International Journal of Heat and Fluid Flow 90 (2021) 108823

Contents lists available at ScienceDirect

International Journal of Heat and Fluid Flow

journal homepage: www.elsevier.com/locate/ijhff

Aerodynamic drag reduction by means of platooning configurations of light

commercial vehicles: A flow field analysis
J.J. Cerutti a, G. Cafiero a, b, *, G. Iuso a
a
Department of Mechanical and Aerospace Engineering, Flow Control Group, Politecnico di Torino, Corso Duca degli Abruzzi 24, 10129 Torino, Italy
b
Centre for Aerodynamics and Environmental Flow, Department of Mechanical Engineering Sciences, University of Surrey, GU2 7HX Guildford, UK

A R T I C L E I N F O A B S T R A C T

Keywords: Platooning configurations of two, three and four commercial vehicles were tested at a Reynolds number based on
Platooning the vehicle’s length (L) of230000. The platoon configurations were obtained using an instrumented model, and
Drag reduction three wooden replicas located at different positions with respect to the instrumented one. The reference model
3D wake dynamics
presents a slant angle at the leading edge, which can produce, in principle, a significantly different flow field
compared to the generally studied Ahmed body. Drag, static pressure distributions and pressure fluctuations
measurements were carried out. Additionally, planar PIV measurements were performed to investigate the near
wake of the two-vehicles platoon configuration.
For the two-models platoon, drag reductions of 30% and 43% were evidenced for the front and for the rear
vehicle, respectively, at an inter-vehicle distance (d) equal to half the vehicle’s length, and corresponding to an
average drag reduction of 36.5%. For increasing distance, the benefit associated with the platooning configu­
ration reduces, reaching an average drag reduction of 20% atd/L = 3. We relate the vehicle’s drag to the flow
field organization and to the distribution of the modal energy through Proper Orthogonal Decomposition of the
microphonic probes located on the base of the instrumented vehicle. We also evidence that the key element that
is responsible for the pumping of the wake is the large vortex that generates near the top edge of the vehicle’s
base.
We show that the slant angle does not affect the drag reduction of the leading vehicle of the platoon, whereas it
can lead to larger differences in the case of the rear vehicle.
For three and four-vehicles platoons, consistently larger values of the average drag reduction are experienced
(≈35%) and were also obtained for distances >1L. A simple model describing the overall drag reduction for a
generic number of vehicles is presented and discussed.

1. Introduction vehicles accounting for the 30% of the total transport energy use
(Bonilla, 2020). It is then mandatory to identify innovative approaches
The ambitious goals set by the European Union to achieve significant to tackle the pollution associated with a such large volume of vehicles.
reductions of the pollution and sound emission throughout the entire In this field, the increasing capabilities of artificial intelligence has
transportation market have stimulated flourishing research in the field. provided a strong push towards the development of autonomous
The interest is strongly cross-disciplinary and requires efforts at different driving. The application of these new developments paves the way to­
levels. Examples can be found in the air transportation, where engine wards the implementation of the “platoon” marching of heavy vehicles.
and airframe manufacturers are pushing towards distributed propulsion, Platoon marching is generally referred to as the case of a convoy of a
with significant benefit in terms of emissions and sound pollution. given number of vehicles that are marching in a line.
However, a large share of the commercial routes in Europe still While this feature is extensively employed in motorsport, where it is
heavily relies on the road transportation. It is estimated that the 45.9% generally referred to as slipstreaming, as well as by cyclists, the hazard
of the goods are moved across Europe using light and heavy commercial associated with the close proximity of the vehicles hinders its

* Corresponding author at: Department of Mechanical and Aerospace Engineering, Flow Control Group, Politecnico di Torino, Corso Duca degli Abruzzi 24, 10129
Torino, Italy.
E-mail address: [email protected] (G. Cafiero).

https://doi.org/10.1016/j.ijheatfluidflow.2021.108823
Received 9 November 2020; Received in revised form 13 April 2021; Accepted 3 May 2021
Available online 11 June 2021
0142-727X/© 2021 Elsevier Inc. All rights reserved.
J.J. Cerutti et al. International Journal of Heat and Fluid Flow 90 (2021) 108823

Fig. 1. a) Schematic representation of the model with indication of the location of the pressure taps (red) and microphones (blue). b) Sketch of the model’s
positioning in the test section, with schematic representation of the location of the drag balance and the boundary layer suction slit. Drawing not to scale. (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

implementation for commercial vehicles. coefficient of the single models, as well as the overall drag coefficient of
A pioneering study of the aerodynamics of the platooning configu­ the platoon? What is the effect of the platoon configuration on the wake
ration was performed by Zabat et al. (1995). Drag reductions of both the dynamics and its structure, as a function of the inter-vehicle distance?
leading and trailing model were evidenced for inter-vehicle distances (d) Both from the flow field data obtained with particle image velocimetry
smaller than the vehicle’s length (L). The authors also evidenced that the and from the proper orthogonal decomposition of the microphonic
drag reduction of the leading model disappears when d/L > 1. probes we are interested in determining the leading mechanism driving
Even though numerous investigations have shown its potential the drag variation as a function of the inter-vehicle distance. Further­
benefits (Michaelian and Browand, 2000; Tsuei and Savas, 2000; Davila more, comparing our results with the literature, we can draw some
et al., 2013; Vegendla et al., 2015; Salari and Ortega, 2018; Hamiga and conclusions on the effect of the slant angle on the drag reduction of the
Ciesielka, 2018, to cite some), the main concern arises from the fact that platoon.
the largest drag reductions are evidenced at short inter-vehicle spacings, Besides providing relevant information on the overall drag reduc­
with potential risks associated with the response time in case of emer­ tion, the flow field organization of the near wake of vehicles marching in
gency. As mentioned, developments in autonomous driving and control platoon configuration at different inter-vehicle distances is of para­
(Gehring and Fritz, 1997; Alam et al., 2013; Al Alam et al., 2011; Ber­ mount importance for the appropriate characterization of the dispersion
genhem et al., 2012; Yu et al., 2016 to cite some) will enable the large- of pollutants as well as for the noise emission of light and heavy duty
scale implementation of this methodology, with the associated renewed vehicles.
interest in the understanding of the underlying physics leading to drag The paper is organized as follows. In section 2 we describe the model,
reduction, particularly from the aerodynamics point of view. the wind tunnel and the experimental technique. In section 3 we present
Under some specific platoon configurations, the resulting aero­ the obtained results in terms of drag and pressure distribution.
dynamic performance may actually worsen, as demonstrated by the Furthermore, we focus on the near wake flow field topology and on the
investigations by Pagliarella et al. (2007) and Watkins and Vino (2008). modal analysis of the wake. In section 4 we propose a revised model to
The authors investigated a range of geometry modifications on the obtain the optimal number of vehicles in the convoy to maximize the
typical Ahmed body. In particular, they looked at the effect of the slant drag reduction. Finally, we draw our conclusions.
angle and they investigated a range of configurations where the leading
and rear model had pre-critical (smaller than 30◦ ) and post-critical 2. Experimental setup and measurement techniques
(>30◦ ) values of the slant angle. The authors also argued that the drag
reductions experienced on the rear vehicle are progressively reduced as The model is a 3D printed 1 : 10 scaled light commercial vehicle,
a consequence of the pressure increase associated with the large vortices with length (L), width (W) and height (H) equal to 412 mm, 170 mm and
shed by the leading vehicle, impinging on the front surface. 200 mm, respectively. The selected geometry is representative of an
More recently, a joint experimental and large eddy simulations (LES) existing light commercial vehicle currently manufactured by Fiat FCA. A
study by Uystepruyst and Krajnović (2013) evidenced potential benefit schematic of the model is shown in Fig. 1. The tests were performed in an
for the vehicles in a platoon of four cuboids at small values of the inter- open-circuit wind tunnel at the laboratory of Aerodynamics “Modesto
vehicle distance. Other numerical efforts, requiring less computational Panetti” at Politecnico di Torino. The test section of the wind tunnel has
power and based on the use of detached eddy simulations (DES), a height of 900 mm, width of 1200 mm and length of 6500 mm. When
allowed the investigation of a much larger platoon (8 vehicles) with an the model is mounted in the wind tunnel, the resulting blockage is less
inter-vehicle spacing of 0.5L. They showed good agreement with the than 3.5%.
available experimental results obtained with moving models rather than The freestream speed (V∞ ) is set to 9m/s corresponding to a Reynolds
in wind tunnel tests (He et al., 2019). The authors evidenced large values number based on the model’s length ReL = ρVμ∞ L = 230000. The Rey­
of the drag reduction for all the intermediate vehicles, with an associ­ nolds number independence was ensured by investigating a broader
ated highly turbulent flow. range of freestream speeds and checking the variation of the drag co­
Despite the large number of investigations that have dealt with the efficient. The results confirmed only minimal variations of the drag
problem, many of them have typically focused on the effect of the pla­ coefficient for a 30% reduction of a freestream speed.
tooning configuration on the drag coefficient, with lower interest on the The model is positioned within the test section at a distance from the
statistical description of the resulting flow field for a broad range of inlet equal to 1900 mm, as sketched in Fig. 1. The experimental setup is
inter-vehicle distances. the same as the one reported in Cerutti et al. (2019) Cerutti et al. (2020),
In this paper, we aim at addressing the following aspects: for a 1:10 Cafiero et al. (2019).
scale model of a commercial light vehicle, with a leading edge slant The model is instrumented with 64 pressure taps that populate its
angle, what is the effect of the platooning configuration on the drag

2
J.J. Cerutti et al. International Journal of Heat and Fluid Flow 90 (2021) 108823

Fig. 2. a) Percentage variation of the drag coefficient ΔCD for the two models platoon configuration as a function of the distance between the models d/L: front
model (Blue), rear model (Red), overall drag variation (dashed line). b) Pressure coefficient distribution on the base of the front (top row) and the rear model (bottom
row). The isolated case is also reported as a reference. The location of the pressure taps is indicated with black dots. The contour plots are obtained by interpolating
the measurements on a structured grid. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

lateral and rear surfaces, as schematically depicted in Fig. 1 by the red suction of the boundary layer on the drag coefficient of the isolated
dots. Furthermore, pressure taps are also located in the symmetry plane model and found that the drag coefficient without and with suction are
of the model. The pressure taps are connected to a Scanivalve ZOC 33 different only by less than 2%.
pressure transducer interfaced with a PC via TCP/IP protocol. Sixteen The model is connected through a hollowed beam to a load cell
microphones in pin-hole configuration, reported in blue in Fig. 1, are (Dacell UU-K002 load cell, full scale of 2 Kgf and accuracy equal to
mounted on half of the base of the model. The signal of these micro­ 0.002%FS). The mounting configuration is shown in Fig. 1b. The beam is
phones were calibrated using a high quality Bruel and Kjaer probe, as enclosed within an aerodynamically shaped hull, as a NACA0018
described by Sardu et al. (2016). profile.
The incoming boundary layer profile was characterized in the case of The voltage signals from the load cell and the pressure fluctuations
an isolated model. In order to minimize the boundary layer thickness on sensors were sampled by a National Instruments data acquisition board
the wind tunnel floor, a suction slit was located 1L upstream of the characterized by A/D resolution of 16bit. The drag measurements were
model. The suction speed was fixed at VS /V∞ = 1. The incoming performed taking into account a total sampling time of three minutes to
boundary layer developing on the wind tunnel floor was measured using ensure statistical convergence.
a hot wire probe positioned at a distance x/L = 0.1 ahead of the model. The platoon configurations are reproduced using three equal non-
The boundary layer displacement thickness (δ* ) compared to the ground instrumented dummy models. Given the length of the test section as
clearance is δ* /G = 0.07, which is in good agreement with other in­ well as the length of the van model, non-dimensional distances ranging
vestigations on commercial vehicle’s aerodynamics (Castelain et al., from d/L = 0.125 to d/L = 3 were investigated.
2018; Hucho and Sovran, 1993). We also checked the influence of the Planar PIV (2D2C) measurements were performed for the two-model
platoon and focused on the investigation of the symmetry plane of the

3
J.J. Cerutti et al. International Journal of Heat and Fluid Flow 90 (2021) 108823

Fig. 3. Pressure coefficient distributions for the isolated, front (left) and rear (right) vehicles. Distributions on the lateral surface as a function of the normalised
distance d/L. The vector plots represent the value of the pressure coefficient on the roof, front part, rear part and underbody of the vehicle along the symmetry plane.
The contour plots are obtained by interpolating the measurements onto a structured grid. Blue and red vectors are related to positive and negative values of the
pressure coefficient, respectively. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

model (XZ). The illumination of the tracer particles is obtained using a wind tunnel walls and on the model. Images deformation and velocity
Dantec Dynamics Dual Power Nd:YAG laser characterized by 200 mJ/ vector fields interpolation are carried out using spline functions
pulse and 15 Hz maximum repetition rate. The laser beam was shaped (Astarita, 2006, 2008). A Blackmann weighting window is used during
into a sheet having a thickness of about 1 mm, using a spherical and a the correlation process to tune the spatial resolution of the PIV process
cylindrical lens. The PIV images were captured using one Andor sCMOS (Astarita, 2007). A multi-step algorithm is implemented to perform the
5.5Mpix camera equipped with a Tokina 100 mm Macro lens operated at cross-correlation. The initial interrogation window size of 48x48 pixel is
an aperturef# = 16. The resulting digital resolution is about 10pix/mm. reduced down to 24 × 24 pixel, corresponding to 2.4 × 2.4 mm (i.e. 68
In order to ensure a typical displacement between the two frames of independent vectors across the model’s width), with 75% overlap.
about 12 pixels, the time delay between the two laser pulses was fixed to Considering the typical error in the detection of the correlation peak in
140μs. For each test, 3000 images were captured to ensure the conver­ the PIV process, we can estimate the velocity vector to be accurate to
gence of the relevant statistics. within 1% of their value.
The background image as well as the historical minimum is removed
from the PIV images to attenuate the effect of laser reflections on the

4
J.J. Cerutti et al. International Journal of Heat and Fluid Flow 90 (2021) 108823

Fig. 4. Drag coefficient of the front (top) and rear (bottom) vehicle when marching in platoon configuration (CD ) normalized with respect to the drag coefficient of
the isolated case (CD0 ) and comparison with available literature using similar models.

3. Results leading vehicle.

Conversely, the progressive reduction in the drag benefit experi­
In this section we start by discussing the results in terms of drag enced by the rear model at larger values of the inter-vehicle distance is
reduction obtained in a platoon constituted by 2 vehicles. We also report associated with the vortical structures generated by the leading vehicle;
the pressure distributions, the flow field organization of the near wake as argued by Pagliarella et al. (2007) the quasi-streamwise structures
and the spectral analysis of the pressure fluctuations. In addition to this, produced in the near wake of the leading vehicle (Cerutti et al., 2020;
we present the analysis reporting the drag variations for 3 and 4 models Pavia et al., 2018a) increase the pressure on the front of the rear vehicle,
in platoon configurations. thus progressively reducing the drag benefit. This only occurs for values
Here and in the following the reference frame is defined such that the of d/L sufficiently large to allow the production of the quasi-streamwise
X axis is aligned with the streamwise direction, the Z axis is aligned with structures, namelyd/L > 0.5.
the vertical direction (pointing upwards from the wind tunnel floor) and The pressure distributions on the model’s base (Fig. 2b) and along
Y completes the triad. the symmetry plane of the model (Fig. 3), expressed in terms of the
(p− p∞ )
pressure coefficient Cp = q∞ , with q∞ = 12 ρV∞
2
being the dynamic
pressure, are helpful to understand the results of Fig. 2a.
3.1. Drag variation and pressure distribution
As it can be inferred from Fig. 2b, for the isolated model the base
shows a value of the Cp nearly constant and equal to − 0.25 with a
The percentage drag variation ΔCD % = (1 − CpD /CD ), with CD being
downwards directed pressure gradient. Similar observations are valid
the drag coefficient of the isolated case and CpD the drag coefficient in the for the case ofd/L = 3, where the mutual interaction of the two models
platoon configuration, is reported in Fig. 2a as a function of the inter- progressively reduces, as also confirmed by the CD values.
vehicle distance d/L. As detailed in the previous section, we only had As the distance between the models reduces, the Cp distribution
one instrumented model. The dummy model was alternatingly posi­ across the base of the front model appears significantly affected. Atd/L
tioned at different locations ahead or behind it, obtaining the red and = 0.5, the pressure gradient across the base switches its orientation,
blue lines of Fig. 2a, respectively. Furthermore, to give a direct grasp of suggesting that the mean direction of the recirculation bubble is
the overall drag associated with the platoon configuration, the average reversed because of the effect of the approaching rear vehicle. At the
between these two values is also plotted and referred to as overall drag smallest investigated distance, d/L=0.125, the presence of a strong
variation (ODV, Zabat et al., 1995): lateral pressure gradient indicates the disappearance of the recirculation
()
∑n region and that the flow is now preferentially directed laterally, along
( ) d
d i=1 ΔCDi % L the Y direction.
ODV n, = (1) As expected, the strong effect that is experienced on the front vehicle
L n
is not reflected on the rear one, when d/L varies. However, a closer in­
where n represent the total number of vehicles, and i indicates the i-th spection of Fig. 2b reveals that, as d/L reduces, also on the vehicle’s base
model. there is evidence of a reversed pressure gradient.
Fig. 2a evidences that independently from the value of d/L an overall The analysis of the pressure distribution along the model’s symmetry
drag reduction is attained, with a largest value of 35% atd/L = 0.125 and plane and its lateral surface (Fig. 3) provides further information on the
the minimum value of 18% atd/L = 3. mechanism leading to the drag reduction documented in Fig. 2a. The
The rear vehicle benefits of the highest values of drag reduction with front model does not show any significant variations with respect to the
a maximum of 70% atd/L = 0.125 and 42% atd/L = 0.5, with a nearly isolated case: the flow withstands a pressure increase in the vicinity of
constant value in the range 0.75<d/L < 2.5. The large values of drag the model’s nose, with a pressure decrease occurring on the front slant
reduction on the rear vehicle when the inter-vehicle distance is small are surface as well as on the underbody of the model.
mostly related to the low pressure region originating in the lee of the Conversely, the rear model shows a progressive reduction of the low

5
J.J. Cerutti et al. International Journal of Heat and Fluid Flow 90 (2021) 108823

Fig. 5. Time averaged wake structure of the front model: contour representation of the dimensionless mean streamwise velocity U/V∞ with overlaid streamlines.
Data are measured in the symmetry plane of the model, Y = 0.

pressure region occurring on the top surface of the model. As it is also results. The larger scatter in the data regarding the rear vehicle is
documented by the pressure distribution on the lateral surface of the actually expected, being the data dependent on the characteristics of the
vehicle, when d/L reduces the flow field configuration is similar to the separated wake flow generated by the leading vehicle, as well as the
one that would be obtained by a vehicle of length equal to 2L, as indi­ effect of the front and rear vehicle geometries (Hammache and Bro­
cated by the absence of the low pressure region on the front slant that is wand, 2004).
typical of the nose of isolated vehicles. However, the obtained results show sufficient agreement with the
Besides with the pressure reduction on the front of the rear model, literature data, thus providing a reliable validation of our
this also justifies the overall drag reduction experienced in the pla­ measurements.
tooning configuration. It is well known that in the case of an isolated
vehicle, conditioning the flow to obtain a wake elongated in the 3.1.1. PIV analysis: Near wake structure of two-van platoon
streamwise direction is an effective methodology for drag reduction. We now turn our attention to the investigation of the flow field in the
This was also recently demonstrated by Minelli et al. (2020) who wake of the leading and rear vehicle when marching in platoon
investigated the suppression of wake’s lateral instability using active configuration, at values ofd/L = 0.125, 0.5, 3 and in the isolated case (as
flow control and found that the resulting wake was extended in the reference). We are particularly interested in evidencing the effect of the
streamwise direction and featured 20% drag reduction. This is also platooning on the large scale structures and their organization as a
exploited to reduce the pressure drag of bluff bodies with passive function of d/L.
methodologies such as boat tailing, or the introduction of short taper on In Fig. 5Figure 6 are presented the contour representations of the
the base of bluff bodies (Perry et al., 2016) time averaged streamwise velocity component U normalised with
In Fig. 4Figure 3 we compare our results in terms of drag coefficient respect to the freestream velocity V∞ for the front and the rear vehicles,
of the front (top) and rear (bottom) vehicle when marching in platoon respectively, obtained in the symmetry plane of the model (Y/W = 0).
configuration (CD ) normalised with respect to the drag coefficient of the The isolated case is also reported as a reference.
isolated case (CD0 ) to those obtained by (Romberg et al., 1971) using two The isolated case features the typical near wake structure of square
car models, (Zabat et al., 1994) using two, three and four van models, back bodies (withH/W > 1), with an asymmetric recirculation region
(Pagliarella et al., 2007) and Watkins and Vino (2008) both using and two counter rotating structures: a large one located near the top end
Ahmed bodies. of the model’s base (V2) and a smaller one (V1), whose core is slightly
A common feature of all the investigated cases is the loss of effec­ displaced downstream, at X/W ≈ 0.5. The former structure is generated
tiveness of the platoon for the front vehicles at values of the inter-vehicle by the flow undergoing separation from the top surface of the vehicle;
spacing>1L. The rear vehicle is the one characterized by the largest the latter is associated with the intense shear region between the high
scatter in the comparison between the available and the literature re­ momentum underbody flow and the velocity deficit in the vehicle’s
sults; nevertheless, our results are in good agreement with the literature wake.

6
J.J. Cerutti et al. International Journal of Heat and Fluid Flow 90 (2021) 108823

Fig. 6. Time averaged wake structure of the rear model: contour representation of the dimensionless mean streamwise velocity U/V∞ with overlaid streamlines. Data
are measured in the symmetry plane of the model, Y = 0.

A saddle point (S) can be detected at approximately Z/W = 0.4. The must be stressed that, only in the case of isolated vehicle, V3≡V1 and
streamwise extent of the recirculation region, defined as the region of V4≡V2.
the near wake where the streamwise velocity is less than or equal to zero At inter-vehicle distanced/L = 3, the near wake shows some signif­
(U/V∞ ≤ 0), is approximately 1W. icant differences with respect to the isolated case: the wake exhibits a
Atd/L = 3 the wake structure is similar to the one observed in the nearly symmetric configuration, with the lower vortex (V3) featuring a
isolated case. The same behaviour was observed for the base pressure in substantial growth. The saddle point (S) is consequently shifted upwards
Fig. 2b. A slight upward shift of the saddle point can be detected, which to Z/W = 0.5. This behaviour can be addressed to the less energetic
indicates that the near wake is only partly affected by the adverse underbody flow in the case of platoon configuration, thus allowing an
pressure gradient induced by the rear vehicle. upstream shift of the bottom vortex towards the model’s base.
As d/L decreases to 0.5, the lower structure (V1) becomes signifi­ Atd/L = 0.5, the near wake of the rear vehicle is substantially
cantly stronger. The approaching rear vehicle forces the flow towards affected by the different topology of the incoming flow. The interplay
the top surface, which seems to trigger a flow separation on the wind between the underbody flow and the one that undergoes the pressure
tunnel floor. This results into an upward shift of the saddle point (S) reduction on the top surface of the model generates a large recirculation
which is now located at Z/W = 1. Furthermore, the reversed direction of region with a core centred at Z/W = 1 and displaced downstream
the pressure gradient evidenced in Fig. 2b reflects into the fact that the (X/W = 0.5) and results to be elongated towards the wind tunnel floor.
base is largely dominated by V1, while the top vortex (V2) is progres­ The lower structure is attached to the base and reduced in size and
sively pushed upwards. enclosed by the larger top structure.
At the shortest distance, d/L = 0.125, a significant distortion of the At the smallest inter-vehicle distance investigated, d/L=0.125, the
flow structure can be detected. The adverse pressure gradient induced effect of the flow undergoing the pressure reduction on the vehicle’s roof
by the rear vehicle causes the expulsion of both the top vortex (V2) and significantly affects the rear wake topology. As observed in Fig. 6, atd/L
the saddle point (S). Given the space constraint imposed by the rear = 0.125 the flow is forced towards the vehicle’s roof, generating a flow
vehicle, the lower vortex (V1) is also reduced in size. It must be pointed separation from the wind tunnel floor on the aft part of the rear vehicle.
out that the tendency of the flow to separate from the floor could have an The low momentum fluid in the wake of the rear vehicle is entrained
effect on the overall wake balance. However, with the present data it is within the flow coming from the vehicle’s roof, effectively causing: (i) an
not possible to determine the extent of this effect. alteration of the top vortex (V4), which reduces its size and gives way to
The wake of the rear vehicle is reported in Fig. 6, where we also the growth of the lower structure (V3); (ii) a flow separation (V5). This
report the already discussed isolated case for comparison purposes. We feature, that is found also in the isolated case for similar values of the
indicate as V3 and V4 the bottom and top vortices, respectively. We non-dimensional ground clearance (G/W, Grandemange et al., 2013a) at
labelled these structures differently from V1 and V2 in the isolated case larger distances from the vehicle (X/W = 1.5), occurs much closer to the
to avoid confusion when the inter-vehicle distance reduces. However, it base in platooning configuration.

7
J.J. Cerutti et al. International Journal of Heat and Fluid Flow 90 (2021) 108823

argue that the near wake in platoon configuration is generally more

turbulent than the equivalent isolated case. While this increase affects
the whole range of measured non-dimensional frequencies for the right
and bottom microphones, featuring a broadband increase of the PSD
intensity with respect to the other values ofd/L, in the case of the top
microphone a large share of the spectral energy is associated with the
high frequencies; this suggests that atd/L = 0.125 the flow is charac­
terized by smaller scale structures with respect to the other cases. It is
also relevant to notice that for the front vehicle the peak associated with
the vortex shedding is progressively suppressed when d/L decreases.
The pressure fluctuation signals measured on the base of the rear
vehicle show much less dependence on d/L, with values of the PSD very
close to the isolated wake. The only exception is related to the bottom
microphone atd/L = 0.125, where the PSD shows an intensity reduction
across the whole range of investigated frequencies. While this may seem
surprising, it is possible to argue that the strong differences experienced
in the near wake of the rear vehicle are mostly associated with features
that are located further downstream, with an effective lower activity in
the close proximity of the base.
Fig. 7. Schematic representation of the location of the reference microphones.
It is also worth to mention that, for all the investigated cases, a
second peak occurs at St ≈ 0.7. Identifying the origin of such peak is not
The base pressure associated to the latter wake configuration straightforward. It is the authors’ opinion that this high frequency peak
(Fig. 2b), along with the separated flow on the front part of the vehicle is associated with small scale structures generated in the shear region
gives rise to highest value of drag reduction, as large as 70%. between the low momentum recirculation bubble and the incoming flow
from the vehicle’s roof.
3.1.2. Spectral analysis of the pressure signals
The time series recorded by the microphonic probes mounted on the 3.1.3. Proper orthogonal decomposition analysis of the pressure
model’s base were analysed in the frequency domain to characterize the fluctuations
features of the wake unsteadiness. Three representative pressure taps The simultaneous measurements of the twelve microphones across
were considered, located near the top, right and bottom edge of the base, the model’s base were used to perform a Proper Orthogonal Decompo­
as schematically represented in Fig. 7. sition (POD)-based analysis. The purpose of this analysis is twofold:
In Fig. 8 we report the power spectral density (PSD) in the case of the firstly, it allows us to determine the most energetic modes in the isolated
front and rear vehicle for values ofd/L = 0.125, 0.5, 3 and isolated wake. case, serving the purpose of comparison with the literature; secondly, it
In the isolated case, the PSD shows a peak at a value of the Strouhal allows us to determine what is the effect of the distance d/L on the
number St = fW/V∞ ≈0.12. This value is typical of the vortex shedding wake’s most energetic modes.
phenomenon from a broad range of bluff bodies, as widely documented The POD technique decomposes a temporal signal into modes and
in the literature (Grandemange et al., 2013a, 2013b; Volpe et al., 2015; modal coefficients, as explained by Berkooz et al. (1993). This decom­
and Pavia et al., 2018a, 2018b). position considers that a given signal p’(Y, Z, t) may be decomposed into
The PSD of the front vehicle shows increasing and broader levels of the sum of a number of modes k, which are constituted by a temporal
the spectral energy as the rear vehicle approaches; it is then possible to coefficient ak (t) and a spatial eigenfunction φk (Y, Z) as follows:

Fig. 8. Power spectral density of the pressure taps’ fluctuations in three representative locations of the base. Data are shown for the front and rear vehicle at values of
d/L = 3, d/L = 0.5, d/L = 0.125. The isolated case is also reported as reference.

8
J.J. Cerutti et al. International Journal of Heat and Fluid Flow 90 (2021) 108823

Fig. 9. POD analysis applied to the pressure signals for the front model. Contour plots obtained for the cases of isolated vehicle and d/L = 3, d/L = 0.5 and d/L
= 0.125.

∑
p’(Y, Z, t) = ak (t)φk (Y, Z) (3) evolution. This also suggests that the most relevant shedding mecha­
k nism surviving the adverse pressure gradient imposed by the rear
vehicle is the horizontal one.
The eigenfunctions φK (x, y) need to satisfy the condition
∫ Furthermore, the expulsion of the structure originating near the top
〈p’(Y, Z, t)⋅p’(Y, Z, t)T 〉t ⋅φj (Y, Z)dydz = λj ⋅φj (Y, Z) (4) edge of the base (V2) evidenced by the flow field results of Fig. 6 along
with the disappearance of the flat mode in the POD results of Fig. 9
where j ranges from 1 to the total number of microphones, 12. suggest that this structure can be considered as the main responsible for
Fig. 10-Fig. 9 show the first four POD modes across the base of the the pumping mechanism. A different behaviour is also observed in the
vehicle in the isolated case (first row) and for values of d/L = 3, 0.5 and case ofd/L = 0.5 for the modes 3 and 4, where the HSB and the sym­
0.125, for the front and rear vehicle, respectively. metric modes observed in the isolated andd/L = 3 cases are replaced by
In the isolated case, the first and most energetic mode is a flat mode, more complex patterns.
often referred to as pumping, characterized by a low frequency From Fig. 10 we can immediately observe that, in the case of the rear
streamwise oscillation of the whole wake. Second and third mode, are vehicle, the modes are not significantly affected by the value of d/L.
representative of a vertical symmetry breaking (VSB) mode and hori­ The comments for the isolated case will then be extended also to the
zontal symmetry breaking (HSB) mode, which are associated with the other cases.
top down and right left shedding of the wake. Similar results were evi­ The only remarkable difference on the rear vehicle is obtained in the
denced by Pavia et al. (2018a), Pavia et al. (2018b). analysis of the fourth mode. While the isolated case, as well asd/L = 3
Across the base of the front vehicle, while the pumping mode is and 0.5 show a symmetric mode (suggesting in-phase shedding), atd/L
retained atd/L = 3 andd/L = 0.5, at the shortest distance it is replaced by = 0.125 the fourth mode exhibits a more complex behaviour which
an HSB mode. As it was observed from the time averaged flow field cannot be referred to neither as symmetric nor asymmetric mode.
results, the wake atd/L = 0.125 is significantly affected and, in the Besides the qualitative representation, it is also relevant to underline
symmetry planeY/W = 0, undergoes a preferential bottom-top how the energy is distributed across these four modes, depending on the
value of d/L. In Fig. 11 the energy distributions of the POD modes are

9
J.J. Cerutti et al. International Journal of Heat and Fluid Flow 90 (2021) 108823

Fig. 10. POD analysis applied to the pressure signals measured on the base of the rear model. Contour plots obtained for the cases of isolated vehicle and d/L = 3, d/
L = 0.5 and d/L = 0.125.

shown for the front and rear vehicles. As it can be observed, the first four 3.2. Three and four-van platooning
modes account for about 95% of the total energy.
For the front model, atd/L = 3 (Fig. 11a) no significant differences Fig. 12 shows the percentage drag variation of the platoon configu­
can be detected with respect to the isolated case on the way the energy is ration when three (a) and four (b) vehicles are considered, respectively.
distributed across the modes, in good agreement with the flow field and Given the limitations imposed by the size of the test section, the
pressure distribution results. maximum allowed distance between models is d/L = 1.5. Despite this
For values ofd/L < 3, the first mode shows decreasing values of en­ limitation, the results allow to make significant observations about the
ergy: this is expected since the top vortex (V2), that we associate to the benefits that can be obtained in the platoon configuration.
pumping mechanism, is progressively reduced as d/L reduces and dis­ As it can be seen from Fig. 2: a) Percentage variation of the drag
appears whend/L = 0.125. This share of modal energy is eventually coefficient ΔCD for the two models platoon configuration as a function
distributed across the higher order modes, evidencing a more uniform of the distance between the models d/L: front model (Blue), rear model
distribution as d/L decreases. (Red), overall drag variation (dashed line). b) Pressure coefficient dis­
For the rear model, in the light of the results of Fig. 10, the only tribution on the base of the front (top row) and the rear model (bottom
relevant difference in the way the energy is distributed across the most row). The isolated case is also reported as a reference. The location of the
energetic modes is observed atd/L = 3. The first mode results much more pressure taps is indicated with black dots. The contour plots are obtained
energetic than the other cases. As observed from the time-averaged PIV by interpolating the measurements on a structured grid.
results this case is characterized by a symmetric configuration with the a and b, the presence of more vehicles in the platoon is associated
sizes of the top and bottom structures being comparable, which might with a drag increase for the trailing vehicle, which makes this last
justify this energy enhancement. However, the current results do not location the least favourable in the platoon. In fact, those vehicles which
allow to draw a clear conclusion on this point which will remain open for are located in the middle are those which exhibit the highest values of
future investigations. drag reduction (model 2 in Fig. 12a and model 2 and 3 in Fig. 12b).

10
J.J. Cerutti et al. International Journal of Heat and Fluid Flow 90 (2021) 108823

Fig. 11. Energy of the POD modes normalised with respect to the total modal energy of the flow: a) front model and b) rear model. The dashed lines refer to the
isolated vehicle.

Fig. 12. Percentage variation of the drag coefficient ΔCD for the (a) three-models platoon and (b) four-models platoon configurations as a function of the distance
between the models d/L.

Conversely, the front model does not show any significant dependency modelling of the drag reduction obtained by a generic n-vehicles pla­
on the number of elements in the platoon. toon. This simplistic model is indeed based on the idea that the inter­
In the case of three vehicles (Fig. 12a) the middle one benefits of a mediate vehicles will be characterized by the same values of drag
further share of drag reduction of about 5% with respect to the rear reduction independently from the value of n. This aspect will be further
vehicle and 15% with respect to the front one. This result is found quite expanded in the next section.
independent on the value of the inter-vehicle distance. A last remark is related to the drag reduction associated with small
In the case of four models, the two intermediate vehicles benefit from values of d/L. In the light of the advancements in autonomous driving,
the highest drag reductions, evidencing an extremely similar behaviour the close proximity might represent a solid possibility leading to sig­
and reaching atd/L = 0.125 the same drag reduction (85%) of the nificant values of drag reductions if appropriately exploited. However,
middle vehicle for the case with three vehicles. in the current scenario, this path can be hardly considered as the high
Fig. 12 shows that, as the number of vehicles in the platoon increases, benefit evidenced in terms of drag reductions could be overthrown by
the drag variation for the intermediate vehicles is independent from the safety and hazard concerns.
total number of vehicles in the platoon, thus paving the way to the

11
J.J. Cerutti et al. International Journal of Heat and Fluid Flow 90 (2021) 108823

Fig. 13. a) Percentage drag variation for the front, tail and middle vehicles as a function of d/L for the two, three and four vehicles’ platoon; the dashed lines are
representative of the spline interpolation of the data, assuming that atd/L = 5 the drag variation is equal to zero. b) Overall drag variation plotted for an increasing
number of vehicles in the platoon n and fixed values ofd/L = 0.125, d/L=0.5 andd/L = 3. c) Overall drag variation for an increasing number of vehicles in the platoon
n, rescaled with respect to the value obtained for n→∞. The dashed red line is representative of a user defined target of drag reduction set to 90% of the maximum
possible drag reduction. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

()
4. A simple modelling for a convoy with n-vehicles
d/L between models and for a number of vehicles equal to n. ΔVFront dL ,
() ()
The investigation in a wind tunnel on the platooning configuration is
ΔVTail dL and ΔVMiddle dL are the drag variations at a value of d/Lof the
limited by the maximum number of vehicles that the length of the test
section allows. In order to compensate this, Zabat et al. (1995) devel­ front, rear and intermediate models, respectively.
oped an overall drag variation law for a generic platoon with n-vehicles. Fig. 13b shows the overall drag variation for a growing number of
Considering the data from the front, middle and rear vehicles of the 3 models n at fixed values of d/L (0.125, 0.5 and 3). For a number of
and 4 vehicles platoon configurations reported in Fig. 12, and the result models n > 8, for the analysed values of d/L it is not possible to obtain
of Fig. 4 showing that ford/L = 5 there is no significant effect in terms of further improvements of the drag reduction.
drag reduction(%ΔCD = 0) for the two vehicles convoy, it is possible to As it can be observed, for a given distance d/L the drag reduction
perform a best fit using a spline function of these data as shown in increases with the number of vehicles in the platoon reaching an
Fig. 13a. In particular, all the front, rear and intermediate vehicles’ data, asymptotic value which depends on the value of d/L. The asymptotic
independently on the number of vehicles in the convoy, where consid­ value can be calculated from equation (5), for n→∞ obtaining
() ()
ered and used for the fit.
Under these assumptions, the overall drag variation can be written ODV dL = DVMiddle dL , which is effectively representative of the
n→∞
as: highest value of drag reduction that could be obtained for that particular
() () () configuration.
ΔDFront Ld + ΔVRear Ld + (n − 2)ΔVMiddle Ld Fig. 13c shows the overall drag variations normalized with respect to
ODV(d/L, n) = (5) the asymptotic value pertaining to each value of d/L as a function of the
n
number of vehicles n. This is helpful to determine what is the minimum
where ODV(d/L, n) is the overall drag variation for a constant distance number of vehicles in the platoon necessary to obtain a given value of

12
J.J. Cerutti et al. International Journal of Heat and Fluid Flow 90 (2021) 108823

the maximum drag reduction, for a fixed value of the distance d/L. For References
the safety considerations that were previously raised, a 25% drag
reduction (90% of the maximum achievable) can be obtained with a Al Alam, Assad, Ather Gattami, Karl H. Johansson, and Claire J. Tomlin. 2011. 44 IFAC
Proceedings Volumes (IFAC-PapersOnline) Establishing Safety for Heavy Duty
convoy of 18 vehicles with inter-vehicle distance equal tod/L = 3. Vehicle Platooning: A Game Theoretical Approach. IFAC.
Alam, Assad, Jonas Martensson, and Karl H. Johansson. 2013. “Look-Ahead Cruise
Control for Heavy Duty Vehicle Platooning.” IEEE Conference on Intelligent
5. Conclusions Transportation Systems, Proceedings, ITSC (Itsc): 928–35.
Astarita, T., 2006. Analysis of Interpolation Schemes for Image Deformation Methods in
An experimental study aimed at determining the flow field charac­ PIV: Effect of Noise on the Accuracy and Spatial Resolution. Experiments in Fluids 40
(6), 977–987.
teristics of light commercial road vehicles marching in platoon config­
Astarita, T., 2007. Analysis of Weighting Windows for Image Deformation Methods in
uration has been carried out. Drag measurements, as well as pressure PIV. Experiments in Fluids 43 (6), 859–872.
distribution and flow field organization in the near wake are considered. Astarita, T., 2008. Analysis of Velocity Interpolation Schemes for Image Deformation
The drag measurements allow us to confirm that the effectiveness of Methods in PIV. Experiments in Fluids 45 (2), 257–266.
Bergenhem, C., Hedin, E., Skarin, D., 2012. Vehicle-to-Vehicle Communication for a
the platooning configuration reduces for inter-vehicle distancesd/L > 3. Platooning System. Procedia - Social and Behavioral Sciences 48, 1222–1233.
This also reflects in the flow field organization observed atd/L = 3, as Berkooz, G., Holmes, P., Lumley, J.L., 1993. The Proper Orthogonal Decomposition in
well as in the distribution of the modal energy obtained from the mi­ the Analysis of Turbulent Flows. Annual Review of Fluid Mechanics 25 (1), 539–575.
http://www.annualreviews.org/doi/10.1146/annurev.fl.25.010193.002543.
crophonic probes. The range of inter-vehicle distances between 0.5 and Bonilla, D., 2020. Road Freight Transport and Energy Use: The USA, China, the EU,
1 can be particularly appealing for drag reduction purposes, especially Japan and Germany. In 5–32.
in the light of future advancements in autonomous driving. We show Cafiero, Gioacchino, Juan José Cerutti, and Gaetano Iuso. 2019. “Near Wake Structure of
a Square Back Road Vehicle.” In 13th International Symposium on Particle Image
that the presence of a leading edge slant angle, does not affect signifi­ Velocimetry – ISPIV 2019,.
cantly the drag reduction obtained on the leading vehicle of the platoon. Castelain, T., Michard, M., Szmigiel, M., Chacaton, D., Juvé, D., 2018. Identification of
However, the comparison with literature results suggests that the drag Flow Classes in the Wake of a Simplified Truck Model Depending on the Underbody
Velocity. Journal of Wind Engineering and Industrial Aerodynamics 175, 352–363.
reduction attainable by the rear vehicle can be dependent on the leading Cerutti, J.J., Cafiero, G., Iuso, G., 2019. Aerodynamics of Two Square Back Vehicles in
edge slant angle. Platooning Configuration. Measurements and Visualization, Naples (Italy).
We detail the flow field structures that can be observed in the two Cerutti, J.J., Sardu, C., Cafiero, G., Iuso, G., 2020. Active Flow Control on a Square-Back
Road Vehicle. Fluids 5 (2), 55. https://doi.org/10.3390/fluids5020055.
vehicle platoon configuration and see that the effect of the reduction of
Davila, A., Aramburu, E., Freixas, A., 2013. Making the Best out of Aerodynamics:
the inter-vehicle distance on the near wake of the front vehicle is mostly Platoons. SAE Technical Papers 2, 4–9.
associated with the expulsion of the otherwise dominant vortex (labelled Gehring, Ottmar, and Hans Fritz. 1997. Practical Results of a Longitudinal Control
Concept for Truck Platooning with Vehicle to Vehicle Communication. In: IEEE
as V2). The joint analysis of the PIV and POD data from the microphonic
Conference on Intelligent Transportation Systems, Proceedings, ITSC: 117–22.
probes also suggests that the V2 structure is the one responsible for the Grandemange, M., Gohlke, M., Cadot, O., 2013a. Bi-Stability in the Turbulent Wake Past
pumping mechanisms, given that at values ofd/L = 0.125 there is the Parallelepiped Bodies with Various Aspect Ratios and Wall Effects. Physics of Fluids
suppression of the flat mode. 25 (9), 095103. https://doi.org/10.1063/1.4820372.
Grandemange, M., Gohlke, M., Cadot, O., 2013b. Turbulent Wake Past a Three-
The nature of the substantial drag reduction of the rear vehicle in the Dimensional Blunt Body. Part 1. Global Modes and Bi-Stability. Journal of Fluid
platoon configuration is confirmed to be the reduction of the static Mechanics 722, 51–84.
pressure on the front of the vehicle. At values ofd/L = 0.125 the flow is Hamiga, W., Ciesielka, W., Banaś, M., Pająk, T., 2018. Numerical Modeling of Airflow
over Column of Vehicles Using Ansys® Package. E3S Web of Conferences 46, 00025.
substantially affected by the adverse pressure gradient induced by the https://doi.org/10.1051/e3sconf/20184600025.
rear vehicle and undergoes a bottom-top evolution. Hammache, M., Browand, F., 2004. On the Aerodynamics of Tractor-Trailers. Trucks,
The wake of the front model in platoon configuration is found to be Buses and Trains, In Aerodynamics of Heavy Vehicles, pp. 185–205.
He, M., Huo, S., Hemida, H., Bourriez, F., Robertson, F.H., Soper, D., Sterling, M.,
generally more turbulent, with a progressive suppression of the vortex Baker, C., 2019. Detached Eddy Simulation of a Closely Running Lorry Platoon.
shedding experienced atSr = 0.12 in the isolated case. Journal of Wind Engineering and Industrial Aerodynamics 193, 103956. https://doi.
In the case of three and four-van platoon, we observe that regardless org/10.1016/j.jweia.2019.103956.
Hucho, W., Sovran, G., 1993. Aerodynamics of road vehicles. Annual Review of Fluid
of the number of vehicles in the platoon, the intermediate vehicles show
Mechanics 25, 485–537.
similar values of drag reduction. We use this information to extrapolate Michaelian, M., Browand, F., 2000. Field Experiments Demonstrate Fuel Savings for
the overall drag variation for a growing number of vehicles and quickly Close-Following. University of California at Berkeley 25.
Minelli, G., Dong, T., Noack, B.R., Krajnović, S., 2020. Upstream Actuation for Bluff-Body
determine the minimum number of vehicles necessary to obtain a target
Wake Control Driven by a Genetically Inspired Optimization. Journal of Fluid
value of the drag reduction for a fixed inter-vehicle distance. Mechanics 893. https://doi.org/10.1017/jfm.2020.220.
The importance of the flow field characterization of vehicles Pagliarella, R.M., Watkins, S., Tempia, A., 2007. Aerodynamic Performance of Vehicles
marching in platoon is also confirmed by the strong interest of the in Platoons: The Influence of Backlight Angles. SAE Technical Papers 1 (724).
Pavia, G., Passmore, M., Sardu, C., 2018a. Evolution of the Bi-Stable Wake of a Square-
research councils which encourage and support research in the field, Back Automotive Shape. Experiments in Fluids 59 (1). https://doi.org/10.1007/
with the development of dedicated test rigs (TRAIN at the University of s00348-017-2473-0.
Birmingham, for example). Furthermore, the European Union has Pavia, G., Passmore, M., Sardu, C., 2018b. Evolution of the Bi-Stable Wake of a Square-
Back Automotive Shape. Experiments in Fluids 59 (1), 20. http://link.springer.
identified with the smart mobility one of the key drivers of the forth­ com/10.1007/s00348-017-2473-0.
coming programme Horizon Europe. Perry, A.-K., Pavia, G., Passmore, M., 2016. Influence of Short Rear End Tapers on the
In this work, besides providing significant support for previous nu­ Wake of a Simplified Square-Back Vehicle: Wake Topology and Rear Drag.
Experiments in Fluids 57 (11). https://doi.org/10.1007/s00348-016-2260-3.
merical investigations and exploring a broader range of inter-vehicle Romberg, G.F., Chianese, F., Lajoie, R.G., 1971. Aerodynamics of Race Cars in Drafting
distances, we also evidence the importance of future developments in and Passing Situations. In SAE Technical Papers.
control and autonomous driving. Exploiting the capabilities of the pla­ Salari, Kambiz, and Jason Ortega. 2018. “Experimental Investigation of the Aerodynamic
Benefits of Truck Platooning.” SAE Technical Papers 2018-April: 1–11.
tooning configuration in close proximity might pave the way towards a
Sardu, C., Lasagna, D., Iuso, G., 2016. Noise Filtering for Wall-Pressure Fluctuations in
much more sustainable road transport and to associated reductions in Measurements around a Cylinder with Laminar and Turbulent Flow Separation.
the vehicles’ emissions. Journal of Fluids Engineering, Transactions of the ASME 138 (6).
Tsuei, L., Savas, Ö., 2000. A Wind Tunnel Investigation of the Transient Aerodynamic
Effects on a Four-Car Platoon during Passing Maneuvers. SAE Technical Papers 1
(724).
Declaration of Competing Interest Uystepruyst, D., Krajnović, S., 2013. LES of the Flow around Several Cuboids in a Row.
International Journal of Heat and Fluid Flow 44, 414–424.
Vegendla, Prasad et al. 2015. “Investigation of Aerodynamic Influence on Truck
The authors declare that they have no known competing financial
Platooning.” SAE Technical Papers 2015-Septe.
interests or personal relationships that could have appeared to influence
the work reported in this paper.

13
J.J. Cerutti et al. International Journal of Heat and Fluid Flow 90 (2021) 108823

Volpe, R., Devinant, P., Kourta, A., 2015. Experimental Characterization of the Unsteady Mechanical Engineers, Part D: Journal of Automobile Engineering 230 (9),
Natural Wake of the Full-Scale Square Back Ahmed Body: Flow Bi-Stability and 1273–1285.
Spectral Analysis. Experiments in Fluids 56 (5), 1–22. Zabat, Michael, Stefano Frascaroli, and F. K. Browand. 1994. “Drag Measurements on 2,
Watkins, S., Vino, G., 2008. The Effect of Vehicle Spacing on the Aerodynamics of a 3 and 4 Car Platoons.” In SAE Technical Papers.
Representative Car Shape. Journal of Wind Engineering and Industrial Zabat, M., Stabile, N., Farascaroli, S., Browand, F., 1995. The Aerodynamic Performance
Aerodynamics 96 (6-7), 1232–1239. Of Platoons: A Final Report. California Partners for Advanced Transit and Highways
Yu, K., Liang, Q., Yang, J., Guo, Y., 2016. Model Predictive Control for Hybrid Electric (PATH).
Vehicle Platooning Using Route Information. Proceedings of the Institution of

14