Aerospace and Electronic Systems - March 2019 - 42

ARIES: An Autonomous Controller For Multirobot Cooperation
a multiple robot team cooperates to perform missions
based on independent tasks. Each robot can sense with
some probability the effect of their own actions and the
actions of other team members through perception and
explicit broadcast communications. Then, each robot
selects its own tasks, considering that those tasks are
aligned for the benefit of the whole team. ALLIANCE
allows the robotic team to be fault tolerant, reliable, and
adaptable through the use of motivational behaviors.
These behaviors are designed to allow robot team members to perform tasks only as long as the task demonstrated a significant advance toward the team goal. Our
approach does not implement motivational behaviors,
instead we model simple behaviors to design our own
policies that aim to prevent robot failures from current
states, i.e., a leader failure does not have to trigger a follower failure or vice versa. As well, we model a fault tolerance domain on each robotic agent to prevent the
propagation of failure states. Other example is the Mission Oriented Operating Suite (MOOS) architecture [10],
which describes a behavior based and distributed controller to coordinate multiple AUVs for oceanic exploration
by using interval programming and solving multiobjective optimizations problems where the objective is
to minimize the path cost of each AUV. Our approach
follows a similar module-based philosophy to provide
scalability and flexibility for further developments, but it
does not implement interval programming neither multiobjective optimization for the path planning.
Other architectures distribute the mission goals of the
multiple robot team to follow a coordinated and robust execution flow. For instance, we can mention the M+ architecture [11], which distributes the mission goals in a
hierarchical manner where the low-level goals are locally
managed by each robot and high-level goals are managed
by a central node. M+ represents an extension of the LAAS
architecture [2] for a multiple robot frame. Our approach
follows a similar hierarchical approach (leader-follower
paradigm) where the deliberation process is centralized in
the leader, i.e., the leader is responsible for the planning of
cooperative high-level goals. As in M+, the low-level goals
aim to solve opportunistic robot failures. An architecture
that follows a similar approach is the High-level Distributed DecisioN (HiDDeN) [12], which is a distributed deliberative architecture that manages the execution of a
hierarchical plan for a multiple robot team in a specific air-
sea scenario. The hierarchical plan is distributed for its execution among the supervisor of the robots. Each robot has
its own supervisor, which contains the robots actions and
synchronization tasks with the others. Besides, the supervisors include a hierarchical repair process to provide fault
tolerance to the system. Both hierarchical plan and repair
processes are instantiated as hierarchical task networks
(HTN) [13], a common planning approach in cooperative
architectures. Our approach does not implement HTN for
42

the goals distribution. Our approach uses the planning
domain definition language (PDDL) [14] to model the
behavior of the multiple robot team following a coordination among their actions.
In practice, many architectures do not conform to a
strict paradigm dichotomy, e.g., many behavior-based
architectures employs hierarchical distribution of goals
and vice versa. In this way, many architectures has been
designed as hybrid systems which integrate several features in order to overcome complex problems. As a hybrid
paradigm, we want to highlight the leader-follower
approach, which has been extensively studied in the literature. The work of Chaimowicz et al. [15] proposes a
tightly coupled architecture where the communication
protocols allow the roles distribution and the coordination
among the robots. Also, it provides a mechanism for
dynamical distribution of roles. Parker et al. [16]
presented a cooperative architecture for the navigation
assistance task. Here, the leaders are sensor-rich robots
modeled to assist in the navigation of sensor-limited
robots (followers) which do not have on-board capabilities
for obstacle avoidance or localization. Our architecture
uses the hybrid leader-follower paradigm to share functionalities among the robotic systems in order to perform
complex missions. It does not provide dynamic roles
assignment but it can be easily adapted to any robotic system. CoT-ReX [17] is a hybrid architecture adapted to the
problem of underwater detection and localization. It uses
AUVs to gather information about the targets locations
while the Autonomous Surface Vehicle (ASV) acts as a
communication hub among all the AUVs. CoT-ReX has
been built under the T-REX system for the planning and
execution control of the mission. Nevertheless, CoT-ReX
keeps a centralized approach by executing a single T-REX
agent for the whole architecture, while our approach goes
further by providing a distributed execution in which each
robotic system executes its own T-REX agent.

A T-REX OVERVIEW
The purpose of this section is to introduce the T-REX
architecture [18], [19] to the reader. T-REX is a goaloriented system that follows the timeline-based planning
paradigm [20]-[22]. As Mayer et al. [23] describe, the
timeline-based planning paradigm aims to control complex physical systems through the synthesis of desired
temporal behaviors (called timelines) over robotic features
with associated temporal functions. In timeline-based
planning, these time-varying robotic features are called
state variables, e.g., the navigation system. Also, the values that a state variable can take over time are called as
tokens. Then, a timeline referred to a state variable is a
sequence of tokens that defines its temporal evolution,
such as the navigation control of a rover reporting its

IEEE A&E SYSTEMS MAGAZINE

MARCH 2019

Aerospace and Electronic Systems - March 2019

Table of Contents for the Digital Edition of Aerospace and Electronic Systems - March 2019