Validation of Truck Platoon Slipstream Effects

Abstract

Due to slipstream effects, platooning leads to a significant decrease of the fuel consumption of the heavy-duty vehicles (HDV). Measurements with a platoon consisting of three vehicles were performed at the Zalazone proving ground. The goal of these measurements was to get the static pressure at the front and the rear of the second vehicle to calibrate computational fluid dynamics simulation and to measure the fuel consumption directly. Measurements were done at a vehicle speed of 80 km/h and varying inter-vehicle distances. Platooning leads to a reduction of the pressure coefficients in the centre of the HDV front and an increase of the pressure coefficient at the top and the rear of the HDV. Furthermore, a reduction of the fuel consumption of the leading vehicle of 7.9% at an inter-vehicle distance of 6 m and 3.7% at a distance of 22 m was determined. A comparison to CFD simulation showed a similar fuel reduction for an inter-vehicle distance of 6 m and 22 m. CFD simulation showed an increase of fuel consumption at an inter-vehicle distance of 15 m. This increase was experimentally not validated. Also, results for the following vehicle are presented.

The original version of this chapter was revised by including a new coauthor. The correction to this chapter can be found at https://doi.org/10.1007/978-3-030-88682-0_18

1 Introduction

Platooning is defined as electronically coupling a group of vehicles at close inter-vehicle distances to either increase the capacity of the road or reduce fuel consumption due to slipstream effects. It is estimated that fuel consumption and exhaust gas emissions can be reduced up to 10% by platooning [1,2,3,4]. The share of fuel cost on the total costs in the transportation business is 21.2% [5]. The earnings of truckage companies are usually very low. Between 2015 and 2017, the earnings were between 1.8 and 3.0% [6]. Thus, even a small reduction of fuel consumption can lead to significant increase in earnings. There are three approaches to estimate the potential of reducing fuel consumption by platooning:

• 1. Real-world measurement—For real-world measurements, heavy-duty vehicles (HDV) are equipped with a limited measurement system, especially a GPS and a system for acquiring data of the on-board diagnostic system (OBD). Thus, long-term measurements are possible. However, it is very difficult to synchronise data of different HDVs of the platoon, it is impossible to get ambient parameters, and above all, there are strict legal restriction to the minimum inter-vehicle distance.

• 2. Computational fluid dynamics simulation (CFD)—Although the possibilities of computational fluid dynamics increased significantly in recent years, CFD has to rely on calibration data provided by measurements.

• 3. Measurements on a proving ground—For measurements on a proving ground, the platoon’s vehicle can be fully equipped, ambient conditions (wind speed and wind direction e.g.) are well known, measurements can be repeated several times, and it is possible to investigate very low inter-vehicle distances. In this work, the minimum distance was chosen to be 6 m. The mayor drawback of proving ground measurements is the limited length of the test section. In order to increase the aerodynamic effects, the vehicles have to go as fast as possible. At a test section of 400 m and a vehicle speed of 20 m/s, there are just 20 s to align the vehicles and conduct the measurements.

In the following, measurements on the Zalazone Proving Ground [7] with a platoon consisting of three HDVs are described. The main goal was to provide data of the static pressure at the vehicle surface to validate the CFD simulation and to quantify the influence of the inter-vehicle distance on the fuel consumption.

2 Materials and Methods

The proving ground, the vehicles and the sensor set-up are described in this section. Also, the mathematical models to calculate pressure coefficient and fuel consumption are presented.

2.1 Proving Ground

The measurement campaigns were done at the handling course of the Zalazone Proving Ground (see Fig. 6.1). This handling course has a total length of 2200 m and the straight section of 400 m, which was used as a test section for platooning measurements. It has a slight slope of approximately 1 m. This elevation has to be taken into account for the pressure measurements. A platoon with three vehicles of a length of 17 m and a distance of 11 m is illustrated for clarity in Fig. 6.1.

Fig. 6.1

Test section at the Zalazone Proving Ground, base map and map data from OpenStreetMap, © OpenStreetMap contributors under the CC-BY-SA license, https://www.openstreetmap.org/copyright

2.2 Heavy-Duty Vehicles

The platoon consisted of three heavy-duty vehicles (HDV) of type Volvo FH 540. The trailer of the HDV was empty. The weight of the HDV was approximately 17,000 kg. Due to this lightweight, the engines of the HDV were operated at an engine load of 25%.

2.3 Sensors

Data Acquisition System

The HDVs of the platoon were equipped with data acquisition systems which had to be synchronised below millisecond level (see Fig. 6.2). At all HDVs, a data acquiring system based on the on-board diagnostic system (OBD) and a GPS was installed. Among the data provided by the OBD, fuel rate, engine speed and engine load were used for the analysis of the fuel consumption. The second HDV of the platoon was additionally equipped with the following measurement systems:

• 1. Eight sensors were measuring the static pressure at the front of the HDV, and four Prandtl tubes were installed at hood air intakes.

• 2. In order to monitor the engine’s cooling systems, the radiator was equipped with a coolant volume flow sensor, a differential pressure sensor and a temperature difference sensors.

• 3. The static pressure at the vehicle rear was measured at seven locations.

• 4. The distance to the preceding HDV was measured by a laser distance sensor. A similar system was installed at the third HDV.

Fig. 6.2

Sensor system on the HDVs of the platoon

Pressure Sensors

Eight flat pressure probes (Type FDStat60 by SVMTec) were mounted at the front and seven at the rear of the vehicle. These probes were connected to the pressure sensors (Honeywell, 006MDSA3, range ± 600 Pa) by small tubes. However, these sensors are differential pressure sensors. Thus, a reference pressure is needed. Clearly, this reference pressure can neither be taken from ambient (as the HDVs are moving) or from the drivers cabin as an underpressure is induced by the passing air flow. Therefore, the pressure sensors were connected to a thermally isolated reference pressure vessel. In Fig. 6.3, the positions of the pressure probes are indicated.

Fig. 6.3

Flat pressure sensors mounted on the front (left) and rear of the second HDV (right)

Ambient

Speed and direction of the ambient air were measured using a 3D ultrasonic anemometer. As can be seen in Fig. 6.1, the weather station was located at the end of the test section. As wind speed and direction can change along the test section, this anemometer only gives information regarding the overall wind speed and direction during the measurements.

2.4 Measurement Campaigns

Two measurement campaigns were conducted at Zalazone Proving Ground in July and August 2019. The focus of the first campaign was on the pressure measurements and on the fuel consumption during the second campaign. All measurements were done at a target vehicle speed of 80 km/h or 22.2 m/s. For each campaign, the first measurements were done with a single reference vehicle. Afterwards, the measurements were done with the platoon at different distances. At the first campaign, the inter-vehicle distances were set to 6, 11, 15, 22 and 55 m. For the second campaign, the distances were 7, 15 and 22 m. The measurements were repeated several times, usually five times during the first campaign and nine times during the second campaign.

2.5 Static Pressure

The static pressure $p_{i,\textrm{stat}}$ at position $(x_i,y_i)$ of the ith sensor position on the HDV is the sum of the dynamic pressure and the barometric pressure, given by

$$\begin{aligned} \begin{aligned} p_{i,\textrm{stat}}&= \rho \,c_{\textrm{p},i}\left( x_{i}, y_{i}, d_{\textrm{front}}, d_{\textrm{back}}, \Delta y_{\textrm{front}}, \Delta y_{\textrm{back}} \right) \frac{v^2 + w^2 + 2 v w \cos (\beta )}{2}\\&\quad + a_{h} \Delta h. \end{aligned} \end{aligned}$$

(6.1)

The dynamic pressure (first summand) depends on the ground speed v, the wind speed w, the direction $\beta$ of the wind speed with respect to the driving direction as well as the coefficient of pressure $c_{\textrm{p},i}$, see Fig. 6.4. The conversion of the dynamic pressure to the static pressure and thus also $c_{\textrm{p},i}$ is influenced by:

• 1. Geometry—The conversion depends strongly on the position of the sensor. In the centre of the HDV’s front (e.g. sensor 4 and sensor 6 in Fig. 6.3), nearly all the dynamic pressures are converted to static pressure, while at the roof, even a negative static pressure can be induced (e.g. sensor 8 in Fig. 6.3).

• 2. Distances—The air flow can be blocked by the preceding or following vehicle. The distances to the preceding vehicle $d_{\textrm{front}}$ and following vehicle $d_{\textrm{back}}$ influence the static pressure and therefore the fuel consumption.

• 3. Lateral offset—If one vehicle has a lateral offset to another, the air flow is only partially blocked. Thus, $\Delta x_{\textrm{front}}$ and $\Delta x_{\textrm{back}}$ also influence static pressure, drag and fuel consumption.

The barometric pressure $a_{h} \Delta h$ [second summand in (6.1)] decreases by 12 Pa per metre and can be seen well in the measurement of the sensors at the rear of the HDV.

Fig. 6.4

Factors influencing static pressure and fuel consumption

2.6 Data Preprocessing

While GPS data and OBD data were sampled at 50 Hz, other parameters were measured at a slower rate (e.g. pressure at 2 Hz). Thus, in a first step, all measurement data were interpolated to the same sampling rate. Acceleration was not measured directly, but calculated from the GPS speed. As the involved derivative calculation significantly increased the noise level, acceleration was smoothed by a Savitzky–Golay filter. Data was sampled continuously during the whole lap. The data sampled at the test section was extracted for further analysis. In most of the following charts, the data of the test section is indicated by red shading.

3 Results

The goal of this measurement campaign was to provide data to validate the computational fluid dynamics (CFD) simulation of the platoon and to directly measure fuel consumption and the influence of inter-vehicle distance.

3.1 Static Pressure

In a first step, the measurements were done with a reference vehicle to get the static pressure without influence of the platoon. Afterwards, measurement with a platoon consisting of three HDVs was done with different inter-vehicle distances.

Reference Vehicle

The static pressures measured at eight positions in the front of the reference HDV (see Fig. 6.3) are shown in Fig. 6.5. The data of the whole measurement is shown. The vehicle drove three laps. The data of the test section is indicated by red shading and can be easily identified due to the approximately constant pressure profile. The mean values of the sensors at the test section are also given in Fig. 6.5. As the vehicle speed was 21.6 m/s and the wind speed was 0.8 m/s coming from south east, the dynamic pressure is calculated to 273 Pa. The sensors p$_{\_\mathrm{{stat}}\_4}$ and p$_{\_\mathrm{{stat}}\_6}$ are in the centre of the enginehood, and $c_{\textrm{p}}$ is close to one. Sensor p$_{\_\mathrm{{stat}}\_1}$ located in front of the licence plate also got a $c_{\textrm{p}}$ close to unity. Contrary, at sensor p$_{\_\mathrm{{stat}}\_8}$, even negative pressure values were measured. The pressure at the laterally positioned sensors p$_{\_\mathrm{{stat}}\_2}$ and p$_{\_\mathrm{{stat}}\_3}$ and p$_{\_\mathrm{{stat}}\_5}$ and p$_{\_\mathrm{{stat}}\_7}$ varies between 115 and 240 Pa.

Fig. 6.5

Static pressures at the front of the reference HDV

At the test section, the static pressure values measured at the rear of the HDV varied between −11 and −28 Pa (Fig. 6.6). At all sensors, the static pressure declined by approximately 10 Pa. This is due to the elevation of the test section by one metre.

Fig. 6.6

Static pressures at the rear of the reference HDV

Fig. 6.7

Static pressures at position 4, vehicle speed, height and residuum of the model

Model of the Static Pressure at the Centre of the Front of the Reference Vehicle

For the reference vehicle, Eq. 6.1 simplifies significantly. As $d_{\textrm{front}}$ and $d_{\textrm{back}}$ as well as $\Delta x_{\textrm{front}}$ and $\Delta x_{\textrm{back}}$ vanish, $c_{\textrm{p}}$ is constant. Thus, $c_{\textrm{p}, i}$ and a$_{\textrm{h}}$ can be determined by linear regression, since the static pressure as well as the vehicle speed, the wind speed, the wind direction and height are measured. The results for sensor p$_{\_\mathrm{{stat}}\_4}$ are shown in Fig. 6.7 as an example. Data of all other sensors behaves the same way. The adjusted coefficient of determination is 0.99, indicating that the independent data can very well described by the model of Eq. 6.1. The independent variables speed and height are shown as well as the dependent variable p$_{\_\mathrm{{stat}}\_4}$. Again, the range of the test section is highlighted by a red background. The residuum of the linear regression is also shown in Fig. 6.7. The model explains the measurement data very well for the test section, where the model overestimates the static pressure of sensor 4 only by approximately 4%. In Fig. 6.8, the residuum along the handling course is shown. Length and colour of the arrows indicate the normalised residuum, while the direction of the arrows corresponds to the velocity vector. It can be seen that the residuum is very low at the test section and largest when the vehicle travelled south-west. In Table 6.1, the pressure coefficients are given for the the eight sensors at the front and for the seven sensors at the back of the reference vehicle. In the columns “TS LAP 1”, “TS LAP 2” and “TS LAP 3”, the pressure coefficient is calculated with the data of the test section for each lap, only. In the column “Total”, data of the whole handling course was used to calculate the pressure coefficient. There is a variation of 5% for the test section measurement. The pressure coefficient for the whole handling course is higher. As was shown in Fig. 6.8, $c_{\textrm{p}}$ is overestimated when the vehicle goes south-west.

Fig. 6.8

Normalised residuum (0–100%) and velocity vector along the handling course

Table 6.1 Pressure coefficient of the reference HDV

Table 6.2 Pressure coefficient at the front and rear of the second HDV of the platoon for a varied inter-vehicle distance

Pressure Coefficients for the Platoon

The static pressure for the sensors mounted on the second HDV of the platoon is shown in Fig. 6.9. The vehicle speed was set to 22.2 m/s. However, there was a tailwind of 2 m/s coming from south/southeast. The platoon went for four laps. Again, the data of the test section is highlighted in red. Comparing Figs. 6.5 and 6.9, the flow profile differs significantly. The pressure values of sensor 1, mounted in the middle of the license plate, are reduced from 250 Pa for the reference vehicle to even negative values of lap 4 for the following vehicle. On the other hand, values at sensor 8 near the roof increase from $-20$ to 70 Pa. At the first lap, a strong decrease of the measured pressure can be seen. For this measurement, the drivers did not succeed in aligning the HDV but reduced the distance constantly. During the first lap, the mean distance between the first and the second HDV was 9.6 m, between the second and the third vehicle 15.4 m. The pressure coefficient calculated by linear regression for the second vehicle of the platoon (front as well as rear) is given in Table 6.2. The coefficient of determination calculated by linear regression is 0.934, so somewhat smaller than for the reference vehicle, but nevertheless indicating that 93.4% of the variance can be explained by this model. $c_{\textrm{p}}$ increases with the inter-vehicle distance. However, even at a distance of 55 m, $c_{\textrm{p}}$ is significantly lower than at the reference vehicle.

Fig. 6.9

Static pressures of the sensors in front of the second HDV. Distance to the preceding HDV 6 m

There is a significant increase of the pressure coefficient at inter-vehicle distance of 15 m. All sensors beside p$_{\_\mathrm{{stat}}\_3}$, p$_{\_\mathrm{{stat}}\_7}$ and p$_{\_\mathrm{{stat}}\_8}$ show significant higher pressure values. Interestingly, also CFD simulation shows a significant increase of the pressure coefficient at a inter-vehicle distance of 15 m.

3.2 Fuel Consumption

Reference Vehicle

Acceleration and deceleration are the most fuel-consuming driving patterns. The reference vehicle used cruise control which minimises the number of acceleration and deceleration events. Additionally, it is not necessary to adjust distance and lateral offset to the other vehicles. Therefore, fuel consumption of the reference could be measured accurately. At an engine speed of 1340 RPM, the average fuel consumption at the test section was measured to 5.81 ± 0.26 ml/s. However, measurement were done at different engine speeds. Using data from the first vehicle of the platoon, a correction factor was calculated:

$$\begin{aligned} f_{\textrm{rate}} = 0.0027 N_{\textrm{eng}} + 1.68 \end{aligned}$$

(6.2)

Thus, for the reference vehicle, at an engine speed of 1750 RPM, a fuel rate of $f_{\textrm{rate}}=6.8$ ml/s is given.

First Platooning Vehicle

As for the reference vehicle, the first vehicle does not have to accelerate and decelerate, and the fuel consumption is not strongly superimposed by acceleration effects. It is assumed that the fuel consumption is influenced by acceleration, engine speed, distance and offset of the following HDV. Vehicle speed has a big influence to the fuel consumption. However, at this measurement set-up, vehicle speed was constant at 22.2 m/s. As least square regression can only minimise variance (and there was none), vehicle speed was not included in the model. Additionally, there is a strong correlation between vehicle speed and engine speed. Including both variables to the model resulted in colinearity problems. The model used for estimating the fuel rate $f_{\textrm{rate}}$ is:

$$\begin{aligned} f_{\textrm{rate}} = a_{0}a + a_{1} N_{\textrm{eng}}+ a_{2}\Delta X_{1-2} + a_{3}\Delta Y_{1-2} \, \end{aligned}$$

(6.3)

with the acceleration a in m/s$^{2}$, the rotational speed of the engine $N_{\textrm{eng}}$ in revolutions per minute/1000, the distance between the first and the second HDV $\Delta X_{1-2}$ in m and the lateral offset of the second vehicle $\Delta Y_{1-2}$ in metres. Other models were tested but did not produce statistically significant results. The parameters a$_{\textrm{i}}$ were estimated by linear regression and are shown in Table 6.3. The most influential parameter is acceleration. The fuel rate increases by 12.05 (ml/s)/(m/s$^2$). The fuel rate due to engine speed increases by 3.57 (ml/s)/(RPM/1000). The influence of the distance is described by a$_{2}$. It is small, however statistically significant. The fuel rate increases by 0.018 (ml/s)/m. The influence of the lateral shift of the preceding vehicle is even smaller. The fuel rate increases by 0.0041 (ml/s)/m.

Table 6.3 Regression parameter for the fuel rate for the first vehicle of the platoon

Table 6.4 Regression parameter for the fuel rate for the second vehicle of the platoon

Following Vehicles Within the Platoon

The same model (6.3) was applied to the second vehicle of the platoon. The regression parameter is shown in Table 6.4. Interestingly, the influence of the acceleration is twice as high as for the first vehicle and the influence of the engine speed is the same as for the first vehicle. The parameter a$_{2}$ describing the influence of the distance is smaller. The lateral offset to the first vehicle is statistically not significant. The fuel saving of the platoon compared to a single HDV is shown in Table 6.5. For the leading HDV, the fuel saving is between 7.9% at a distance of 6 m and 3.7% at 22 m. The upper and lower limits of the confidence interval are also given in Table 6.5. The fuel saving of the second HDV is 8.2% at a distance of 6 m, 7.14% at 15 m and 5.2% at 5.2 m. However, the error of margin is much higher than for the first HDV.

Table 6.5 Fuel savings due to platooning for a varied inter-vehicle distance

3.3 Comparison to Simulation Results

A comparison of the fuel consumption of the CFD results with measurement data was done. As shown before, the direct fuel consumption values can only be used for the first truck. In comparison with the following trucks, only the first truck has an evenly sufficient drive (use of the cruise control), which means that the measured reduction in fuel consumption can be clearly attributed to the reduction in air resistance. A comparison of the consumption reduction of this measurement with the CFD simulations is shown in Fig. 6.10. For the smallest distance $6\,{\textrm{m}}$, a very good agreement between measurement and simulation is obtained. A deviation of 4–5% can be observed for the distance $15\,{\textrm{m}}$. Also, the effect that the fuel reduction of distance $15\,{\textrm{m}}$ is nearby the same as for distance $22{\textrm{m}}$ could not be seen for the measured fuel data but can be seen in the derived pressure drag coefficients for front and back from measured data, see Fig. 6.11. As explained in this chapter the measured fuel rate data was quite sensitive to external factors. For the $22\,{\textrm{m}}$ point, the absolute error is smaller than for the $15\,{\textrm{m}}$ point and is around 2%. It can also be observed that the measured values are always greater than the simulated values.

Fig. 6.10

Comparison of simulation and measurement results of the fuel reduction in [%] relative to single truck

Another comparison of the simulation results was done using the pressure signals from the front and back of truck 2. The exact measurements of these pressure signals and position of the sensors is shown in this chapter. The pressure signals where also evaluated for the simulation results. The same position of the pressure sensors like in the measurements was used. Some pressure measurement points are located in areas with local pressure gradients. In order to determine the influence of the local sensor position on the pressure results, the static pressure values of the simulations on the surface of the front and rear sides were averaged for different radii. A displacement of the sensor up to $3\,{\textrm{cm}}$ has no noticeable influence on the result. Since the positioning accuracy of the pressure sensors is within this range, the influence of the positioning can be neglected for all further evaluations. The pressure values were used to calculated an arithmetic average pressure value for the front and back sides. The average pressure value is used to calculate the pressure drag coefficient. For calculation of the pressure drag coefficient, the stagnation pressure and reference pressure were used. The drag coefficient can be divided into two components, namely frictional drag coefficient (viscous drag) and pressure drag coefficient (form drag), whereas for our cases, the frictional drag coefficients are insignificantly small in relation to pressure drag coefficients. In Fig. 6.11, the different pressure drag coefficients for front and back are shown. The solo case with a theoretical distance of infinity is represented here with a distance of $100\,{\textrm{m}}$ because of illustration. The pressure drag coefficient values for the front have a good agreement for all compared cases except at the point $15\,{\textrm{m}}$. Here, the measured value is higher. For the pressure drag coefficient values in the back deviations for all distances can be observed. The measured pressure drag coefficients at the back are greater than the pressure drag coefficients obtained from simulations, which indicates that the simulations might overpredict the negative pressure and underestimate the influence of the trailing truck on the pressure distribution at the back. It seems the deviation of the measured and simulated pressure drag coefficients at the back has a nearby constant offset for all reliable data points and the decreasing trend of the pressure drag coefficient at the rear side of the truck is covered by simulation.

Fig. 6.11

Comparison of simulation and measurement results for the pressure coefficients front and back for truck 2

4 Discussion

The measurement of the fuel consumption of a platoon compared to a single HDV poses a great challenge to the instrumentation of the HDV, the measurement campaign and the data processing.

4.1 Instrumentation

While the instrumentation of an HDV poses a challenge for itself, the sticking point of the instrumentation is the highly synchronised data acquisition. Due to the high speed of the HDV of 22 m/s, even a time shift of ten milliseconds results in an error of the distance measurement of 22 cm. An overview of all sensors can be seen in Fig. 6.12.

Fig. 6.12

Platoon with a sketch of all sensors

The measurement system “DATA.BEAM” is developed at the Virtual Vehicle Research GmbH for distributed measurements with precision time synchronisation. During measurements, all “DATA.BEAM” devices on the trucks are synchronised via 868 MHz wireless communication to an average error of 10 $\upmu$s. Immediately after the measurement, “DATA.BEAM” collects the measurement data. Using this procedure, it is possible to substantially speed up the measurements and reduce the effort of post-processing as all data are already synchronised.

4.2 Measurement Campaign

There are two possibilities measuring fuel consumption of a platoon vehicle: either a measurement campaign at a test section, as it was done in this project, or real world on road measurements. Both approaches have advantages and disadvantages. The measurement campaign at the Zalazone Proving Ground has the main advantage that important parameters like wind speed and wind direction can be measured and the measurement can be repeated under the same boundary conditions. The major drawback was the very short test section of only 400 m. Thus, the HDV passed the test section in less than 20 s. During this short period, the HDV had to be aligned and the distance had to be adjusted. It was very difficult to reach a steady state of the platoon without any acceleration events. This can be seen in Fig. 6.15. The positions of the HDV of the platoon are shown every two seconds. The platoon has been shifted along the y-axis for better visibility. The platoon entered the test section and was aligned after 6 s. After 8 s HDV 3 was too close. The distance to HDV 2 was 3.5 m instead of 7 m. This was corrected by the driver. However, this manoeuvre required some deceleration. The platoon was stable after 12 seconds. After 14 seconds HDV 2 was too close to HDV 1. After 18 seconds, the first HDV reached the end of the test section and the platoon was dissolved.

Fig. 6.13

Alignment of the HDV at the test section

Fig. 6.14

Acceleration events at the test section

To reduce the effect of acceleration, data with an acceleration above a threshold of 0.2 m/s$^2$ was discarded. In Fig. 6.13, an histogram of the acceleration at the test section is shown. According to Tables 6.3 and 6.4, even an acceleration of 0.05 m/s$^2$ leads to an additional fuel rate of 0.5 ml/s for the trailing and 1 ml/s for the following HDV. The fuel rate versus engine speed is shown for a distance of the HDV of 7 m for all laps. The black lines indicate the maximum fuel rate at 100% engine load. The fuel rate for different laps is indicated by a different colours and marker symbols. It can be seen that due to the empty trailer, the measurements took place at low fuel rates. As described, all data above an absolute value of 0.2 m/s$^2$ have been discarded. Nevertheless, the fuel rate varies between 1 and 15 ml/s. The analysis of the least square model suggests that approximately 50% of this variation can be explained by the acceleration. Figure 6.14 an histogram of the acceleration at the test section is shown.

Fig. 6.15

Fuel rate over engine speed for HDV 2 at a distance of 7 m

4.3 Lessons Learned

The measurement campaigns revealed that highly accurate and synchronised data of a moving platoon consisting of three HDV can be provided. Especially, due to the measurement procedure provided by “DATA.BEAM”, it was possible to do the measurement according to a very strict time table. As the measurement data was automatically synchronised and put into data groups, post-processing could be reduced to a minimum. However, it became evident that a test section of 400 m is too short. At a inter-vehicle distance of 22 m, the length of the platoon is already 95 m; thus, a quarter of the test section is occupied by the platoon itself. It is therefore highly recommended to use a test section with a minimum length of 800 m.