IEEE Robotics & Automation Magazine - June 2021 - 75

The efficient supervision and coordination of a heterogeneous
system mandates a decentralized framework that integrates
high-level task planning, low-level motion planning
and control, and robust real-time sensing of the robot's
dynamic environment. Decentralization in multiagent robotic
systems is of utmost importance because it provides flexibility,
scalability, and fault-tolerance capabilities. In this work, we
present the architecture of the decentralized framework
developed within the context of the European Union project
Co4Robots and its application in a multitasking collaboration
scenario involving various heterogeneous robots and humans.
Common Applications
Multiple robots are commonly used in cooperative applications,
such as exploration, surveillance, service robotics, and
cognitive factories [1]. However, dealing with the collaboration
of heterogeneous multirobot systems is a rather tricky
undertaking due to the different kinematic and sensing capabilities
of each robot. One issue of utmost importance in
coordination of robotic teams is multiagent task planning and
control. Significant efforts have been devoted toward this in
the past decades, resulting in a number of high-complexity
algorithms [2].
A standard classification that arises in multiagent planning
and control is centralized versus decentralized
schemes, depending on whether the assignment of actions
to the agents is performed by a central computer unit or
locally by each agent. Current practice in coordination of
robotic teams is based on offline centralized planning, and
related tasks are almost exclusively fulfilled in a predefined
manner, allowing little room for real-time and coordinated
decentralized actions. Centralized planning schemes with
global and local tasks usually provided satisfactory results
[3]; however, they have been proven to be computationally
expensive. On the other hand, decentralized planning significantly
reduces computational complexity [4]. A rather
important issue in heterogeneous decentralized task planning
is role assignment and task allocation, where each
agent's capabilities diverge depending on its teammate and/
or their mutual state [5].
From the control point of view, multirobot cooperative
object manipulation and transportation have been well studied
in the literature, especially in a centralized framework [6].
Despite its performance, centralized control is less robust since
all units rely on a central system, and its complexity increases
rapidly as the number of participating robots increases. On the
other hand, decentralized control approaches usually depend
on heavy, explicit interrobot communication and global offline
knowledge of the desired task [7]. Nevertheless, in such tasks,
implicit interrobot communication arises naturally as a side
effect of the robot's physical interactions (e.g., the interaction
forces between the object and the robot), which can be easily
acquired by appropriate sensors attached to the robot [8].
However, limited studies have been conducted in cooperative
object manipulation and transportation via heterogeneous
robotic systems [9].
Hence, our work is motivated by the need to have multirobot
decentralized systems, where a significant amount of
information can be implicitly acquired via physical interactions
and processed locally, reducing the need for exhaustive,
explicit interrobot communication.
More specifically,
we present a
complete decentralized
framework consisting
of 1) a set of perceptual
algorithms that enable
cooperating robots to
estimate the state of their
highly dynamic environment,
2) a set of control
schemes appropriate for
the mobility and manipulation
capabilities of the
considered robotic platforms,
3) a systematic
real-time decentralized methodology to accomplish complex
mission specifications given to a team of heterogeneous
robots, and 4) the corresponding systematic integration of
these modalities at both conceptual and software implementation
levels. The efficacy of the overall framework is demonstrated
via a complex scenario, which involves three
heterogeneous robots and humans cooperating in loading
and transportation tasks.
System Components
Our purpose is to develop a decentralized framework that will
be able to support logistic tasks in an automated manner by
efficiently allocating a set of heterogeneous robotic agents as
well as humans that collaborate appropriately according to the
specifications of each task. Hence, we envision the employment
of 1) a dexterous seven-degrees-of-freedom (DoF) static
manipulator able to perform loading and unloading actions
of light and heavy objects, 2) a mobile manipulator to extend
the motion flexibility and reachability of the overall framework,
and 3) a mobile robot for the transportation of objects
across different areas of the workspace, endowed with the
ability to ease the loading and unloading procedures by
Decentralization in
multiagent robotic systems
is of utmost importance
because it provides
flexibility, scalability, and
fault-tolerance capabilities.
Figure 1. Heterogeneous agents cooperating in a multitasking
collaboration scenario.
JUNE 2021 * IEEE ROBOTICS & AUTOMATION MAGAZINE *
75
IEEE Robotics & Automation Magazine - June 2021

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - June 2021
