Development of simulated driving cycles for light, medium, and heavy duty trucks: Case of the Toronto Waterfront Area

Transportation Research Part D 34 (2015) 255–266

Contents lists available at ScienceDirect

Transportation Research Part D journal homepage: www.elsevier.com/locate/trd

Development of simulated driving cycles for light, medium, and heavy duty trucks: Case of the Toronto Waterfront Area Glareh Amirjamshidi 1, Matthew J. Roorda ⇑ Department of Civil Engineering, University of Toronto, M5S 1A4, Canada

a r t i c l e

i n f o

Keywords: Driving cycle Microsimulation modeling Emission model MOVES

a b s t r a c t Driving cycles are an important input for state-of-the-art vehicle emission models. Development of a driving cycle requires second-by-second vehicle speed for a representative set of vehicles. Current standard driving cycles cannot reflect or forecast changes in traffic conditions. This paper introduces a method to develop representative driving cycles using simulated data from a calibrated microscopic traffic simulation model of the Toronto Waterfront Area. The simulation model is calibrated to reflect road counts, link speeds, and accelerations using a multi-objective genetic algorithm. The simulation is validated by comparing simulated vs. observed passenger freeway cycles. The simulation method is applied to develop AM peak hour driving cycles for light, medium and heavy duty trucks. The demonstration reveals differences in speed, acceleration, and driver aggressiveness between driving cycles for different vehicle types. These driving cycles are compared against a range of available driving cycles, showing different traffic conditions and driving behaviors, and suggesting a need for city-specific driving cycles. Emissions from the simulated driving cycles are also compared with EPA’s Heavy Duty Urban Dynamometer Driving Schedule showing higher emission factors for the Toronto Waterfront cycles. Ó 2014 Elsevier Ltd. All rights reserved.

Introduction Driving cycles are an important input for state-of-the-art vehicle emission models. Driving cycles have been developed for various cities and vehicle types. A driving cycle is a representative speed-time profile for a study area within which a vehicle can be idling, accelerating, decelerating, or cruising. However, speed-time profiles vary across cities due to each city’s unique topography and road driving behavior and they have been shown to vary by vehicle type, time of day and type of road (Ericsson, 2001; Hung et al., 2007; Kamble et al., 2009; Saleh et al., 2009; Yu et al., 2010). Two categories of driving cycles can be found in the literature: First, synthesized (or model) driving cycles which are built from combining different phases of idling, constant acceleration/deceleration and steady speed. Examples include the European cycle (NEDC) and the Japanese cycle (J10-15). However unrealistic transition between the different phases in these driving cycles could result in erroneous emission estimation (Chugh et al., 2012; Kamble et al., 2009; Pelkmans and Debal, 2006; Tong and Hung, 2010; Weiss et al., 2011). Second, real world driving cycles (or transient driving cycles) are developed by recording speed-acceleration profiles while driving on the real world roadway network. In other words, these cycles are synthesized from real-world speed data. Examples include FTP-75 in the US, and driving cycles for Pune (Kamble et al., 2009), and Hong Kong (Hung et al., 2007). ⇑ Corresponding author. Tel.: +1 416 978 5976; fax: +1 416 978 6813. 1

E-mail addresses: [email protected] (G. Amirjamshidi), [email protected] (M.J. Roorda). Tel.: +1 416 978 5049; fax: +1 416 978 6813.

http://dx.doi.org/10.1016/j.trd.2014.11.010 1361-9209/Ó 2014 Elsevier Ltd. All rights reserved.

256

G. Amirjamshidi, M.J. Roorda / Transportation Research Part D 34 (2015) 255–266

No real world driving cycle has been developed for the Toronto area. The only recent driving cycles developed for the Toronto area were synthesized using CALMOB6 reflecting average speeds from a travel demand model using the Transportation Tomorrow Survey (TTS) (Raykin et al., 2012). The objective of this paper is to develop representative simulated driving cycles, using simulated data, for different combinations of roadway class, time of day and vehicle attributes. There are several applications for a disaggregate set of driving cycles. Emissions and fuel consumption impacts of changing congestion patterns, peak spreading, new infrastructure, and vehicle specific driving behavior could be better addressed. The emissions benefits of new vehicle technology could be assessed more specifically for different roadway types (e.g. when and where would the greatest benefits of plug-in hybrid electric vehicles be attained?). Vehicle routing algorithms could be developed that include congestion-sensitive fuel consumption and emissions in the objective function, for example, to service congested areas with lower emitting vehicles during off-peak hours.

Background and motivation A driving cycle is made up of micro-trips where a micro-trip is defined as the trip between two idling periods. A driving cycle is usually for a 10–30 min interval, which is long enough to contain enough micro-trips to reflect the diversity of real world driving behavior, but short enough to be practical and cost effective in terms of data collection (Hung et al., 2007; Lai et al., 2013; Yu et al., 2010). Development of a driving cycle generally involves three steps: test route selection, data collection, and cycle construction. Test route selection involves selecting the route on which data are to be collected. The data collection step generally involves the collection of the speed of a sample of vehicles at frequent time intervals (usually on a second-by-second basis) using the car-following method, or on-board measurement using GPS devices (Chugh et al., 2012; Coelho et al., 2009; Green and Barlow, 2004; Han et al., 2012; Hung et al., 2007; Kamble et al., 2009; Prasad et al., 2012; Shahidinejad et al., 2010; Sriniwas et al., 2011; Tzirakis et al., 2006; Wang et al., 2008; Yu et al., 2010). Cycle construction consists of 5 steps: (1) define the set of assessment measures used to describe a driving cycle; (2) calculate the assessment measures for the collected data (called target statistics); (3) develop a candidate driving cycle from the pool of micro-trips available (called candidate cycle); (4) calculate the same assessment measures for the candidate cycle (called test statistics); and (5) identify the candidate cycle whose test statistics are closest to the target statistics. The studies in the literature are distinguished based on how the candidate driving cycles are developed, and the set of assessment measures used for comparison. Most of the studies in the literature have used random selection of micro-trips as the method for producing a candidate cycle (Hung et al., 2007; Kamble et al., 2009; Wang et al., 2008; Xiao et al., 2012). A few studies have used the driving data clustering method (Fotouhi and Montazeri-Gh, 2013), where micro-trips are first divided into different categories (or clusters) based on their ‘‘traffic conditions’’ (e.g., congested, free flow) and driving cycles are then developed for each cluster. This method is similar to the random selection method, since clustering only results in developing unique driving cycles for each road type or traffic condition. Yu et al. (2010) used a genetic algorithm for part of the cycle development, where micro-trips are first sorted based on their assessment measures. Then the top 20% of microtrips are selected and a lower and upper limit for the number of required micro-trips in a driving cycle is estimated. A genetic algorithm is then used to develop candidate driving cycles for each number of micro-trips in the cycle (between the upper and lower level). Finally, the cycle with the best performance according to an assessment measure is selected as the final driving cycle. Although this method improves slightly upon the random selection method, the large computational time for a 20% sample of the micro-trips is prohibitive. Consequently, in this research, the random selection method is chosen. The other difference between studies is the type and number of assessment measures used. Driving activity measures and the Vehicle Specific Power (VSP) method have both been used in the literature. Activity measures refer to statistics like speed and acceleration (Hung et al., 2007; Wang et al., 2008). The VSP method, which is a bin based method, focuses on instantaneous power per unit of a vehicle and is a nonlinear function of instantaneous speed, instantaneous acceleration and road grade (Coelho et al., 2009; Yu et al., 2010). The driving activity measures method has been used in this research for cycle construction as it is used more extensively by researchers in the area making the results comparable to the existing body of literature. The method is described in detail in ‘‘Methodology for developing the driving cycles’’. This research is motivated by the fact that collection of real world data to develop driving cycles for different vehicles and road types, for a large enough representative sample of vehicles, would either be too costly or biased (if data were collected on a day with unusual congestion patterns). A new method for developing simulated driving cycles is introduced using traffic simulation. This concept has also been studied by Della Ragione and Meccariello (2010), who evaluated the ability of four car-following models to produce simulated driving cycles using data from four vehicles equipped with GPS in the Naples Metropolitan Area. Their results showed that most of the models produce driving cycles and emission values close to the observed; however concluded that further calibration and investigation is required. Simulated driving cycles use simulated data, collected for all vehicles under consistent and calibrated traffic conditions, from a microscopic traffic simulation for cycle development. In other words, the test route selection step is not necessary for simulated driving cycles, since data can be collected on all routes within a desired roadway classification. Using data from multiple simulation replications also accounts for stochastic variations in traffic conditions, allowing a driving cycle to be more representative. It also allows for analysis of the changes to the driving cycle as a result of future traffic conditions

G. Amirjamshidi, M.J. Roorda / Transportation Research Part D 34 (2015) 255–266

257

or infrastructure or technology changes. However, the quality of the outcomes of the emissions model are directly tied to the quality of the microsimulation, requiring a robust and well-calibrated model, which is discussed in the next section. Model setup Study area In this research, the method of developing a simulated driving cycle is demonstrated to develop road-specific driving cycles for light duty trucks (LDT), medium duty trucks (MDT), and heavy duty trucks (HDT) in the Toronto Waterfront Area (Fig. 1). This area consists of the central business district of Toronto and inner urban areas to the east and west. The network includes arterial, collector and local roads and two freeways that play an important role in transporting goods and people to and from the downtown. Details about how the network was coded, and the demand estimation can be found in (Abdulhai et al., 2002; Roorda et al., 2010, 2011). The network, coded in Paramics V6.9.1, consists of 4012 roadway links, 1841 nodes, 44 internal zones, 35 external gateways, 227 signals, approximately 26 km of freeway, and about 471 km of total road. The network presents an area of about 27 km2. The results of the demand estimation showed that a total of 54,916 passenger vehicles, 4129 LDTs, 1518 MDTs, and 666 HDTs traverse the network between the hours of 8:00 and 9:00 AM. Model calibration and validation Using default parameters for microsimulation models tends to produce unrealistic driving behavior that is too aggressive, with higher acceleration and deceleration rates than are observed in reality (Hallmark and Guensler, 1999; Manjunatha et al., 2013; Zhang et al., 2012). To address this shortcoming, default parameters for microsimulation models should be calibrated to accurately reproduce vehicle dynamics including speed, acceleration and deceleration (Int Panis et al., 2006; Song et al., 2011; Younglove et al., 2005). The network was calibrated using traffic counts, speed measurements and acceleration/deceleration data available between September 2008 and 2009. The model was calibrated for the AM peak hour using a simple genetic algorithm (GA) with the multi-objective function minimizing the Root Mean Square of Errors (RMSE) between model estimates and field measurements for counts, speeds, and standard deviation of acceleration. The model was also validated against observed probe data driving cycles on freeways (using the thirteen driving cycle attributes discussed in Section 4), and their cycle emissions (using MOVES2010b). Details about the calibration and validation can be found in (Amirjamshidi, 2015). Methodology for developing the driving cycles The method used for developing simulated vehicle and/or road-specific driving cycles is shown in Fig. 2. The inputs are the second-by-second speed, location, and vehicle type information of all vehicles from the simulation. The data are first categorized based on road and vehicle type (LDTs, MDTs, and HDTs). The following five road categories are observed in this network based on their speed limits and driving behavior: Freeway, Lake Shore Blvd., University Avenue, Major arterial, and Major arterial with transit. Target statistics are calculated using micro-trips for each vehicle-road type combination. Then a test cycle is generated by appending randomly selected micro-trips until the total test cycle duration is between 10 to 30 minutes and the test statistics are calculated (as shown in the dotted rectangle in Fig. 2); and if it passes all criteria with an acceptable threshold of 15%, it is accepted as a candidate driving cycle. The reason for selecting a 15% threshold was twofold: first to ensure that each

Fig. 1. Toronto Waterfront network.

258

G. Amirjamshidi, M.J. Roorda / Transportation Research Part D 34 (2015) 255–266

Fig. 2. Method for the development of a simulated driving cycle.

assessment measure of the candidate driving cycles is within 15% of the target measure; and second to allow for having multiple candidate driving cycles for each vehicle type. The objective is to find the best driving cycle that minimizes the difference between the drive cycle statistics and the target statistics. The set of assessment measures selected for this study are adapted from those used in the literature (Hung et al., 2007; Saleh et al., 2009; Wang et al., 2008). Table 1 shows the 13 assessment measures used to compare attributes of a candidate driving cycle to the mean attributes of all micro-trips. The stopping criteria are satisfied when either 20 candidate cycles are identified, or a maximum of 10,000 test cycles are tested. Under these criteria, there will generally be more than one candidate driving cycle and hence another goodness-of-fit measure is required to select the best candidate as the final driving cycle. The Performance Value (PV) (Eq. (1)), also adapted from Hung et al. (2007) and Saleh et al. (2009), represents the sum of the normalized absolute difference between the test statistics and the target statistics.

Table 1 Assessment measures used for cycle construction. Parameter

Abbreviation

Unit

Average speed Average running speed (excludes points where speed is equal to zero) Average acceleration Average deceleration Time proportion of driving modes in idling Time proportion of driving modes in accelerating Time proportion of driving modes in decelerating Time proportion of driving modes in cruising Time proportion of driving modes in creeping (speed <4 kph) Average micro-trip duration Average percentage of acceleration-deceleration changes Root mean square acceleration Root mean square of positive kinetic energy over weight

V Vr A D Pi Pa Pd Pcrs Pcrp D Acc-dec RMSA RMSPKE

m/s m/s m/s2 m/s2 % % % % % Sec % m/s2 m/s

G. Amirjamshidi, M.J. Roorda / Transportation Research Part D 34 (2015) 255–266

PV ¼

X k

  ðtest statk Þ  ðtar statk Þ  for k ¼ 1; . . . ; 13 wk   ðtar statk Þ

259

ð1Þ

In this equation test statk is the kth assessment measure for the candidate driving cycle, tar statk is the kth target assessment measures, and wk shows the importance of the kth measure relative to the other 12. Current literature on driving cycles considers all measures equally important (wk = 1) (Hung et al., 2007; Kamble et al., 2009; Saleh et al., 2009; Yu et al., 2010). The PV is calculated for all candidate driving cycles, and the candidate with the lowest PV is selected as the final driving cycle for the particular vehicle/road type.

(a) Speed (kph)

80 60 40 20 0 0

200

300

400

500

600 700 800 Time (sec)

900

1000 1100 1200 1300 1400 1500

80 Speed (kph)

(b)

100

60 40 20 0 0

100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 Time (sec)

Speed (kph)

(c)

Speed (kph)

(d)

100 80 60 40 20 0 0

100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 Time (sec)

0

100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800

50 40 30 20 10 0 Time (sec)

Speed (kph)

(e)

Speed (kph)

(f)

50 40 30 20 10 0 0

100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 Time (sec)

0

100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 Time (sec)

60 50 40 30 20 10 0

Fig. 3. Simulated driving cycles developed for (a) HDT Freeways, (b) MDT Freeways, (c) LDT Freeways, (d) HDT Major Arterials, (e) MDT Major Arterials, and (f) LDT Major Arterials.

260

G. Amirjamshidi, M.J. Roorda / Transportation Research Part D 34 (2015) 255–266

Table 2 Assessment measures for the developed driving cycles. Assessment measure

Freeway HDT

MDT

LDT

Lake Shore Blvd. HDT

MDT

LDT

HDT

University Ave. MDT

LDT

HDT

Major arterials MDT

LDT

HDT

Arterials with Transit MDT

LDT

V (kph) Vr (kph) Acc (m/s2) Dcc (m/s2) D (sec) Acc-dcc (%) Pa (%) Pd (%) Pi (%) Pcrs (%) Pcrp (%) RMSA (m/s2) RMSPKE (m/s)

40.9 41.9 0.147 0.282 159 15 62.8 32.8 1.9 2.5 2.9 0.324 9.17

39.7 40.6 0.135 0.336 132 16 69.4 27.8 1.6 1.2 3.5 0.344 8.82

52.7 54.0 0.293 0.544 112 17 61.2 33.0 2.1 3.8 3.5 0.597 11.34

28.4 33.3 0.280 0.605 33 20 59.3 27.5 11.7 1.4 23.7 0.571 7.04

25.7 29.4 0.265 0.647 38 19 62.8 25.7 10.4 1.2 23.5 0.521 6.44

34.8 40.6 0.569 0.752 30 28 45.7 34.6 11.5 8.2 24.1 0.940 8.87

12.4 16.7 0.403 0.598 21 25 46.0 31.1 22.3 0.6 39.2 0.583 3.47

13.1 17.0 0.376 0.629 19 26 50.4 30.1 19.1 0.4 39.4 0.637 3.67

13.9 18.5 0.598 0.747 18 29 44.3 35.5 19.5 0.7 41.6 0.815 4.07

16.6 20.5 0.398 0.585 22 25 49.3 33.5 15.6 1.6 31.8 0.603 4.37

15.2 19.3 0.360 0.647 23 25 52.2 29.1 17.7 1.0 35.6 0.636 4.07

18.4 23.5 0.580 0.705 20 32 43.8 36.1 17.1 3.0 40.4 0.824 5.27

16.9 21.6 0.405 0.557 26 31 43.4 31.6 18.6 6.4 34.4 0.534 4.40

15.8 20.1 0.355 0.590 20 31 48.0 28.9 17.5 5.7 38.6 0.539 4.25

19.4 23.6 0.572 0.572 22 36 36.8 36.8 14.1 12.3 28.4 0.704 4.89

Results Toronto simulated driving cycles Based on the method described in Section 4, simulated driving cycles were developed using second-by-second speed data from 15 simulation runs for LDTs, MDTs, and HDTs on each road category. As an example, Fig. 3 shows the simulated driving cycles on freeways and major arterials for each vehicle category. The assessment parameters for all the simulated driving cycles are presented in Table 2. The following observations can be made from comparing the effect of road type on a vehicle’s driving cycle. 1. Average speed is higher on the freeway, then Lake Shore Blvd. compared to the arterials, as expected. 2. Arterials have higher accelerations compared to the freeway and Lake Shore Blvd. 3. Lake Shore Blvd. and all the arterials have higher decelerations, which reflects more aggressive driving and more interactions with other vehicles. 4. Average micro-trip duration for the freeway cycle is much longer than the other driving cycles, mainly because there are no signals; and stops would only be a result of congestion. 5. Longer time spent idling, lower speeds, and significantly less cruising time on University Ave suggests a higher level of traffic on this street. 6. The freeway cycle is smoother than other cycles (higher speeds, longer micro-trip duration, less time spend creeping and idling; and lower acceleration and deceleration values). However all freeway cycles show bottlenecks during the AM peak as part of the cycle that is more congested than the rest of the cycle (e.g. the bottleneck in the area of Gardiner expressway and Spadina Ave.). 7. The effect of signals can also be seen in the cycles (more stops on University Avenue and other arterials). 8. For all vehicle classes, the cycle on University Avenue is more aggressive. This could be because University Avenue has twice the number of lanes of other arterials; so drivers would want to drive at higher speeds. However there are lights and traffic; requiring drivers to slow down that might result in more aggressive driving. It also has higher acceleration and deceleration values, and the lowest cruising time. 9. Driving cycles for arterials with and without transit are similar. This could be due to the fact that in the AM peak hour; both road categories are at capacity. The only major difference between the two cycles is in the percent of time spent cruising, which is higher on arterials with transit. This can be a reflection of how vehicles adjust their speeds and behavior with the street cars, hence reducing the vehicle’s aggressiveness. Although cycles for arterials with and without transit are similar in many ways; it was however decided not to combine the two road categories due to the following reasons: (1) there are still differences in the assessment measures for the two road types, (2) emission analysis showed that they produce different CO2 emission factors, and (3) it would be expected that the two road categories show more difference for off peak hours. The following results are noted from comparing the driving cycles for different vehicle classes: 1. In all cases speed and running speed for LDTs is the highest of the three; with a very small difference between the average speeds of HDTs and MDTs. 2. LDTs have much higher average deceleration compared to MDTs and HDTs on freeways. This difference is less pronounced on Lake Shore Blvd., University Avenue and major arterials, and is small on arterials with transit.

G. Amirjamshidi, M.J. Roorda / Transportation Research Part D 34 (2015) 255–266

261

Table 3 LDT freeway cycles vs. available international light duty freeway driving cycles.

1 Not all 13 statistics were used in Wang et al. (2008), or defined the same way in Hung et al. (2007); therefore, only some parameters are used for comparison.

3. A similar relationship is noted for average acceleration, except that differences in average acceleration are greater than differences in average deceleration. 4. In most cases; average micro-trip duration is longest for HDTs, followed by MDTs and LDTs suggesting that LDTs drive more aggressively compared to other truck classes. 5. For all cases; the percent of time in cruising mode is highest for LDTs, followed by HDTs. This likely occurs because LDTs accelerate more quickly, leaving more time for vehicle cruising. Comparison of the LDT Toronto driving cycles with international cycles The developed LDT freeway simulated driving cycles are compared against two categories of driving cycles. One category includes other real world driving cycles representing peak hour driving of light duty vehicles in other cities, for which most of the assessment measures are reported in the literature (Hung et al., 2007; Wang et al., 2008). Freeway cycles are also compared to the Artemis freeway cycle2 (André, 2004; Dieselnet, 2011), the high and extra-high phases of the Worldwide Harmonized Light Vehicles Test Cycle (WLTC)3 class 2 and class 3 (Tutuianu et al., 2013), and the Highway Fuel Economy Driving Schedule (HWFET) developed by the EPA to measure the fuel consumption of light duty vehicles (U.S. Environmental Protection Agency, 2010). The following observations can be made (from Table 3): First, compared to the HWFET and the Artemis freeway cycles, Toronto freeways experience lower speeds, shorter micro-trips, and longer idling and creeping periods representing more congestion on the Toronto freeways which would be expected since the HWFET and Artemis cycles was not prepared specifically for peak traffic conditions. Second, higher values for average deceleration and RMSA, time proportions spent accelerating, decelerating, and the shorter cruising period suggests more aggressive driving for the Toronto freeway cycle compared to the HWFET cycle. These differences are smaller between the Toronto and Artemis freeway cycles; suggesting that the Artemis cycle’s aggressiveness is closer to the Toronto cycle. Third, compared to the WLTC class 2 and 3 cycles, the Toronto cycle has lower speeds compared to WLTC class 3 and similar speeds compared to class 2 cycles. This suggests that the high phase in the WLTC class 2 cycle is more representative of the Toronto freeway level of congestion during peak traffic. However, higher values for RMSA, lower average deceleration and longer time spent decelerating, shorter micro-trip duration, along with shorter time idling and cruising, suggests that the Toronto cycle is more aggressive than the high phase of the WLTC class 2. A similar difference in level of aggressiveness is observed between the Toronto freeway cycle and the high and extra high phases of WLTC class 3, with the difference that the WLTC class 3 cycle is less congested than the Toronto cycle given its higher speeds. Fourth, compared to other real world cycles developed specifically for the peak traffic conditions in China and Hong Kong, higher speeds and less time spent idling suggest less congestion in the Toronto freeway cycle. Lastly, lower average acceleration and longer time spent accelerating, along with an average deceleration and time proportion spent decelerating that 2 The Artemis driving cycles were developed based on statistical analysis of data collected over 10 years of real world driving of European vehicles. The cycles were developed within the Assessment and Reliability of Transport Emission Models and Inventory Systems (ARTEMIS) project representing urban, rural and freeway driving patterns. 3 The WLTC cycles are developed by the UN-ECE working party on pollution and energy transport program (GPRE) for estimating emissions and fuel consumptions of light duty vehicles. It consists of three test cycles based on the vehicle’s power-to-mass ratio. The cycles represent driving on urban, extra urban, freeway, and motorways through low, middle, high, and extra high phases (Dieselnet, 2012). In this study the low and middle phases are compared against Toronto urban cycles; and the high and extra high phases against Toronto freeway cycles.

262

G. Amirjamshidi, M.J. Roorda / Transportation Research Part D 34 (2015) 255–266

are in the same range of the other cities, suggests that the Toronto freeway cycle is less aggressive than other real world cycles shown in Table 3. Simulated LDT cycles for the other road categories are also compared against real world driving cycles representing peak hour driving for arterial cycles, Artemis urban cycle, WLTC class 1, the low and middle phases of the WLTC class 2 and class 3, and test cycles used in the US for vehicle emission testing (Table 4). Comparing the Asian cycles with LDT urban cycles shows that, in general, the simulated LDT cycles have higher average deceleration (except compared to Shanghai); lower speed, and higher average acceleration compared to Beijing, Chongqing, and Tianjin; and smaller acceleration than Shanghai, Chengdu and Hong Kong. These along with the fact that the Toronto cycles spent a longer percentage of time in the accelerating, decelerating and creeping modes, and less time in the idling mode; suggests that the Toronto cycles are more aggressive and experience more stop-and-go traffic which could be explained by the fact that downtown Toronto consists of many signalized intersections that are very close to one another. The Lake Shore Blvd. cycle is compared against the Hong Kong cycle; since they have very similar speeds. The comparison shows that other than the smaller average deceleration, the Hong Kong cycle has a longer average micro-trip duration, much smaller percentage of time spent in the creeping mode, and a smaller root mean of square acceleration. These suggest that the Hong Kong cycle is less aggressive and that the Lake Shore Blvd. cycle is probably more affected by the signals and its resulting bottlenecks which result in a more stop and go traffic in parts of the cycle, and less congested traffic after the signal (hence similar values of speeds). There are also observable differences between the simulated Toronto cycles and US cycles used for emission testing. Given that the NYCC cycle has lower average speeds, higher accelerations, and in most cases lower decelerations and higher average micro-trip duration, longer time spent idling, relatively shorter time spent cruising or accelerating, and a longer time spent creeping suggests that the NYCC cycle represents similar behavior but more congested traffic conditions than the Toronto simulated cycles. Compared to the simulated driving cycles, the UDDS cycle has higher speed values, longer average micro-trip duration, smaller average acceleration and deceleration values, smaller RMSA value, and shorter percentage of time spent in the creeping mode, suggesting that the UDDS cycle in less aggressive and less congested than the peak Toronto cycles. Similar to the UDDS cycle, the US06 cycle is less congested allowing for higher average speeds and average micro-trip duration. However higher average acceleration and deceleration values, longer amount of time spent accelerating and decelerating, along with a higher RMSA values suggests that the US06 cycle is more aggressive than the Toronto LDT cycles. The LA92 cycle shows higher values of speed, acceleration, much longer average micro-trip duration, and time spent cruising. These along with a smaller value of RMSA and a similar deceleration value suggests that the LA92 cycle represents less congested, but not necessarily less aggressive driving than the Toronto cycles. The main difference between the three WLTC cycles is their level of aggressiveness. The WLTC class 3 cycle is the most aggressive followed by the class 2 cycle and the class 1 cycle. Comparing the WLTC cycles with Toronto simulated LDT urban cycles shows that, in general, the Toronto cycles have lower average speeds (except for Lake Shore Blvd.), higher accelerations, much higher decelerations, higher values of RMSA, and lower average micro-trip duration. This shows that the Toronto cycles are more aggressive compared to the WLTC cycles. The comparison also suggests that the WLTC cycles are more congested than the Lake Shore Blvd. cycle, but less congested than other Toronto urban cycles. Lastly, the Artemis urban cycle’s average and running speeds are close to Toronto cycles on major arterials with and without transit. However, the Artemis cycle’s higher acceleration and deceleration, and shorter time spent accelerating suggests that the Artemis cycle is more aggressive than the Toronto cycles. MDT and HDT vs. the Heavy Duty Urban Dynamometer Driving Schedule In this section the simulated HDT and MDT driving cycles are compared against the Heavy Duty Urban Dynamometer Driving Schedule (HD-UDDS), developed by the EPA, and used for estimating tailpipe emissions for urban heavy duty driving (U.S. Environmental Protection Agency, 2010). Parts of this cycle reflect highway driving; however most of the cycle still represent arterial driving. Therefore the comparison focuses on the differences between the HD-UDDS cycle and simulated arterial driving cycles with more focus on Lake Shore Blvd. as the arterial that, if not congested, allows for higher speed driving (Table 5). The table shows that the HD-UDDS cycle has higher average speeds, longer average micro-trip durations, shorter amount of time spent accelerating or decelerating, and longer time in the cruising mode; suggesting that the HD-UDDS cycle is less congested than the simulated cycles. Also the HD-UDDS cycle shows congestion with the high percentages of time spent idling and creeping without taking into account the effect it would have on the micro-trip durations; which is better represented in the simulated driving cycles by showing more stop-and-go traffic. This confirms that real-world road and vehiclespecific driving cycles are better representative of actual driving than the current dynamometer test recommended by the EPA. Applications of the simulated driving cycles As mentioned in the introduction, there are several applications for a disaggregate set of driving cycles. This section shows how driving cycles can be used for emission analysis and fleet route optimization (also known as the vehicle routing

1

City driving. City driving under low speed and stop-and-go traffic condition. 3 High acceleration aggressive driving. 4 Driving under less aggressive speeds and acceleration compared to the US06 cycle. 2

G. Amirjamshidi, M.J. Roorda / Transportation Research Part D 34 (2015) 255–266

Table 4 LDT urban cycles vs. available international urban cycles for light duty vehicles and light duty driving cycles used for emission testing.

263

264

G. Amirjamshidi, M.J. Roorda / Transportation Research Part D 34 (2015) 255–266

Table 5 MDT and HDT urban cycles vs. the Heavy Duty Urban Dynamometer Driving Schedule (HD-UDDS). Assessment measure

Simulated HDT Toronto driving cycles Lake Shore Blvd.

University Ave.

Arterials

Arterials with transit

Lake Shore Blvd.

Simulated MDT Toronto driving cycles University Ave.

Arterials

Arterials with transit

V (kph) Vr (kph) Acc (m/s2) Dec (m/s2) D (s) Pa (%) Pd (%) Pi (%) Pcrs (%) Pcrp (%) RMSA (m/s2)

28.45 33.28 0.280 0.605 33 59.3 27.5 11.7 1.4 23.7 0.571

12.40 16.74 0.403 0.598 21 46.05 31.06 22.27 0.62 39.18 0.583

16.57 20.53 0.398 0.585 22 49.30 33.53 15.60 1.57 31.79 0.603

16.87 21.59 0.405 0.557 26 43.38 31.59 18.63 6.40 34.37 0.534

25.72 29.40 0.265 0.647 38 62.8 25.7 10.4 1.2 23.5 0.521

13.05 17.02 0.376 0.629 19 50.4 30.1 19.1 0.4 39.4 0.637

15.23 19.35 0.360 0.647 23 52.2 29.1 17.7 1.0 35.6 0.636

15.76 20.14 0.355 0.590 20 48.0 28.9 17.5 5.7 38.6 0.539

EPA HD-UDDS cycle

30.17 45.17 0.480 0.574 71 27.1 22.6 31.9 18.4 36.6 0.464

Table 6 CO2-Eq emission factor (gr/km) for gasoline and diesel MDTs and diesel HDTs on different road categories. Road category

Freeway Lake Shore Blvd. University Ave. Major arterials Major arterials with transit

Vehicle category MDT gasoline

MDT diesel

HDT diesel

539.12 705.518 1139.2 995.837 963.619

761.072 977.057 1504.92 1328.69 1286.71

1189.7 1676.03 2404.95 2112.23 1972.07

Table 7 HD-UDDS CO2-Eq emission factor (gr/km) for diesel HDT/MDT and gasoline MDTs on urban unrestricted roads. Vehicle category

MDT gasoline

MDT diesel

HDT diesel

CO2-Eq Emission Factor (gr/km)

729.189

801.237

1540.62

problem (VRP)). As such, EPA’s MOVES2010b emission model was used to estimate running CO2-Eq emission factors for MDT and HDT driving cycles (Table 6), and the HD-UDDS cycle (Table 7). Since no data on fleet composition were available for downtown Toronto, CO2-Eq were estimated for gasoline and diesel medium duty trucks and heavy duty diesel trucks on different road categories available using a 2001 vehicle model.4 As can be seen from Table 6, diesel medium duty trucks produce higher CO2-Eq emission factors compared to gasoline vehicles by an average of 35%; and heavy duty trucks result in about 60% higher CO2-Eq emission levels compared to medium duty trucks. Also comparing the emission factors of simulated cycles with the HD-UDDS cycle shows that the HD-UDDS cycle produces emission factors higher than freeway driving, lower than Lake Shore Blvd., and much lower than University Ave. and arterial with and without transit. Current research is looking at the impact of using road-specific driving cycles in vehicle routing problems with different objectives of: time-optimal, distance-optimal, and emission-optimal, using a hypothetical fleet of HDT vehicles servicing random customers with random demands within the study area. So far, results of a Monte Carlo simulation have shown that emission-optimal VRP would reduce the total network emissions by 5.0% with an increase of only 1.9% in total distance compared to the distance-optimal; and an emission reduction of 1.1% with an increase of only 0.43% in total travel time compared to the time-optimal scenario.

Conclusion This paper developed and demonstrated a method for efficiently developing driving cycles that represent specific combinations of roadway class, time of day and vehicle attributes using simulated data. Use of simulation allows data to be collected under consistent traffic conditions for all vehicle and road types. Thirteen statistics representing speed, time, acceleration, speed squared, and acceleration squared were used to ensure the development of a good quality driving cycle. Analyzing the impact of road type on a driving cycles showed that freeway cycles are generally smoother and less aggressive, and have higher average speeds, lower accelerations and decelerations, longer micro-trips, and less time spent creeping 4

Average fleet age for medium and heavy duty trucks was 8 years in Ontario for the year 2009 (Statistics Canada, 2009).

G. Amirjamshidi, M.J. Roorda / Transportation Research Part D 34 (2015) 255–266

265

and idling. Also for the AM peak simulated cycles, arterials with and without transit showed similar characteristics, since both road categories are at capacity. Results also showed that on all roads, simulated LDT cycles have higher average speeds, acceleration and deceleration, and the shortest average micro-trip duration; suggesting that LDT cycles are more aggressive and spent less time cruising. Clearly, driving patterns differ for different cities and different vehicle types, indicating that the development of city-, time-, road type- and vehicle-specific cycles is justified. Differences between the assessment measures of the simulated cycles and available international driving cycles show the uniqueness of Toronto’s driving conditions. Currently, US cycles are used by car manufacturers in Canada for emission testing, however these results show the need for more specific driving cycles for specific emissions analysis scenarios. Finally, emissions estimated using MOVES and the simulated driving cycles showed that diesel vehicles produce an average of 35% higher CO2-Eq emission factors compared to gasoline vehicles; and heavy duty trucks result in about 60% higher emission levels compared to medium duty trucks. Acknowledgement The authors thank the Ministry of Transportation of Ontario, the Ontario Graduate Scholarship (OGS), and Natural Sciences and Engineering Research Council of Canada (NSERC) for their financial support. The Ministry of Transportation of Ontario also provided count and speed data. Intellican Transportation Systems Inc. provided the original Toronto Waterfront microscopic traffic simulation network. The Data Management Group and the City of Toronto provided traffic counts. References Abdulhai, B., Georgi, A., Roorda, M.J., Tarabain, A., and Tang, E., 2002. Microsimulation of the Toronto Waterfront Revitalization Plan: Impact Assessment of Roadway Configuration Alternatives. Final Report for the Toronto Waterfront Revitalization Corporation. André, M., 2004. The ARTEMIS European driving cycles for measuring car pollutant emissions. Sci. Total Environ. 334, 73–84. Amirjamshidi, G., 2015. Assessment of Commercial Vehicle Emissions and Vehicle Routing of Fleets using Simulated Driving Cycles. Doctor of Philosophy, Department of Civil Engineering, University of Toronto, Canada. Chugh, S., Kumar, P., Muralidharan, M., Kumar B, M., Sithananthan, M., Gupta, A., Basu, B., Kumar Malhotra, R., 2012. Development of Delhi Driving Cycle: A Tool for Realistic Assessment of Exhaust Emissions from Passenger Cars in Delhi. SAE Technical paper 2012–01-0877. Michigan, United States. Coelho, M.C., Frey, H.C., Rouphail, N.M., 2009. Assessing methods for comparing emissions from gasoline and diesel light-duty vehicles based on microscale measurements. Transport. Res. Part D Transport Environ. 14D, 91–99. Della Ragione, L., Meccariello, G., 2010. The evaluation of a new kinematic emissions model on real and simulated driving cycles. SAE Int. J. Fuels Lubricants 3, 521–531. Dieselnet, 2011. Common Artemis Driving Cycles [last Updated September, 2011]. [Accessed 2014, October, 10]. Dieselnet, 2012. Worldwide Harmonized Light Vehicles Test Procedure (WLTP) [last Updated December, 2012]. [Accessed 2014, October, 12]. Ericsson, E., 2001. Independent driving pattern factors and their influence on fuel-use and exhaust emission factors. Transport. Res. Part D Transport Environ. 6, 325–345. Fotouhi, A., Montazeri-Gh, M., 2013. Tehran driving cycle development using the k-means clustering method. Scientia Iranica. Green, J., Barlow, T.J., 2004. Traffic Management and Air Quality: Realistic Driving Cycles for Traffic Management Schemes. Department of Transport TRL Report TRL596, UK. Hallmark, S.L., Guensler, R., 1999. Comparison of speed-acceleration profiles from field data with NETSIM output for modal air quality analysis of signalized intersections. Transport. Res. Rec. 1664, 40–46. Han, D.S., Choi, N.W., Cho, S.L., Yang, J.S., Kim, K.S., Yoo, W.S., Jeon, C.H., 2012. Characterization of driving patterns and development of a driving cycle in a military area. Transport. Res. Part D 17, 519–524. Hung, W.T., Tong, H.Y., Lee, C.P., 2007. Development of a practical driving cycle construction methodology: a case study in Hong Kong. Transport. Res. Part D Transport Environ. 12D, 115–128. Int Panis, L., Broekx, S., Liu, R., 2006. Modelling instantaneous traffic emission and the influence of traffic speed limits. Sci. Total Environ. 371, 270–285. Kamble, S.H., Mathew, T.V., Sharma, G.K., 2009. Development of real-world driving cycle: case study of Pune, India. Transport. Res. Part D Transport Environ. 14, 132–140. Lai, J., Yu, L., Song, G., Guo, P., Chen, X., 2013. Development of city-specific driving cycles for transit buses based on VSP distributions: case of Beijing. J. Transport. Eng. 139 (7), 749–757. Manjunatha, P., Vortisch, P., Mathew, T.V., 2013. Methodology for the Calibration of VISSIM in Mixed Traffic. Transportation Research Board 92nd Annual Meeting. Washington, DC. Pelkmans, L., Debal, P., 2006. Comparison of on-road emissions with emissions measured on chassis dynamometer test cycles. Transport. Res. Part D Transport Environ. 11, 233–241. Prasad, S.A., Manickam, M., Balakrishnan, D., Saravanan, N., 2012. Development of Synthetic Drive Cycle for Fuel Economy Prediction. SAE Technical Paper 2012-01-2048. Raykin, L., Roorda, M.J., MacLean, H.L., 2012. Impacts of driving patterns on tank-to-wheel energy use of plug-in hybrid electric vehicles. Transport. Res. Part D Transport Environ. 17, 243–250. Roorda, M.J., Hain, M., Amirjamshidi, G., Cavalcante, R., Abdulhai, B., Woudsma, C., 2010. Exclusive truck facilities in Toronto, Ontario, Canada: analysis of truck and automobile demand. Transport. Res. Rec. 2168, 114–128. Roorda, M.J., Amirjamshidi, G., Misra, A., Mostafa, T.S., 2011. A Vehicle Emission Simulation Tool for Evaluating Green Commercial Vehicle Policy. Prepared for the Toronto Atmospheric fund, and the Ministry of Transportation of Ontario. University of Toronto. Saleh, W., Kumar, R., Kirby, H., Kumar, P., 2009. Real world driving cycle for motorcycles in Edinburgh. Transport. Res. Part D Transport Environ. 14, 326– 333. Shahidinejad, S., Bibeau, E., Filizadeh, S., 2010. Statistical development of a duty cycle for plug-in vehicles in a North-American urban setting using fleet information. IEEE Trans. Veh. Technol. 59, 3710–3719. Song, G., Yu, L., Tu, Z., 2011. Distribution characteristics of vehicle-specific power on urban restricted-access roadways. ASCE J. Transport. Eng. 138, 202–209. Sriniwas, A., Amal, J., Nandagopalan, C., Simhachalan, A., Pandey, N., 2011. A Real World Drive Cycle for India. SAE Technical paper 2011-26-0077. Statistics Canada, 2009. Canadian Vehicle Survey: Annual. 53–223-X. Tong, H., Hung, W., 2010. A framework for developing driving cycles with on-road driving data. Transport Rev. 30, 589–615.

266

G. Amirjamshidi, M.J. Roorda / Transportation Research Part D 34 (2015) 255–266

Tutuianu, M., Marotta, A., Steven, H., Ericsson, E., Haniu, T., Ichikawa, N., Ishii, H., 2013. Development of a World-Wide Worldwide Harmonized Light Duty Driving Test Cycle (WLTC). (GRPE-68-03). Sponsoring Entity: UNECE. Tzirakis, E., Pitsas, K., Zannikos, F., Stournas, S., 2006. Vehicle emissions and driving cycles: comparison of the Athens driving cycle (ADC) with ECE-15 and European driving cycle (EDC). Global NEST J. 8, 282–290. U.S. Environmental Protection Agency, Dynamometer Driver’s Aid [last Updated June 16, 2010]. http://www.epa.gov/nvfel/testing/ dynamometer.htm[Accessed 2010, July 20th]. Wang, Q., Huo, H., He, K., Yao, Z., Zhang, Q., 2008. Characterization of vehicle driving patterns and development of driving cycles in Chinese cities. Transport. Res. Part D Transport Environ. 13, 289–297. Weiss, M., Bonnel, P., Hummel, R., Provenza, A., Manfredi, U., 2011. On-road emissions of light-duty vehicles in Europe. Environ. Sci. Technol. 45 (19), 8575– 8581. Xiao, Z., Dui-Jia, Z., Jun-Min, S., 2012. A synthesis of methodologies and practices for developing driving cycles. Energy Procedia 16, 1868–1873. Younglove, T., Scora, G., Barth, M., 2005. Designing on-road vehicle test programs for the development of effective vehicle emission models. Transport. Res. Rec. 1941, 51–59. Yu, L., Zhang, X., Qiao, F., Qi, Y., 2010. Genetic algorithm-based approach to develop driving schedules to evaluate greenhouse gas emissions from light-duty vehicles. Transport. Res. Rec. 2191, 166–173. Zhang, J., Hounsell, N., Shrestha, B., 2012. Calibration of bus parameters in microsimulation traffic modelling. Transport. Planning Technol. 35, 107–120.