Aerospace and Electronic Systems - March 2019 - 47

Ropero et al.

Figure 7.
T-REX agents in ARIES following the leader-follower approach.

Currently, the functional support is directly integrated in
the T-REX agents. However, it is possible to easily extend
the T-REX agents to use different technologies for the
functional layer such as GenOM [32] or ROS [33], for
instance.
Currently, the ARIES fault-tolerant system is able to
control specific nonnominal states. On one hand, the
Leader Agent can manage unachieved goal(s) and communication failures. For unachieved goals, the Leader
follows a two steps procedure: first, the executive layer
isolates the event and tries to execute a reactive rule that
may solve the issue; if the first step fails, the second step
followed by the executive layer is to forward the issue to
the deliberative layer. This last implies a replanning
and tasks reassignment to the Followers. In such case,
the Followers replace their previous plan with the new
plan given by the Leader. For the communication failure,
the Leader waits for any heartbeat signal from any Follower. If none Follower sends a heartbeat signal, the
Leader performs replanning and tries to achieve the mission goal on its own. On the other hand, the Follower
Agent can manage the unachieved goal and communication failure in a different way. For the unachieved goal
state, the cooperative layer of the Follower reports the
error to the deliberative layer of the Leader, asking for a
new plan considering the Follower(s) current states. For
the communication failure, the cooperative layer of the
Follower reports the error to its deliberative layer, which
deals with the failure from a selfish attitude without considering the other Follower(s) states. This is an initial
approach and it is not a complete failure management.
Building a complete failure management system is an
on-going research work.

LEADER AGENT
The Leader Agent is the T-REX agent that provides the
cooperative planning capabilities to an ARIES instance.
MARCH 2019

It is made up of three T-REX reactors, each one placed in
a different layer as shown in Figure 7(a): the Planner
Reactor in the deliberative layer, the COOP Reactor in the
cooperative layer, and the Generic Executive Reactor
(GER) [34] in the executive layer. In the following, we
describe the functional scope of each layer.
At the top level, the deliberative layer is formed by the
Planner Reactor, a PDDL-based reactor which uses two
input files to represent the world knowledge: the domain
and the problem file (see Figure 8). The domain file
contains the description the possible state variables that
can be achieved by every robotic system in the team as
shown in Section "FROM TEMPORAL ACTION-BASED
TO TIMELINE-BASED PLANNING." The interaction
among the state variables of different robotic systems into
the domain file defines the cooperation model of the robot
team. The problem file includes the initial facts defining
the starting tokens of every state variable for each robotic
system in the team and, also, the high-level goals we want
to achieve. This information is used by a PDDL planner to
find a sequence of tokens which allows the multiple robot
team to reach the goal states from their initial states. Also,
the Planner Reactor declares every state variable of the
Leader and every Follower as external. The externals of the
Leader are used by the Planner Reactor to publish goals to
the GER (B.1 in Figure 8) as we explained in Section "A TREX OVERVIEW."
We have integrated the Unified Path Planning and
Task Planning Architecture (UP2TA) [35] as the PDDL
planner of ARIES. UP2TA is a planner that interleaves
path-planning and task-planning for mobile robotics applications. Nevertheless, ARIES can be easily adapted to different PDDL planners, such as Fast Forward (FF) [36] or
Subgoal Partitioning Planning [37], thanks to the use of an
standardized planning language and the support provided
by the PDDL-lib.
The Planner Reactor, as a deliberative reactor, has a
certain latency > 0 and look-ahead p > 1 within the
timing model of the T-REX engine (see Section "A

IEEE A&E SYSTEMS MAGAZINE

Aerospace and Electronic Systems - March 2019

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