exploited. The ADS-B exploitation is currently
object of intense work by several international
institutions (see e.g. SESAR Work Package 5 (Sesar,
2012)). The ARCA algorithm thus does not need
additional data or interaction, but it exploits already
available signals for a new and valuable purpose.
In the second section of the paper the algorithm
is briefly reviewed. In the third section the available
experimental result are reviewed together with the
used experimental setup. In the fourth section the
implementation on board the Embedded Processing
Unit (EPU) is presented together with the CANOpen
implementation of the on board communications
among the various devices composing the system;
hardware-in-the-loop tests are also reported. In the
fifth section of the paper the conclusions are drawn.
2 THE ARCA ALGORITHM
The ARCA algorithm is inspired by Game Theory
and in particular it is based on the Satisficing Game
Theory approach (SGT) (Stirling, 2003). Basically,
SGT computes the best choice among various
alternatives by means of two utility functions,
rejectability and selectability, which respectively
represent costs and benefits for each agent while
making a choice. There may exist dependence
between utility functions of different agents that
implements an “altruistic” behaviour.
This approach has already proven to be effective
in dealing with several real-world problems (e.g.
packet routing (Boyan et al., 1994) and (Choi et al.,
1996), transport logistics (Wolpert et al., 1999),
automated car driving (Stirling, 2003), airborne self-
separation (Bellomi et al., 2008)) by providing
robust and dependable solutions that can be achieved
with limited computational resources.
In the UAV context each UAV can be considered
as an agent working under the SGT framework.
Selectability and rejectability of each agent are
respectively the benefits and costs of the UAV’s
available manoeuvring choices. Benefits are
essentially proportional to the optimality of the
possible route to reach the final destination (sortest
possible delay). Costs are proportional to the risk of
infringing the separation with other vehicles. Each
aircraft computes the rejectability and selectability
of each considered manoeuvre and then selects the
option maximising the difference between the
selectability and rejectability utilities. Plainly
speaking, each aircraft selects the best path with
respect to the minimum risk.
2.1 2-D ARCA Algorithm
Let us consider a single flight level. At each time
step, each UAV collects information from all other
aircraft within its communication range (viewable
aircraft). This information includes position, speed,
destination, actual heading, flight time (basically an
ADS-B frame). Each UAV may choose one of five
directions: flying straight, moderate or sharp turn to
the left, moderate or sharp turn to the right (-10°,-
5°,0°,5°,10°). For each UAV a priority set is defined
as that of all viewable aircraft with higher ranking
than itself that could be conflicting for some heading
choices (parents). The rank may be assigned using
different criteria, e.g. it can be based on the aircraft
accumulated delay, using the flight time information
in the ADS-B data.
The rejectability of agents is unconditioned: each
UAV matches the linear extension of each of its
directional options with the linear projections of
current headings of all aircraft in its priority set.
Each projected conflict adds a weight to the
rejectability function related to that directional
option, depending on distance in time and severity of
the conflict. Then a normalization is performed. This
increases the rejectability of flight options that lead
to conflicts (small separations), with more weight
for incidents closer in time. On the other hand the
selectability function reflects goal achievement. The
selectability is influenced by the preferences of other
agents, it is composed of two terms: the base
selectability (current UAV heading preferences) and
the parent selectability (higher priority agents
preferences). In non restricted airspace the
commercial aircraft have always higher priorities
than the UAVs for obvious safety purposes.
After computing the rejectability and the
selectability, each agent chooses its heading change,
maximizing the difference between selectability and
rejectability. Each satisficing agent looks for the
highest gain, with the lowest risk, taking also into
account preferences of other agents, thus obtaining a
solution that could be effective for the whole system.
2.2 3-D ARCA Algorithm
The three dimensional version of the algorithm takes
into account two additional choices for the flight
path: it is possible to climb or descend one flight
level (1000 ft). The altitude change choice is taken
through the continuous monitoring of the minimum
value for rejectability with respect to direction. If
this minimum is larger than a given threshold, it
implies that the current flight level is too crammed
ICINCO 2012 - 9th International Conference on Informatics in Control, Automation and Robotics
178