Signal Processing - November 2016 - 75

In this article, we give an overview of the coupling between
improve safety, energy, and traffic efficiency on the existing
control, communication, and sensing, as visualized in Figure 1.
and planned road infrastructure.
We provide a survey of the different control approaches and
Many of the aforementioned problems are caused by
their associated signal processing challenges. We hope that
the human involvement in the coordination of traffic. Studthis article can provide an introduction for signal processing
ies have shown that over 90% of traffic accidents are comprofessionals to the control-theoretic aspects of vehicle coorpletely, or in part, due to human error [2]. This has led to a
dination and pave the way for a tighter collaboration.
progressive shift in responsibilities from the human driver
to dedicated control systems, most recently in the form of
autonomous vehicles, which aim to provide more efficient,
Problem formulation
comfortable, and virtually accident-free road traffic. AutonoThe problem of coordinating a set of vehicles can be phrased
mous vehicles are still limited in terms of their sensing and
as calculating the best control trajectories for the individual
coordination capabilities, as their actions depend on the
vehicles that allow them to safely reach their destination in
on-board sensory data and models of other
finite time (e.g., within a few tens of secvehicles' behavior. As an example, a sumonds). In general, any solution should meet
We give an overview of
mary of the Urban Grand Challenge [3]
the basic requirements of safety (i.e., no
the coupling between
mentioned that a number of incidents could
collisions occur) and liveness (i.e., desticontrol, communication,
have been avoided if vehicles could anticinations are reached eventually), while
and sensing for control
pate the behavior of other vehicles, and
optimizing some performance metric.
of automated vehicles.
that vehicles should cooperate for autonoThe most important requirement is safety.
mous driving to reach its full potential. The
Hence, vehicles may never be steered to
benefit of cooperation was already recognized in a parallel
states from which future collisions are unavoidable. Second,
track in vehicle automation, and platooning, which instead of
the coordination algorithm must guarantee that all vehicles are
complete autonomy promotes information sharing between
allowed to both enter and exit the coordination area in finite
vehicles and joint decision making. In a platoon, the vehicles
time so that permanent stops and traffic deadlocks are avoidrely on vehicle-to-vehicle (V2V) and vehicle-to-infrastruced. Finally, a performance criterion is necessary to favor one
ture (V2I) communication to share information regarding the
among multiple solutions. In summary, a coordination probenvironment and internal states and choose safe and efficient
lem can be stated as a constrained optimal control problem,
control policies jointly [4].
where a performance criterion is optimized with respect to the
The two tracks for automating vehicles have thus followed
vehicles' control input trajectories, subject to safety and livedifferent approaches: one (the platooning track) explicitly
ness requirements:
relies on communication among vehicles, while the second
minimize
performance criterion
(1a)
(the autonomous vehicle track) does not. With the adoption
of the IEEE 802.11p Standard, as well as the possibilities of
subject to
safety constraints,
(1b)
V2V and V2I communication and other services under the
liveness constraints.
(1c)
future 5G wireless standard [5], the two tracks are expected to merge, leading to a new type of large-scale wireless
The constrained optimal control framework clearly allows
networked control system. This merging will likely take
one to conveniently accommodate performance, safety, and
place in a piecemeal fashion, with
first a ubiquitous availability of wireless communications, and only later
the gradual introduction of cooperaCommunicate
tive autonomous vehicles [6]. These
Map, Actors, and
vehicles will drive autonomously, but
Control Actions
at the same time be able to leverage
their communication capabilities for
cooperative planning and control, as
well as cooperative perception and
sensing, thus eliminating many of the
traffic-safety and efficiency problems.
The design and operation of such netCooperative
Cooperative
works of cooperating vehicles place
Estimation,
Control for
enormous demands on the control,
Localization,
Safety-Critical
communications, and sensing subsysMapping,
Traffic
tems, as they must operate in harmony
and Prediction
Optimization
across different brands and types of
Figure 1. Vehicle coordination relies on a tight interaction between control, communication, and sensing.
vehicle, with limited margin for error.
IEEE SIgnal ProcESSIng MagazInE

November 2016

Table of Contents for the Digital Edition of Signal Processing - November 2016