Content uploaded by Marc-Antoine Fénart
All content in this area was uploaded by Marc-Antoine Fénart on Sep 10, 2021
Content may be subject to copyright.
Road infrastructure digitalization for automated vehicles
BETEND Loan1, Prof. FENART Marc-Antoine1
1HES-SO – HEIA-FR, Fribourg, Switzerland
Automated vehicles developers are investing and working on improving the algorithms that enable the vehicle to
detect markings and signs that it must respect. However, due to problems detection we encounter nowadays, these
algorithms do not yet allow us to envisage a level of automation 5 in which the driver no longer intervenes.
Several concrete examples can be mentioned here; Pavement longitudinal joints repairs can mislead the algorithm
into thinking it is detecting a continuous line and thus moving the vehicle away from its ideal position, or even
putting it in danger; In other cases, markings may be contradictory, degraded or non-existent (e.g. in the case of
unmarked central lanes), or the reduction of markings in 30 km/h zones in urban areas may cause problems for
line detection algorithms and Elon Musk recently confirmed this : “Moreover, standard Autopilot would require
lane lines to turn on, which this street did not have” . In Switzerland, several accidents have been reported on
the motorway construction sites involving semi-autonomous vehicles that have not respected the markings or signs
in place, even hitting buffer trucks (set up to protect people working on the site) .
The solution envisaged here consists of digitizing the transport infrastructure, or the transport infrastructure
modifications (construction phases) in order to transmit this information to the automated vehicles. Recent progress
made at the HEIA-FR (SwissMoves Group) has made it possible to define the position of the vehicle with an
accuracy less than ten centimeters. Furthermore, road infrastructures are already highly digitized and relatively
accurate in GIS (Geographic Information Systems) or CAD (Computer Aided Design) tools.
As part of a master's thesis results , we were able to convert the plans drawn by civil engineers (markings) into
a map compatible with the route planning algorithm used by SwissMoves for its automated vehicles, Lanelet2 ,
with the following process:
Fig. 1 Conversion process
The first three steps are carried out by draughtsman. When designing the plan, requirements (drawing process)
have been established to ensure that all necessary information is included in the draw. While this requires extra
work, we have been careful to minimize this. Then the developed converter transforms the draw into a file
compatible with the automated vehicles path planning algorithm. Thanks to this process, the map that the civil
engineer uses to draw lines on the roads and the map that the autonomous vehicles need are identical, which means
that the map is up to date and the drawing work is done only once, as opposed to twice actually.
If the current results show that the conversion of the markings plan into a useful file by the path planning algorithm
works and that the information extracted from the draw is well represented in the map understood by the automated
vehicles, the future goal is to add the conversion of the road signs in order to obtain fully auto-converted signs
(e.g. speed limits, road priorities).
 Elon Musk, Twitter, 19.04.2021
 RTS, “Sécurité des voitures semi-autonomes en question après des accidents”, 06.11.2019
 L. Bétend, J. Supcik, M.-A. Fénart, “AutoMAPi – Tool for processing infrastructure plans for autonomous
vehicles”, Master of Science HES-SO in Engineering, L. Bétend, 04.06.2021
 Fabian Poggenhans et al. “Lanelet2: A High-Definition Map Framework for the Future of Automated
Driving”. In: Proc. IEEE Intell. Trans. Syst. Conf. Hawaii, USA, 2018.
Base road draw Adding information for
the path planning Saving the map in a
the file into a path-
Use this map in the path