192 PART | V The future of ITS applications
open research issues and decisions to be taken especially in the intermediate
autonomy levels, where the autonomous vehicle has to ask the driver to inter-
vene [request to intervene (RTI)], expecting the intervention to occur in a timely
manner (Inagaki & Sheridan, 2018). For example, in SAE’s Autonomy Level
3, the readiness of the driver when the driving automation issues the RTI must
be considered along with the driver’s performance to take over control, and the
time pressure that the RTI puts on the driver (Happee, Gold, Radlmayr, Hergeth
& Bengler 2017). It is also important along with the RTI message to provide
information that will assist the driver to control the car on his/her intervention
(Petermeijer, Cieler & De Winter 2017).
17.2 Technology enablers
In order to achieve the aforementioned autonomy levels, car manufacturing
companies equip vehicles with several automations, intelligent decision-mak-
ing mechanisms and advanced driver-assistance systems (ADAS) that facilitate
driving, take control of the car under specific conditions or notify the driver to
take control when necessary. Sensors become the eyes and ears of the driver
and embedded computing platforms become their brain and nervous system,
and the effort remains on making this technology reliable, fault-tolerant, and
stable at any condition.Several studies (Banks, Eriksson, O’Donoghue, &
Stanton, 2018) and reports (US Department of Transportation, 2017) on Tesla,
BMW, and Mercedes-Benz and other autonomous cars provide useful details
on the operation of the various in-vehicle systems and how they collaborate for
providing autonomy. They also define the areas that require further examination
(NHTSA, 2013 ). At the same time, companies such as Lyft together with car-
makers such as General Motors and technology development companies such
as Waymo, start new projects and introduce smart technologies that aim to make
autonomous driving to the higher levels of autonomy.
The Automatic Emergency Braking (AEB) of Tesla Model S, uses a radar/
camera fusion module, which has been designed to prevent accidents by early
predicting front-to-rear collisions. The module uses an image classification
model that has been trained using a large dataset of rear vehicle images and is
able to detect the distance from the vehicle in front. The output of the image
detection system is fed to the decision making system that includes a forward
collision warning system that monitors how the vehicle is moving forward, a
dynamic brake support system that supplements insufficient braking from the
driver and a crash imminent braking system that automatically activates the
breaks if a hazard is detected. Past NHTSA’s reports on several AEB systems
not only validated the effectiveness of the collision avoidance technologies in
many cases but also identified several crash modes that were not validated such
as the collisions related to straight movement or left turn that crosses another
vehicle’s path. Concerning rear-end collision the study focused on the cases
where the leading vehicle stops, decelerated or was moving during the accident.