads it. The passenger’s events are composed by information as acceleration magnitude, acceleration axes and position of the event reported by the user. 3.2. Jerk threshold detection The Jerk is defined by an international standard ISO2041 [7] as: ‘‘The resultant vector of the acceleration derivate in the time domain’’. Therefore, not only the Jerk’s magnitude varies according to acceleration magnitude, but also, it changes with the duration of the acceleration change. In order to illustrate this concept, it is modeled an acceleration signal as a sinusoidal signal (Eq. (1)):
aðtÞ ¼ A sinðxtÞ;
Table 2 Event type description. Event type
Event description
0 1 2 3 4 5 6 7 8 9
No event Longitudinal acceleration event Transversal acceleration event Vertical acceleration event Longitudinal and transversal event Longitudinal and vertical event Transversal and vertical event All axial acceleration event Passenger’s event Jerk’s event or index comfort indicatora
Type 9 changes their functionality according to the algorithm used.
445
ð1Þ
where A is the acceleration magnitude and x is the acceleration frequency. The Jerk, J(t), is obtained performing the derivation of a(t) (Eq. (2)). Note that, the Jerk’s magnitude depends on the magnitude and frequency of the acceleration signal.
JðtÞ ¼ A cosðxtÞ:
ð2Þ
The frequency spectrum of the acceleration signal is important to design the algorithm to calculate the Jerk. Some factors as filter order and cut-off frequency has to be chosen from the spectrum analysis. In this way, a Wavelet analysis was performed over the acceleration’s signals.
Fig. 2. Vertical acceleration event due to speed bump.
446
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
A Discrete Wavelet Transform (DWT) analysis using the longitudinal acceleration analysis reflects the signal due to the forward advance and the gears changing maneuver. The result shows low frequency components below to 0.3 Hz due to starting maneuver, components below to 0.5 Hz due to changing gears maneuver and higher components (5–40 Hz). Similar analyses were performed using vertical and transversal accelerations. Considering the previous results, the stop and run behavior of the cities and the experimentation in trains with passengers made by Andersson [1], it was chosen the longitudinal acceleration axis to perform the comfort analysis. Thus, it was implemented the same filter used by Andersson [1], i.e. a low-pass filter (LPF) with cut-off frequency (FC) of 0.3 Hz. The block diagram of this detection technique is shown in Fig. 3. The result obtained using the proposed algorithm over a real longitudinal acceleration signal with several start and stop vehicle actions are shown in Fig. 4. The initial threshold was set to 0.4 m/s3 according to those results, the recommendation of Förstberg [6] and the comfort’s research made by Andersson [1]. The results of Andersson showed that 30% of walking passengers perceived a discomfort above 0.4 m/s3. The JATD algorithm is composed by this detection technique and the first one.
3.3. Comfort index measurement The third technique measures the ride comfort according to an indicator called Comfort Index (CI). There are different types of CI, such as: WZ, Ride Index, NMV, NVA, NVD, PCT, PDE and ISO 2631 [6,8]. In this work, it is used the CI described by the international standard ISO 26311, which uses the Root Mean Square (RMS) value of the weighted-acceleration signal in three axes. The frequency range of measurement for comfort evaluation should be from 0.5 Hz to 80 Hz. The basic method to calculate the RMS value is described by following equation:
aW ¼
Z T 12 1 a2W ðtÞdt ; T 0
where aW is the weighted RMS acceleration value (m/s2), T is the duration of measurement in seconds and aW(t) is the weighted acceleration signal. The aW is calculated for each of the three axes, where each axis has a specific weightcurve. For lateral and transversal axes is used the transfer function xd and for vertical axis is used the transfer function xk, those transfer functions are valid for measurements on the seat or vehicle’s floor. The impulse response of xd and xk are given by the standard ISO2631-1 [8]. After the calculation of the RMS value, (based on Eq. (3)) in each axis, it is calculated the value of the total RMS acceleration using the coefficients kX, kY and kZ combining three axes. The expression used is described by Eq. (4):
12 2 2 2 aV ¼ kX a2WX þ kY a2WY þ kZ a2WZ ;
Fig. 3. Block diagram for Jerk threshold detection technique.
ð3Þ
ð4Þ
where aV is the total weighted RMS acceleration. If the comfort is evaluated for standing passengers as this case, the coefficients k will be one.
Fig. 4. Longitudinal acceleration and longitudinal Jerk of a bus journey using the Jerk detection technique.
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
In this technique, it is used the Fast Fourier Transform (FFT) to transform the acceleration signal of each axis in the time domain into the same signal in the frequency domain to apply the weight values. After that it is performed the RMS calculation each second. To perform an evaluation in a longer period of time (Tm), it is used the expression of Eq. (5), which uses the aV computed each second and the average expression for RMS values.
"
aRMST m
#12 Tm 1 X 2 ¼ a : T m i¼0 Vi
ð5Þ
Finally, it is given a comfort index (CI) according to the calculated aRMST m . The comfort index definition in Table 3 is based on the ISO standard. The CI-ATD algorithm is composed by this detection technique and the first one. 4. Software The software is based on Graphical User Interface (GUI), which was developed using the LabVIEW™ platform. This software is composed by three parts. The first one is the Event’s Visualization Interface, in which is visualized the events in an online map. The second one allows the configuration and reading of the sensor nodes. The third one is called Defects on the Pavement Detection Interface (DOPDI). It is a processing data interface, which analyze several travels in the same route to localize the repeated events in the vertical axis. This procedure allows the identification of defects on the pavement over driver’s mistakes. The DOPDI uses a radius of tolerance to recognize several nearby events as a unique discomfort source with the average position. Table 3 Comfort index definition. Index
Range (G)
Event description
0 1 2 3 4 5
Greater than 0.229 0.145–0.229 0.092–0.145 0.057–0.092 0.032–0.057 Less than 0.032
Extremely uncomfortable Very uncomfortable Uncomfortable Fairly uncomfortable A little uncomfortable Not uncomfortable
447
5. Results 5.1. Jerk-Acceleration Threshold Detection (JATD) An early test was performed to explore the maximum value of the longitudinal Jerk. This test was carried out by car, where their journeys were characterized by several breaking and starting maneuvers under moderate speed conditions. In those testes were obtained a maximum Jerk of 0.98 m/s3. The evaluation test was carried out in two urban bus lines, line A and line B. The results of the line A show the acceleration’s events with maximum amplitude of 1.86 G, the acceleration’s threshold was set to ±0.7 G; moreover, most of the acceleration’s events correspond to vertical events type. The maximum Jerk event has amplitude of 0.562 m/s3, among six Jerk’s events. The results of the second line show acceleration’s events with a maximum amplitude of 1.71 G and it detects a unique Jerk event of 0.412 m/s3. The data obtained reveals that 95% of the vertical acceleration’s events are below to 1.3 G and 1.08 G for the line A and B, respectively. It probably means that the track of the line A has a poor condition on the road than the track of the line B. The results of the second test of JATD algorithm are shown in Fig. 5 using the Event’s Visualization Interface.
5.2. Comfort Index with Acceleration Threshold Detection (CIATD) Several tests were performed to check the CI-ATD algorithm, however, in this article are presented just two of the most representative tests. The first test was carried out in two vehicles on the same route; the first vehicle was a university bus and the second one was a particular car. In the test, both vehicles used the same sensor device with the acceleration threshold set to 0.7 G and it calculates the CI each 15 min. The result of the first vehicle journey not only reveals more events than the journey made by the second vehicle; but also, it calculated a lower CI. Most of the events were vertical and longitudinal type as it is shown in Fig. 6. The lowest CI obtained in the first vehicle journey
Fig. 5. JATD Results in an urban bus lines. At the left is shown the line A results. At the right is shown the line B results. Yellow markers indicate Jerk events. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
448
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
Fig. 6. CI-ATD First test results, two vehicles at the same track. At the left, the results from a bus journey. At the right, the results taken from an automobile journey.
was two, which had a magnitude of 0.128 G; whereas, the second one was three with a magnitude of 0.082 G; those CIs follow the scale shown in Table 3. Also, the test localized a street in poor maintenance state detected by both vehicles, which presented high density of acceleration events. The second test of the CI-ATD was performed on an urban bus line of Campinas, Brazil. The test was carried out in real situation with passenger’s loads. The journey had a duration of 46 min and around of 20.4 km distance. The data obtained was analyzed by each motion axis (transversal (X), longitudinal (Y) and vertical (Z) axes). Also, mixed events (XY, XZ and YZ) were analyzed as isolated events in longitudinal, transversal and vertical axes. Those results are shown in Fig. 7 in terms of the occurrence frequency and magnitude.
Also the comfort indexes obtained each 5 min were analyzed. The results revels an average comfort index equals to 2.3, which is uncomfortable according to Table 3. 5.3. User’s feedback results using the UFE module The follow test illustrates the methodology to set up the acceleration threshold based on the user’s experience of a specific population group, as elderly, visual deficient, students’ passengers, etc. Those subjects registered their opinion about the ride through a push-button action. The test is composed by three laps at the same road of 2.8 km, which in the first lap the driver limited the average speed to 30 km/h and when the vehicle passed over a speed bump the speed was reduced to 10 km/h. In the second lap was limited to 50 km/h and 20 km/h, respectively. The last lap
Fig. 7. Acceleration events analysis of bus journey in the same track (longitudinal, transversal and vertical events).
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
was performed with an average speed of 60 km/h and obstacle speed of 40 km/h. This test was carried out in a particular car with two test’s subjects and it was repeated three times with different passengers. Since, the test’s subjects were an unclassified group of six people, the sample size is not enough for statistical significance of the results, but it is enough to show the purpose of the user’s feedback function. The transmission device was installed on the floor of the vehicle and it was set with the CI-ATD algorithm, which was programmed with a low acceleration threshold of 0.5 m/s2. Fig. 8 shows the route of the test and the events detected by the sensor node due to the poor state of the road, the driver style or the user’s perception registered through one button. The analysis of the events was performed according to the acceleration type (vertical, longitudinal and
449
transversal). The user’s perception events were compared with their respective acceleration events. Fig. 9 shows the results for this comparison, which reveals that 70% of the users perceived a disturbance with a maximum acceleration of 0.7 G, 0.8 G and 1.4 G for lateral, longitudinal and vertical events, respectively. 5.4. Defects on the Pavement Detection Interface (DOPDI) The DOPDI software was test performing four travels by particular car at the same track to show its functionality. In those testes was used the CI-ATD algorithm. The tolerance radius was set to 10 m. After the DOPDI program was executed using the data of the four travels, it produces the output file with the position of the vertical repeated event markets. The unified vertical events contain informa-
Fig. 8. Test track with several speed bumps and one roundabout.
Fig. 9. User’s perception events and CI-ATD events comparison detected using CI-ATD with UFE module.
450
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
Fig. 10. DOPDI analysis results using tolerance radius of 10 m.
tion about the quantity of the repeated events and its average position. Fig. 10 shows the result for this analysis, where several vertical discomfort sources were detected.
6. Conclusions This paper presented a successful system to localize comfort disturbances using the public transportation system to collects, process and transmit the relevant information. This information allows the detection of dangerous driving style behaviors such as excessive accelerations. This system was successfully tested in real conditions and also used to determinate the sensibility of the passengers to the acceleration magnitude and the pavement defects on the road. This information is important for both transportation and governmental agencies in aim to improve the service and the road’s state. Two algorithms using three techniques to assess the comfort in a journey were presented in this article. The first technique, the acceleration threshold detection with user’s feedback, is appropriate to localize geographically the discomfort sources due to excessive accelerations. The second technique, Jerk threshold detection, adds other criteria to the detection of discomfort’s events based on the longitudinal acceleration changes (Longitudinal Jerk). This technique is adequate to assess heavy traffic or transportation systems with unscheduled starts-stops. The rating output depends on the quantity of events and its magnitude, hence, it is difficult to compare among different journeys and establish which was the most comfortable. The low frequency characteristic of Jerk signal impacts negatively in the localization accuracy, also, the Jerk magnitude depends on the acceleration’s signal to noise rate (SNR). Therefore, filter influences in the Jerk computation. Some previously researches [1,2,6,13], that involve Jerk do not define a unique method to calculate it from a vehicle’s acceleration. Hence, most of times, it is impossible to compare the results obtained in those studies among them. In this way, it is essential to establish a standard way
to calculate the Jerk in comfort’s applications following similar parameters on the filtering stage. The obtained results using this algorithm has Jerk magnitudes similar to the results obtained by Förstberg and Andersson [1,6] using similar filtering parameters. However, Bagdadi [2] shows interesting results using an alternative Jerk algorithm that can be implemented by ESCM. The comparison concern is solved using the technique, called Comfort Index Measurement. Therefore, the comparison among drivers or journeys becomes into a simple numerical comparison. Although, using just this technique, the geographic location of discomfort sources is not possible. Hence, this technique is merged with the first technique (CI-ATD). For a simple test, using UFE module, the user’s sensibility is lower for the vertical axis than longitudinal and lateral axis of acceleration, but this information has to be tested with more subjects before to turn into valid result, statistically. This test could be used to determine the acceleration or Jerk threshold value in specific populations as children, pregnant, elderly or people with special necessities, which information can be really interesting to improve the quality of special transport services e.g. scholar buses. Implementing a specific hardware instead of using the embedded sensors in mobile devices allows independence on harvesting data from the user’s availability. Also, it ensures the quality of the data gathering necessary to accomplish the ISO2631 recommendations and decreases the quantity of data processed due to their on-board signal processing. Acknowledgements The authors acknowledge the Brazilian National Council of Scientific and Technological Development – CNPq under Universal Project No. 480864/2011-0 and CAPES for the financial support. Also, the authors acknowledge the bus Company Itajaí in Campinas City, São Paulo State, Brazil, for the test vehicle provided and the team of Microelec-
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
tronic Sensors Laboratory staff, LSM, at the UNICAMP for the academic support. References [1] E. Andersson, N. Nilstam, Development of advanced high speed vehicles in Sweden, Proceeding of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering 198 (1984) 229–237. [2] O. Bagdadi, A. Várhelyi, Development of a method for detecting Jerks in safety critical events, Accident Analysis & Prevention (2012). [3] M. Bertolotto, G. O’Hare, R. Strahan, A. Brophy, A. Martin, E. McLoughlin, Bus catcher: a context sensitive prototype system for public transportation users, in: Proceedings of the Third International Conference on Web Information Systems Engineering (Workshops), 2002, pp. 64–72. [4] J.C. Castellanos, A.A. Susin, F. Fruett, Embedded sensor system and techniques to evaluate the comfort in public transportation, in: 2011 14th International IEEE Conference on Intelligent Transportation Systems (ITSC), 2011, pp. 1858–1863. [5] DETR, A new deal for transport: better for everyone (white paper), 1998. [6] J. Förstberg, Ride Comfort and Motion Sickness in Tilting Trains, Department of vehicle engineering, Royal Institute of Technology (KTH), Stockholm, Sweden, 2000. [7] ISO, Vibration and Shock – Vocabulary, ISO 2041, 1990. [8] ISO, Mechanical Vibration and Shock – Evaluation of Human Exposure to Whole-Body Vibration – Part 1: General Requirements, ISO 2631–1, 1997a. [9] ISO, Mechanical Vibration and Shock – Human Exposure – Vocabulary. ISO 5805, 1997b. [10] ISO, Lifts (elevators) – measurement of lift ride quality, ISO 18738, 2003. [11] P. Kelly, Quality bus transit systems, in: Public Transport Electronic Systems, International Conference on (Conf. Publ. No. 425), 1996, pp. 11–15.
451
[12] C.-Y. Lin, L.-J. Chen, Y.-Y. Chen, W.-C. Lee, A comfort measuring system for public transportation systems using participatory phone sensing, in: Proceedings of PhoneSense 2010, Zurich, Switzerland, 2010. [13] J.J. Martinez, C. Canudas-de-Wit, Model reference control approach for safe longitudinal control, in: American Control Conference, Proceedings of the 2004, vol. 2753, 2004, pp. 2757–2762. [14] G. Reveriego, C. Urdiales, F. Sandoval, Vehicle risks assessment using a nokia N95, in: Intelligent Transportation Systems, 2009. ITSC ‘09. 12th International IEEE Conference on, 2009, pp. 1–6. [15] J.-P. Rodrigue, C. Comtois, B. Slack, The Geography of Transport Systems, second ed., Routledge, London, New York, 2009. [16] M.O. Rodrigues, An Evaluation of the Level of Service in the Public Transportation System of the City of São Carlos, School of Engineering of São Carlos, University of São Paulo, São Carlos, Brazil, 2006. [17] K. Strandemar, On Objective Measures for Ride Comfort Evaluation, Department of Signals, Sensors and Systems. Royal Institute of Technology (KTH), Stockholm, Sweden, 2005. [18] A. Tennøy, Why we fail to reduce urban road traffic volumes: does it matter how planners frame the problem?, Transport Policy 17 (4) (2010) 216–223 [19] TRB, Highway Capacity Manual, fifth ed., Transportation Research Board, Washington, DC, 2010. [20] Volvo, Its4mobility, in: V.B. Corporation (Ed.), Technical Description, Göteborg, Sweden, 2009. [21] S. Ye, Research on Urban road traffic congestion charging based on sustainable development, Physics Procedia 24 (Part B(0)) (2012) 1567–1572. [22] K.D. Zoysa, C. Keppitiyagama, G.P. Seneviratne, W.W.A.T. Shihan, A public transport system based sensor network for road surface condition monitoring, in: Proceedings of the 2007 Workshop on Networked Systems for Developing Regions, ACM, Kyoto, Japan, 2007, pp. 1–6.
Contents lists available at ScienceDirect
Measurement journal homepage: www.elsevier.com/locate/measurement
Embedded system to evaluate the passenger comfort in public transportation based on dynamical vehicle behavior with user’s feedback Juan C. Castellanos ⇑, Fabiano Fruett School of Electrical Engineering, University of Campinas, Av. Albert Einstein 400, CEP: 13083-852 Campinas, SP, Brazil
a r t i c l e
i n f o
Article history: Received 24 January 2013 Received in revised form 27 August 2013 Accepted 30 August 2013 Available online 19 September 2013 Keywords: Global Positioning Systems Accelerometer Comfort Intelligent transportation systems
a b s t r a c t This paper shows the development of a system (Hardware, Firmware and Software) focused to assess the dynamic motion factors that affect the comfort in public transportation systems. The data is collected, on-board processed and transported using the public transportation system vehicles as mobile smart sensors. Therefore, the acceleration measurement using a tri-axial accelerometer, the position detection using Global Positioning System (GPS) and the appropriate algorithms allow the system to detect rude driver styles and defects on the pavement. The firmware is composed by two algorithms. The first one is based on the detection of acceleration and Jerk magnitudes out of the comfort range, which is called Jerk-Acceleration Threshold Detection (JATD). An algorithm to compute the Jerk with comparable results to prior researches is proposed in this paper. The second algorithm, called Comfort Index with Acceleration Threshold Detection (CI-ATD), is based on the detection of acceleration values out of comfort range and the average ride comfort. The average ride comfort is supported by the recommendation of the international standard ISO2631-1. The comfort range or threshold values can be set using the user’s perception. A software developed in LabVIEW™ interface, visualizes discomfort event in online maps for geographic location of each event. Also, the software implements road unevenness detection, which is based on the collected data analysis. The system was successful tested in a conventional bus line on its daily ride, the results reveals that most of the events are due to vertical acceleration disturbances. Also, a preliminary test indicates higher sensibility for vertical than longitudinal or transversal accelerations. Ó 2013 Elsevier Ltd. All rights reserved.
1. Introduction Transport policies are important for society, because efficiency in the transportation system have a positive economic and social impact as better accessibility to markets, employment and additional investments [15]. Nowadays, one of the principal problems in big cities is the large fleet of vehicles [5,21]. Some policies such as improving the quality of public transportation would collaborate to turn
⇑ Corresponding author. Tel.: +55 (19)88410241. E-mail addresses: [email protected] (J.C. Castellanos), [email protected] (F. Fruett). 0263-2241/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.measurement.2013.08.068
car’s users into public service users; this change of roles will have a positive impact in decreasing traffic jams, atmospheric and noise pollution [5,11,18]. However, the efficiency of policies depends on the service quality. According to the Highway Capacity Manual 2010 [19], the quality of service can be influence from several factors as speed, travel time, reliability, convenience, maneuverability, cost, accessibility, safety, comfort, etc. Statistics from vehicles fails, accidents, number of complaints and specific surveys are used to assess the quality of the transportation service [16]. However, performing comfort statistics is an expensive task in terms of human resources and time, because of surveys and personal
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
interviews [12]. Hence, systems to collect objective data in aim of guarantee or assess safety and comfort are being proposed. Several applications have emerged to improve the quality of transportation system; those could be classified in four topics: mobility, information, security, and comfort. In the first topic, mobility solutions as Bus Rapid Transit (BRT) and subways use fleet management system, as the ITS4mobility [20], to improve the impact in urban mobility. Nowadays, information about buses routes, station localization, arrival and departure times can be accessed through web sites or phone applications, e.g. the Bus Catcher is an application developed to assists mobile users to plan their travels using a PDA, resulting in saving time [3]. Since the last decade, Global Positioning Systems (GPS) are most widely used; this technological advance became the Vehicle Tracking System in one of the most successful solution in the security field. The comfort measurement may be the most challenge topic to be measure, because it depends on the human perception and it is affected by several factors. Förstberg [6] specifies some types of human-vehicle interaction variables that influence the user’s comfort: Human Factors (as age and gender), environmental factors (as temperature, noise, and pressure), spatial factors (as workspace, leg room, seat shape, etc.) and dynamic motion factors. The ISO 5805 [9,17] defines comfort just as ‘‘subjective state of well-being or absence of mechanical disturbance in relation to the induced environment’’. In this paper, the comfort scopes will be limited to mechanical disturbances or dynamic motion factors. Usually, the quality of the comfort is evaluated using subjective rating tests [17], which may include several issues not coupled with the ride as mood, tiredness, etc. Also, the vocabulary used to rate may be different among individuals and changes in perception level or acceptance level may occur. Therefore, the objectivity of the evaluation of comfort is an interesting issue to be implemented, because, it also allows repeatability, comparison between competitors and creation of scales as mentioned by Strandemar [17]. Vehicles interaction with road will be reflected in terms of acceleration, Jerk (the rate of acceleration change) and angular motion, which depend on bus maintenance, driver behavior and road’s state. In elevators, Jerk is used to confirm motion control settings associated to ride quality [10]. Moreover, Andersson et al. [1] demonstrated the influence of the train lateral Jerk in walking, standing and sitting passenger. It is possible to assess the dynamic motion comfort using five methodologies: average ride comfort, estimated ride comfort, comfort disturbances, motion sickness and vibration effects in daily activities [6]. Reveriego [14] collects data using the acceleration, speed and position through a mobile telephone. Those data can be used to evaluate the fore-and-aft and lateral ride comfort using a methodology based on acceleration thresholds. In the same way, Zoysa et al. [22] proposes a system sensor network to monitor road surface condition. It can be used to detect vertical ride comfort using the comfort disturbance with
443
a vertical acceleration threshold. However, those solutions are not designed to process the acceleration data as average ride acceleration, which make difficult the comparison, e.g. between drivers, roads, etc. Lin et al. [12] uses average ride comfort methodology to assess the comfort. Using participatory phones to collect data, a ride index is computed in a server and matched with the transportation database. This system allowed them to compare different bus types in aim to determinate the most comfortable vehicle (bus type) and ride. However, the comfort evaluation depends on the availability voluntaries and their care for measurement, e.g. fixation technique. Also, the battery discharge and costs associated to data communication (GPRS) can discourage the voluntary users and limit the distance of the evaluation. The main contribution of this paper is the implementation of a complete system (Hardware, Firmware and Software) specially designed to assess the dynamic motion comfort in public transportation. The data reported by the system can be highly important for transportation agencies, road state surveillance authorities and efficiency of quality transport polices. The innovative work includes using buses as mobile sensors to collect, process and transport information related with comfort in public transportation (as driver’s behavior, road state, etc.). Also, previously works often used just one measurement methodology. The proposed system uses two complementary measurement methodology (average ride comfort, comfort disturbances) for an objectively comfort assessment without need additional data processing. The average ride comfort methodology follows the entire international standard ISO 2631-1 [8] for comfort measurement. The comfort disturbances detect sudden motion, which is measured through acceleration and Jerk signals above a defined threshold. The threshold value can be determined by the own system using passengers experience feedback hardware under controlled ride conditions. Those two methodologies are complementary, the first methodology provides an average measurement, which can be used to compare tracks, vehicles or drivers and it is in accordance with the standard ISO 2631-1. On other hand, the second methodology allows the source of discomfort events to be localized, geographically and temporally. In this way, algorithms using the two methodologies were created in this paper: Jerk and Acceleration Threshold Detection (JATD) and Comfort Index with Acceleration Threshold Detection (CI-ATD). These methodologies were both embedded in the system software. The hardware, described in Section 2, is based on tri-axial accelerometers, a Global Positioning System (GPS) module, wireless communication modules and user’s feedback module. The firmware, described in Section 3, is divided in the two algorithms JATD and CI-ATD. The software, described in Section 4, is composed by three parts: event’s visualization interface, set-up (configuration) interface and defects on the pavement detection interface. Also, Section 5 presents the results obtained using the system with the two algorithms, user’s feedback module and analysis of data to detect defects on the pavement. Finally, section 6 presents conclusions and discussion associated to this paper.
444
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451 Table 1 Technical specification of sensor node.
2. Hardware The system hardware was based on the former work previously described in [4]. The system of the current work also uses a sensor and base node including a new extension module called User’s Feedback Extension Module (UFE). The UFE module allows the operator to set the acceleration threshold with real user’s experience feedback for specific population e.g. elderly people, passengers with special needs, etc. The hardware description and electrical characteristics of the Embedded System for Comfort Measurement (ESCM) and UFE module are shown below.
Item
Description
Acceleration range Acceleration bandwidth Sample frequency Maximum position accuracy Wireless range Memory type Energy consumption Power supply range Size
±4 G 50 Hz, 100 Hz 500 Hz 5m 140 m (indoor) SD™ Card 250 mA 5–30 V 5 cm 9 cm
2.1. Sensor node
2.3. User’s feedback extension (UFE)
The ESCM sensor node type acquires, processes and transmits the acceleration data which is used to detect discomfort events. Also, the events generated are saved in a memory. At the end of the journey, the sensor node delivers the events data to the base node. The sensor node is installed on the vehicle with the device’s axes align with the motion axes. Fig. 1 shows the sensor node, which contains a 32-bit microcontroller, a GPS module, a digital tri-axial MEMS accelerometer, a temperature sensor, a wireless communication module to 900 MHz and finally a socket for a Secure Digital (SD) memory. Furthermore, the sensor node has two expansion ports. Table 1 describes some technical specifications of the sensor node.
The UFE module is used to determinate the acceleration threshold value according to the passenger’s perception. It is composed by a microcontroller and eight switches. Those modules are connected to the expansion port at the sensor nodes.
2.2. Base node The second ESCM device, the base node, is installed in the central station. This node transmits a broadcast signal to request the event’s transmission from the sensor node. This node is an interface between wireless module and a computer USB port, where the events are visualized and analyzed in the user’s interface based on LabVIEW™.
3. Firmware In this paper, three measurement techniques are used to assess the comfort disturbances or the comfort level. The first one uses the acceleration data to detect accelerations out of a range defined by a threshold value. The second one calculates the Jerk signal to detect Jerks values out of a range defined by a threshold value. The last technique is based on the international standard ISO 2631-1, which describes the vehicles vibration effects on the human body and it defines a comfort index (CI) based on the Root Mean Square (RMS) value from acceleration-weighted data. Using these measurement techniques, two algorithms were created to evaluate the comfort, both of them using acceleration measurement. The algorithms are: the Jerk-
Fig. 1. Sensor node.
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
Acceleration Threshold Detection (JATD) and Comfort Index with Acceleration Threshold Detection (CI-ATD). The three techniques are explained below:
3.1. Acceleration threshold detection with user’s feedback The acceleration threshold detection allows the system to detect disturbances in the comfort due to excessive acceleration. Here, the maximum absolute value of acceleration is computed each second. If this value is higher than the threshold, the firmware gets the position from the GPS module and an event structure is generated, after that the event is saved in the SD memory. The event’s structure is composed by a header, Bus Id, number, type, value and position of the event. Each event’s structure contains a field called ‘‘Event Type’’, which indicates the source of the disturbance. Table 2 shows the possible values for Event type field. The vertical, transversal and longitudinal acceleration signals from a bus journey of 21 min were measured. The signals show several disturbance events due to fast turns, abrupt starts, breaks maneuvers and imperfection or obstacles on the road. Fig. 2 shows an example of this
a
detection technique, where is detected a vertical disturbance event of two seconds produced by a speed bump. In this case, the acceleration has magnitudes higher than the threshold set to 0.5 G. 3.1.1. UFE module firmware The UFE module generates a passenger’s event each time that the user indicates a discomfort using a pushbutton. The firmware of the UFE module is based on a slave device, which holds in a register the information of the buttons pressed until the master device (sensor node) reads it. The passenger’s events are composed by information as acceleration magnitude, acceleration axes and position of the event reported by the user. 3.2. Jerk threshold detection The Jerk is defined by an international standard ISO2041 [7] as: ‘‘The resultant vector of the acceleration derivate in the time domain’’. Therefore, not only the Jerk’s magnitude varies according to acceleration magnitude, but also, it changes with the duration of the acceleration change. In order to illustrate this concept, it is modeled an acceleration signal as a sinusoidal signal (Eq. (1)):
aðtÞ ¼ A sinðxtÞ;
Table 2 Event type description. Event type
Event description
0 1 2 3 4 5 6 7 8 9
No event Longitudinal acceleration event Transversal acceleration event Vertical acceleration event Longitudinal and transversal event Longitudinal and vertical event Transversal and vertical event All axial acceleration event Passenger’s event Jerk’s event or index comfort indicatora
Type 9 changes their functionality according to the algorithm used.
445
ð1Þ
where A is the acceleration magnitude and x is the acceleration frequency. The Jerk, J(t), is obtained performing the derivation of a(t) (Eq. (2)). Note that, the Jerk’s magnitude depends on the magnitude and frequency of the acceleration signal.
JðtÞ ¼ A cosðxtÞ:
ð2Þ
The frequency spectrum of the acceleration signal is important to design the algorithm to calculate the Jerk. Some factors as filter order and cut-off frequency has to be chosen from the spectrum analysis. In this way, a Wavelet analysis was performed over the acceleration’s signals.
Fig. 2. Vertical acceleration event due to speed bump.
446
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
A Discrete Wavelet Transform (DWT) analysis using the longitudinal acceleration analysis reflects the signal due to the forward advance and the gears changing maneuver. The result shows low frequency components below to 0.3 Hz due to starting maneuver, components below to 0.5 Hz due to changing gears maneuver and higher components (5–40 Hz). Similar analyses were performed using vertical and transversal accelerations. Considering the previous results, the stop and run behavior of the cities and the experimentation in trains with passengers made by Andersson [1], it was chosen the longitudinal acceleration axis to perform the comfort analysis. Thus, it was implemented the same filter used by Andersson [1], i.e. a low-pass filter (LPF) with cut-off frequency (FC) of 0.3 Hz. The block diagram of this detection technique is shown in Fig. 3. The result obtained using the proposed algorithm over a real longitudinal acceleration signal with several start and stop vehicle actions are shown in Fig. 4. The initial threshold was set to 0.4 m/s3 according to those results, the recommendation of Förstberg [6] and the comfort’s research made by Andersson [1]. The results of Andersson showed that 30% of walking passengers perceived a discomfort above 0.4 m/s3. The JATD algorithm is composed by this detection technique and the first one.
3.3. Comfort index measurement The third technique measures the ride comfort according to an indicator called Comfort Index (CI). There are different types of CI, such as: WZ, Ride Index, NMV, NVA, NVD, PCT, PDE and ISO 2631 [6,8]. In this work, it is used the CI described by the international standard ISO 26311, which uses the Root Mean Square (RMS) value of the weighted-acceleration signal in three axes. The frequency range of measurement for comfort evaluation should be from 0.5 Hz to 80 Hz. The basic method to calculate the RMS value is described by following equation:
aW ¼
Z T 12 1 a2W ðtÞdt ; T 0
where aW is the weighted RMS acceleration value (m/s2), T is the duration of measurement in seconds and aW(t) is the weighted acceleration signal. The aW is calculated for each of the three axes, where each axis has a specific weightcurve. For lateral and transversal axes is used the transfer function xd and for vertical axis is used the transfer function xk, those transfer functions are valid for measurements on the seat or vehicle’s floor. The impulse response of xd and xk are given by the standard ISO2631-1 [8]. After the calculation of the RMS value, (based on Eq. (3)) in each axis, it is calculated the value of the total RMS acceleration using the coefficients kX, kY and kZ combining three axes. The expression used is described by Eq. (4):
12 2 2 2 aV ¼ kX a2WX þ kY a2WY þ kZ a2WZ ;
Fig. 3. Block diagram for Jerk threshold detection technique.
ð3Þ
ð4Þ
where aV is the total weighted RMS acceleration. If the comfort is evaluated for standing passengers as this case, the coefficients k will be one.
Fig. 4. Longitudinal acceleration and longitudinal Jerk of a bus journey using the Jerk detection technique.
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
In this technique, it is used the Fast Fourier Transform (FFT) to transform the acceleration signal of each axis in the time domain into the same signal in the frequency domain to apply the weight values. After that it is performed the RMS calculation each second. To perform an evaluation in a longer period of time (Tm), it is used the expression of Eq. (5), which uses the aV computed each second and the average expression for RMS values.
"
aRMST m
#12 Tm 1 X 2 ¼ a : T m i¼0 Vi
ð5Þ
Finally, it is given a comfort index (CI) according to the calculated aRMST m . The comfort index definition in Table 3 is based on the ISO standard. The CI-ATD algorithm is composed by this detection technique and the first one. 4. Software The software is based on Graphical User Interface (GUI), which was developed using the LabVIEW™ platform. This software is composed by three parts. The first one is the Event’s Visualization Interface, in which is visualized the events in an online map. The second one allows the configuration and reading of the sensor nodes. The third one is called Defects on the Pavement Detection Interface (DOPDI). It is a processing data interface, which analyze several travels in the same route to localize the repeated events in the vertical axis. This procedure allows the identification of defects on the pavement over driver’s mistakes. The DOPDI uses a radius of tolerance to recognize several nearby events as a unique discomfort source with the average position. Table 3 Comfort index definition. Index
Range (G)
Event description
0 1 2 3 4 5
Greater than 0.229 0.145–0.229 0.092–0.145 0.057–0.092 0.032–0.057 Less than 0.032
Extremely uncomfortable Very uncomfortable Uncomfortable Fairly uncomfortable A little uncomfortable Not uncomfortable
447
5. Results 5.1. Jerk-Acceleration Threshold Detection (JATD) An early test was performed to explore the maximum value of the longitudinal Jerk. This test was carried out by car, where their journeys were characterized by several breaking and starting maneuvers under moderate speed conditions. In those testes were obtained a maximum Jerk of 0.98 m/s3. The evaluation test was carried out in two urban bus lines, line A and line B. The results of the line A show the acceleration’s events with maximum amplitude of 1.86 G, the acceleration’s threshold was set to ±0.7 G; moreover, most of the acceleration’s events correspond to vertical events type. The maximum Jerk event has amplitude of 0.562 m/s3, among six Jerk’s events. The results of the second line show acceleration’s events with a maximum amplitude of 1.71 G and it detects a unique Jerk event of 0.412 m/s3. The data obtained reveals that 95% of the vertical acceleration’s events are below to 1.3 G and 1.08 G for the line A and B, respectively. It probably means that the track of the line A has a poor condition on the road than the track of the line B. The results of the second test of JATD algorithm are shown in Fig. 5 using the Event’s Visualization Interface.
5.2. Comfort Index with Acceleration Threshold Detection (CIATD) Several tests were performed to check the CI-ATD algorithm, however, in this article are presented just two of the most representative tests. The first test was carried out in two vehicles on the same route; the first vehicle was a university bus and the second one was a particular car. In the test, both vehicles used the same sensor device with the acceleration threshold set to 0.7 G and it calculates the CI each 15 min. The result of the first vehicle journey not only reveals more events than the journey made by the second vehicle; but also, it calculated a lower CI. Most of the events were vertical and longitudinal type as it is shown in Fig. 6. The lowest CI obtained in the first vehicle journey
Fig. 5. JATD Results in an urban bus lines. At the left is shown the line A results. At the right is shown the line B results. Yellow markers indicate Jerk events. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
448
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
Fig. 6. CI-ATD First test results, two vehicles at the same track. At the left, the results from a bus journey. At the right, the results taken from an automobile journey.
was two, which had a magnitude of 0.128 G; whereas, the second one was three with a magnitude of 0.082 G; those CIs follow the scale shown in Table 3. Also, the test localized a street in poor maintenance state detected by both vehicles, which presented high density of acceleration events. The second test of the CI-ATD was performed on an urban bus line of Campinas, Brazil. The test was carried out in real situation with passenger’s loads. The journey had a duration of 46 min and around of 20.4 km distance. The data obtained was analyzed by each motion axis (transversal (X), longitudinal (Y) and vertical (Z) axes). Also, mixed events (XY, XZ and YZ) were analyzed as isolated events in longitudinal, transversal and vertical axes. Those results are shown in Fig. 7 in terms of the occurrence frequency and magnitude.
Also the comfort indexes obtained each 5 min were analyzed. The results revels an average comfort index equals to 2.3, which is uncomfortable according to Table 3. 5.3. User’s feedback results using the UFE module The follow test illustrates the methodology to set up the acceleration threshold based on the user’s experience of a specific population group, as elderly, visual deficient, students’ passengers, etc. Those subjects registered their opinion about the ride through a push-button action. The test is composed by three laps at the same road of 2.8 km, which in the first lap the driver limited the average speed to 30 km/h and when the vehicle passed over a speed bump the speed was reduced to 10 km/h. In the second lap was limited to 50 km/h and 20 km/h, respectively. The last lap
Fig. 7. Acceleration events analysis of bus journey in the same track (longitudinal, transversal and vertical events).
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
was performed with an average speed of 60 km/h and obstacle speed of 40 km/h. This test was carried out in a particular car with two test’s subjects and it was repeated three times with different passengers. Since, the test’s subjects were an unclassified group of six people, the sample size is not enough for statistical significance of the results, but it is enough to show the purpose of the user’s feedback function. The transmission device was installed on the floor of the vehicle and it was set with the CI-ATD algorithm, which was programmed with a low acceleration threshold of 0.5 m/s2. Fig. 8 shows the route of the test and the events detected by the sensor node due to the poor state of the road, the driver style or the user’s perception registered through one button. The analysis of the events was performed according to the acceleration type (vertical, longitudinal and
449
transversal). The user’s perception events were compared with their respective acceleration events. Fig. 9 shows the results for this comparison, which reveals that 70% of the users perceived a disturbance with a maximum acceleration of 0.7 G, 0.8 G and 1.4 G for lateral, longitudinal and vertical events, respectively. 5.4. Defects on the Pavement Detection Interface (DOPDI) The DOPDI software was test performing four travels by particular car at the same track to show its functionality. In those testes was used the CI-ATD algorithm. The tolerance radius was set to 10 m. After the DOPDI program was executed using the data of the four travels, it produces the output file with the position of the vertical repeated event markets. The unified vertical events contain informa-
Fig. 8. Test track with several speed bumps and one roundabout.
Fig. 9. User’s perception events and CI-ATD events comparison detected using CI-ATD with UFE module.
450
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
Fig. 10. DOPDI analysis results using tolerance radius of 10 m.
tion about the quantity of the repeated events and its average position. Fig. 10 shows the result for this analysis, where several vertical discomfort sources were detected.
6. Conclusions This paper presented a successful system to localize comfort disturbances using the public transportation system to collects, process and transmit the relevant information. This information allows the detection of dangerous driving style behaviors such as excessive accelerations. This system was successfully tested in real conditions and also used to determinate the sensibility of the passengers to the acceleration magnitude and the pavement defects on the road. This information is important for both transportation and governmental agencies in aim to improve the service and the road’s state. Two algorithms using three techniques to assess the comfort in a journey were presented in this article. The first technique, the acceleration threshold detection with user’s feedback, is appropriate to localize geographically the discomfort sources due to excessive accelerations. The second technique, Jerk threshold detection, adds other criteria to the detection of discomfort’s events based on the longitudinal acceleration changes (Longitudinal Jerk). This technique is adequate to assess heavy traffic or transportation systems with unscheduled starts-stops. The rating output depends on the quantity of events and its magnitude, hence, it is difficult to compare among different journeys and establish which was the most comfortable. The low frequency characteristic of Jerk signal impacts negatively in the localization accuracy, also, the Jerk magnitude depends on the acceleration’s signal to noise rate (SNR). Therefore, filter influences in the Jerk computation. Some previously researches [1,2,6,13], that involve Jerk do not define a unique method to calculate it from a vehicle’s acceleration. Hence, most of times, it is impossible to compare the results obtained in those studies among them. In this way, it is essential to establish a standard way
to calculate the Jerk in comfort’s applications following similar parameters on the filtering stage. The obtained results using this algorithm has Jerk magnitudes similar to the results obtained by Förstberg and Andersson [1,6] using similar filtering parameters. However, Bagdadi [2] shows interesting results using an alternative Jerk algorithm that can be implemented by ESCM. The comparison concern is solved using the technique, called Comfort Index Measurement. Therefore, the comparison among drivers or journeys becomes into a simple numerical comparison. Although, using just this technique, the geographic location of discomfort sources is not possible. Hence, this technique is merged with the first technique (CI-ATD). For a simple test, using UFE module, the user’s sensibility is lower for the vertical axis than longitudinal and lateral axis of acceleration, but this information has to be tested with more subjects before to turn into valid result, statistically. This test could be used to determine the acceleration or Jerk threshold value in specific populations as children, pregnant, elderly or people with special necessities, which information can be really interesting to improve the quality of special transport services e.g. scholar buses. Implementing a specific hardware instead of using the embedded sensors in mobile devices allows independence on harvesting data from the user’s availability. Also, it ensures the quality of the data gathering necessary to accomplish the ISO2631 recommendations and decreases the quantity of data processed due to their on-board signal processing. Acknowledgements The authors acknowledge the Brazilian National Council of Scientific and Technological Development – CNPq under Universal Project No. 480864/2011-0 and CAPES for the financial support. Also, the authors acknowledge the bus Company Itajaí in Campinas City, São Paulo State, Brazil, for the test vehicle provided and the team of Microelec-
J.C. Castellanos, F. Fruett / Measurement 47 (2014) 442–451
tronic Sensors Laboratory staff, LSM, at the UNICAMP for the academic support. References [1] E. Andersson, N. Nilstam, Development of advanced high speed vehicles in Sweden, Proceeding of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering 198 (1984) 229–237. [2] O. Bagdadi, A. Várhelyi, Development of a method for detecting Jerks in safety critical events, Accident Analysis & Prevention (2012). [3] M. Bertolotto, G. O’Hare, R. Strahan, A. Brophy, A. Martin, E. McLoughlin, Bus catcher: a context sensitive prototype system for public transportation users, in: Proceedings of the Third International Conference on Web Information Systems Engineering (Workshops), 2002, pp. 64–72. [4] J.C. Castellanos, A.A. Susin, F. Fruett, Embedded sensor system and techniques to evaluate the comfort in public transportation, in: 2011 14th International IEEE Conference on Intelligent Transportation Systems (ITSC), 2011, pp. 1858–1863. [5] DETR, A new deal for transport: better for everyone (white paper), 1998. [6] J. Förstberg, Ride Comfort and Motion Sickness in Tilting Trains, Department of vehicle engineering, Royal Institute of Technology (KTH), Stockholm, Sweden, 2000. [7] ISO, Vibration and Shock – Vocabulary, ISO 2041, 1990. [8] ISO, Mechanical Vibration and Shock – Evaluation of Human Exposure to Whole-Body Vibration – Part 1: General Requirements, ISO 2631–1, 1997a. [9] ISO, Mechanical Vibration and Shock – Human Exposure – Vocabulary. ISO 5805, 1997b. [10] ISO, Lifts (elevators) – measurement of lift ride quality, ISO 18738, 2003. [11] P. Kelly, Quality bus transit systems, in: Public Transport Electronic Systems, International Conference on (Conf. Publ. No. 425), 1996, pp. 11–15.
451
[12] C.-Y. Lin, L.-J. Chen, Y.-Y. Chen, W.-C. Lee, A comfort measuring system for public transportation systems using participatory phone sensing, in: Proceedings of PhoneSense 2010, Zurich, Switzerland, 2010. [13] J.J. Martinez, C. Canudas-de-Wit, Model reference control approach for safe longitudinal control, in: American Control Conference, Proceedings of the 2004, vol. 2753, 2004, pp. 2757–2762. [14] G. Reveriego, C. Urdiales, F. Sandoval, Vehicle risks assessment using a nokia N95, in: Intelligent Transportation Systems, 2009. ITSC ‘09. 12th International IEEE Conference on, 2009, pp. 1–6. [15] J.-P. Rodrigue, C. Comtois, B. Slack, The Geography of Transport Systems, second ed., Routledge, London, New York, 2009. [16] M.O. Rodrigues, An Evaluation of the Level of Service in the Public Transportation System of the City of São Carlos, School of Engineering of São Carlos, University of São Paulo, São Carlos, Brazil, 2006. [17] K. Strandemar, On Objective Measures for Ride Comfort Evaluation, Department of Signals, Sensors and Systems. Royal Institute of Technology (KTH), Stockholm, Sweden, 2005. [18] A. Tennøy, Why we fail to reduce urban road traffic volumes: does it matter how planners frame the problem?, Transport Policy 17 (4) (2010) 216–223 [19] TRB, Highway Capacity Manual, fifth ed., Transportation Research Board, Washington, DC, 2010. [20] Volvo, Its4mobility, in: V.B. Corporation (Ed.), Technical Description, Göteborg, Sweden, 2009. [21] S. Ye, Research on Urban road traffic congestion charging based on sustainable development, Physics Procedia 24 (Part B(0)) (2012) 1567–1572. [22] K.D. Zoysa, C. Keppitiyagama, G.P. Seneviratne, W.W.A.T. Shihan, A public transport system based sensor network for road surface condition monitoring, in: Proceedings of the 2007 Workshop on Networked Systems for Developing Regions, ACM, Kyoto, Japan, 2007, pp. 1–6.