PreprintPDF Available
Preprints and early-stage research may not have been peer reviewed yet.


More widespread adoption requires swarms of robots to be more flexible for real-world applications. Multiple challenges remain in complex scenarios where a large amount of data needs to be processed in real-time and high degrees of situational awareness are required. The options in this direction are limited in existing robotic swarms, mostly homogeneous robots with limited operational and reconfiguration flexibility. We address this by bringing elastic computing techniques and dynamic resource management from the edge-cloud computing domain to the swarm robotics domain. This enables the dynamic provisioning of collective capabilities in the swarm for different applications. Therefore, we transform a swarm into a distributed sensing and computing platform capable of complex data processing tasks, which can then be offered as a service. In particular, we discuss how this can be applied to adaptive resource management in a heterogeneous swarm of drones, and how we are implementing the dynamic deployment of distributed data processing algorithms. With an elastic drone swarm built on reconfigurable hardware and containerized services, it will be possible to raise the self-awareness, degree of intelligence, and level of autonomy of heterogeneous swarms of robots. We describe novel directions for collaborative perception, and new ways of interacting with a robotic swarm.
End-to-End Design for Self-Reconfigurable
Heterogeneous Robotic Swarms
Jorge Pe˜
na Queralta1, Li Qingqing1, Tuan Nguyen Gia1, Hong-Linh Truong2, Tomi Westerlund1
1Turku Intelligent Embedded and Robotic Systems (TIERS) Lab, University of Turku, Turku, Finland
2Department of Computer Science, Aalto University, Finland
Emails: 1{jopequ, qingqli, tunggi, tovewe},
Abstract—More widespread adoption requires swarms of
robots to be more flexible for real-world applications. Multiple
challenges remain in complex scenarios where a large amount
of data needs to be processed in real-time and high degrees of
situational awareness are required. The options in this direction
are limited in existing robotic swarms, mostly homogeneous
robots with limited operational and reconfiguration flexibility.
We address this by bringing elastic computing techniques and
dynamic resource management from the edge-cloud computing
domain to the swarm robotics domain. This enables the dy-
namic provisioning of collective capabilities in the swarm for
different applications. Therefore, we transform a swarm into a
distributed sensing and computing platform capable of complex
data processing tasks, which can then be offered as a service.
In particular, we discuss how this can be applied to adaptive
resource management in a heterogeneous swarm of drones, and
how we are implementing the dynamic deployment of distributed
data processing algorithms. With an elastic drone swarm built
on reconfigurable hardware and containerized services, it will
be possible to raise the self-awareness, degree of intelligence,
and level of autonomy of heterogeneous swarms of robots. We
describe novel directions for collaborative perception, and new
ways of interacting with a robotic swarm.
Index Terms—Swarm Robotics; Edge Computing; Elasticity;
Dynamic Resource Management; Autonomous Robots
A recent trend in cyber-physical systems and the Internet
of Things (IoT) domain is to shift towards more distributed
computation, a trend that has crystallized through the edge
computing paradigm [1]. Similarly, recent advances in con-
tainerization, elastic computing, and dynamic resource man-
agement are materializing a decentralized cloud [2], [3]. Mul-
tiple researchers have explored the possibilities of integrating
Multi-Agent Systems (MAS) theory within the IoT and cloud
computing towards IoT MAS and Cloud MAS [4]. On the
other side, the combination of MAS and robotics has brought
swarms of robots with the potential for enhancing human
responses in safety-critical applications such as firefighting [5],
or post-disaster scenarios [6], among others.
We extend these two approaches (swarm robotics and
IoT MAS) towards the intersection of edge computing and
robotics, and argue that with appropriate management and
distribution of computing, sensing and communication re-
sources within a robotic swarm, higher degrees of intelligence
and operational flexibility and robustness can be achieved.
This concept is illustrated in Fig. 1. The capabilities of
robotic swarms are currently limited by different factors, from
the lack of methods and distributed collaborative sensing
Fig. 1: Intelligent and self-reconfigurable robot swarms can
be designed at the intersection of the multi-agent
systems (MAS), Robotics and Edge-Cloud Computing
algorithms [7], to static and inflexible resource management
with non-optimal utilization of hardware resources due to
separate hardware and software design. This includes em-
bedded hardware with relatively constrained computational
resources due to payload constraints and uniform resources in
swarms [8], [9], [10]. Some previous works have addressed
these limitations through computational offloading and the
definition of edge-cloud robotics architectures [11]. More
recently, heterogeneous robotic swarms have been proposed
to extend the flexibility and intelligence, opening the door to
a wider array of more complex application possibilities [7].
Nonetheless, multiple challenges remain in terms of managing
the collaboration within heterogeneous swarms [12].
A. Related Works
Multiple research efforts within the fields of multi-robot
systems and cloud computing have been directed towards
distributed task allocation and distributed load management.
For instance, autonomous mobile programs (AMPs) were an
early introduction of a dynamic computational load manage-
ment framework [13]. AMPs provided a distributed approach
where autonomous agents were able to make decisions on a
shared computational load, being aware of their own com-
putational capabilities. AMPs share similarities with early
load balancing techniques based and colony optimization
for cloud computing [14]. More recently, multi-agent load
balancing for resource allocation in a distributed computing
environment has been proposed [15]. From the point of view
of task allocation in multi-robot systems, K-means clustering
and auction based mechanisms were introduced in [16]. In
terms of spatial allocation, a workspace partitioning method
was presented in [17] for indoor environments. In our work,
we aim at combining these two approaches considering full
reconfigurability through tight integration of methods from
the edge computing domain and algorithms for cooperation
in multi-robot systems. This is, to the best of our knowledge,
the first paper presenting such an approach.
In our recent works, we have presented initial results to-
wards the definition of reconfigurable robotic swarms. In [18],
we define the concept of resource ensembles from the perspec-
tive of the edge computing domain. This concept serves as
the basis for abstracting edge resources and building dynamic
management models on top of them. From the point of
view of collaborative swarms of robots, we have presented
a blockchain-based approach in [12]. In this work, we uti-
lize a blockchain as a medium for achieving consensus for
bandwidth allocation and data quality ranking in a distributed
multi-robot system. This, again, serves as the starting point
towards distributed sensing and data processing in swarms of
robots, where sensing, network and computational resources
are abstracted and managed through collaborative decision
B. Contribution and Structure
The techniques we propose have a clear impact on swarms
of drones. Current solutions for drone swarms require opera-
tors to either manually control drones or perform analysis of
streamed data at a ground control center. This is a limitation
for the deployment of drones in large areas where there have
been natural disasters such as fires, or where people have gone
missing, as the human resources necessary are too large. Even
if the data is processed by a computer at the ground station,
the need for a high-bandwidth channel between drones and the
base station still limits significantly their operational capabil-
ities. Therefore, there is an evident need for more intelligent
drones that are able to perform data analysis independently
and autonomously navigate large areas. Reconfigurable drone
swarms empower complex edge data analysis at the swarm
level through distributed edge computing. At the same time,
this allows for long-term autonomous operation when energy
constraints allow, as well as reconnaissance in remote areas
with poor network connectivity.
The main objective of this paper is to introduce a new design
approach that enables efficient and dynamic resource man-
agement and autonomous reconfiguration of heterogeneous
robotic swarms (Sections II & III). We discuss the optimization
of the various computing resources and sensing capabilities
of robots in the swarm through a hardware (HW)/software
(SW) co-design approach for the development of specialized
robots. In particular, we incorporate elastic principles for
coordinating and engineering collective capabilities of multiple
heterogeneous robots. This elasticity takes into account the
coordination at the level of multi-agent systems, but also the
specific resources and algorithms utilized for robotic percep-
tion and navigation, among others.
Furthermore, we discuss how our proposed approach en-
ables the reconfiguration of drone swarms and realize a
distributed collective intelligence (Section IV). This requires
embedding intelligence and information processing in the
drones themselves. With current technology, deploying multi-
ple drones requires coordination among their operators, bind-
ing valuable resources from the actual mission. Even more,
when multiple drones are being deployed in parallel [19], [20].
However, neither isolated intelligent drones nor simple task
list orchestration is sufficient [21], [22]. Thus, it is essential
to establish a collective intelligence that enables autonomous
coordination and collaboration among the drones.
The remainder of this paper is organized as follows. Section
II describes the main concepts involved in swarm reconfigura-
bility, interfacing and control. In Section III, we introduce a
three-layered end-to-end design architecture for reconfigurable
robotic swarms. Section IV then delves into enabled technolo-
gies, focusing on our initial work towards a reconfigurable
drone swarm. In Section V, we discuss potential application
scenarios. Section VI concludes the work.
We address the aforementioned challenges through an end-
to-end design: from the robot hardware and local decision
making on the computational and sensing resources to provid-
ing the swarm and its capabilities as a service for end-users.
Two key aspects in our work are:
Reconfigurable hardware resources for flexible and
resource-rich computation platform at the swarm level
Distributed data processing at the edge for collective
swarm intelligence.
Reconfigurability and distributed intelligence provides more
computational resources for multi-modal sensor fusion and
distributed computation. Our model leverages methodologies
and engineering techniques for distributed and dynamic man-
agement of elastic resources for swarms improving quality of
results. The models are:
1) Swarm-as-a-Service Model: A swarm provides services
to end-users with a control interface or API through what we
call a Swarm as a Service (SwaaS). SwaaS offers an edge
service model for swarm applications. In this view, each robot
with its specialization is an edge services provider (or just
an edge provider). Edge provider’s sensing, computational
and external communication resources are considered edge
resources. Edge resources (hardware and software) are co-
designed to make robots richer in terms of the resources
that they can provide. With this approach, we bring various
` `
(a) A cloud robotics system.
` `
(b) A swarm robotics system.
(c) A reconfigurable robotic swarm.
Cloud computing server
Human command inputs
Mesh network
Sensing node
Computational node
Fig. 2: Illustration of different connectivity modalities for a
swarm of drones, illustrating sensing and computing
roles together with interaction modalities.
concepts of edge computing and services models to the field
of multi-robot systems.
2) Application-Specific Resource Ensembles Model: Ap-
plication Specific Resource Ensembles (ASREs) [18] define
a specific organized set of edge resources (ASRE template
or pattern) in swarms forming the edge infrastructure. An
essential part of ASREs is the coordination and monitoring
of resources together with swarm control and coordination.
Resource management techniques enable service discovery,
service end-to-end communication segmentation, and dis-
tributed task computation under unreliable and uncertain envi-
ronments by incorporating uncertainty and elasticity [23]. For
our knowledge, these have not been applied before to drone
swarms. For example, [24], [25] are dedicated for containers
and virtual machines.
In our end-to-end vision, the central point is to provide dy-
namic and flexible swarms as an elastic heterogeneous multi-
robot system. In the system individual robots have different
sensing and computational capabilities, a mesh network takes
care of intra-swarm communication, and distributed algorithms
enable the swarm to perform collective decision making as if
it were a single unit. By elastic, we mean that the different
resources of robots are abstracted and can be reconfigured
depending on the application needs [26]. An illustration of
this concept, compared to current cloud robotic and swarm
robotic systems, appears in Fig. 2. In Fig. 2a, each drone is
independently connected to a cloud server, where it offloads
part of its data processing. Any external control in this case
goes through the cloud application, but direct control of drones
could be enabled as well (for example, if the movement
of the drone is controlled via a radio controller, and then
mapping or other algorithms are run in the cloud). In Fig. 2b,
the drones form together a swarm. Each individual drone in
the swarm has the same role initially, and both sensing and
data processing occur individually at each drone. Algorithms
describing the collaboration between drones could then run
at the swarm level, but each drone would be still a separate
entity. Finally, our approach is illustrated in Fig. 2c, where
all drones form a swarm as well. The key difference is that
the swarm and its applications and resources are abstracted
from individual drones and defined in a distributed manner
at the swarm level. The communication with a controller or
cloud services occurs from the swarm as a whole, and not
from individual drones as separate entities. Moreover, sensing
and computational resources, and the corresponding roles,
are assigned dynamically among the swarm members. In the
example illustrated in Fig. 2, half of the drones take a sensing
role as their main role while data processing is offloaded to
other drones assigned as computing nodes.
We design an architecture with three layers and building
blocks as illustrated in Fig. 3. The layers are:
Physical Swarm Layer. The actual control of robots is
carried out at the physical layer, where the different hardware
and mesh communication solutions are defined. The definition
Sensing Resources
Hardware Resources
Intra-Swarm Communication
Robot control and dynamics
Local collision avoidance and
local path planning
Swarm Controller
decision making
Collaborative sensing
and sensor fusion
Spatial coordination
Multi-agent control
Regions of
Distributed Edge Services Layer
Application Layer
ASREs Templates
and Patterns Pool
Physical Swarm Layer
Resource Manager
Dynamic and Elastic
Resource Provisioning
ASREManagement Services
Fig. 3: Proposed Architecture and Building Blocks for SwaaS and ASREs
of a set of computing resources, sensors and actuators needs
to be carried out as part of the swarm design in order to
enable higher degrees of optimization when the sensing and
computational resources are shared at the swarm level and
applications run in a distributed manner.
Edge Services Layer. This is the main focus of the swarm
design in terms of self-reconfigurability. This layer includes
a Resource Manager (RM) for configuring and managing
resources provided by the Physical Swarm Layer. Based on
application requirements, the RM will assemble resources
from edge providers into ASREs for SwaaS. The RM supports
the automatic creation of ASREs by requesting, provisioning
and orchestrating suitable resources.
This layer represents all the distributed services and pro-
cesses running within the swarm and executed at each in-
dividual robot. These processes are classified in three main
types: (1) spatial coordination (e.g., distributed formation
control [27], [28]), (2) collaborative sensing (e.g. cooperative
mapping [29]), and (3) collaborative decision making (e.g.
). These three apparently different parts of swarm control
and decision making have a high synergy and their optimal
operation depends on feedback from each other. These three
topics have mostly been studied separately in the previous
works [30], [31], [32]. Therefore, we have focused on the
design and development of techniques for efficient communi-
cation between these processes. In summary, the key novelty
is that ASREs and the RM implicitly manage collaboration
within the swarm.
At runtime, ASREs will form an elastic and resilient edge
mesh of services across robots in a swarm. The RM will
dynamically provision new resources from different providers
elastically. This kind of elasticity is carried out in an end-to-
end and bi-directional manner: the resources are provisioned
dynamically when new services are required or when the
available resources change. The RM learns and optimizes the
provisioning based on the reliability of resources, performance
variations, bottlenecks, and failures. The RM provisioning is
hidden from the application.
Application Layer. The application layer provides an in-
terface for controlling and interacting with the swarm. We
refer to this interface as the Swaas API. An external party or
swarm controller can select from a pool of ASRE templates,
which define the different patterns in which the swarm can be
configured for different applications. By choosing an ASRE
template through the SwaaS API, the swarm controller is
implicitly selecting a set of resources and services. These
resources are then provisioned and managed within the swarm
itself through the RM. The services are provided based on the
available distributed algorithms for sensing and coordination.
In this section, we discuss the specific technologies that
enable the realization of the self-reconfigurable robotic swarm
architecture defined in the previous section. We are in the
process of applying these technologies to a swarm of drones.
A. Reconfigurability-enabling Technologies
Currently, most small mobile robots and aerial drones rely
mainly on CPUs in order to perform all the navigation and
mission related computation, and microcontrollers for low-
level control such as flight controllers [33]. We are leveraging
existing technologies from other domains to enable reconfig-
uration within a drone swarm.
As a computing platform, we utilize FPGAs with embedded
processors to extend the existing algorithms with custom
hardware accelerators. FPGAs are reconfigurable hardware
accelerators that can be exploited in computationally intensive
and highly parallelizable tasks for autonomous robots, with
higher performance/size and performance/power ratio as we
have shown in previous works [34], [35]. Both the size
and power consumption of hardware are essential aspects to
take into account in drones. The use of FPGAs enables the
RM to not only provision the existing resources but also
dynamically provision new hardware accelerators on-demand.
Some drones are equipped with FPGAs while others have
embedded processors with GPU such as the NVIDIA Jetson
At the software level, we utilize containers to enable dy-
namic resource management and task execution. Container
technologies are known but they have not been exploited for
drones. Our goal is to use containers to enable dynamic re-
sources management and task execution. All algorithms, from
spatial coordination to collaborative sensing, are containerized
and run in a distributed way. With efficient container orches-
tration, we are able to add flexibility and reconfigurability to
the swarm. We bring specific techniques for computational
load distribution, elasticity and resource management from the
edge-cloud domain to the robotics domain.
In order to interface sensors, actuators, communication
and the containerized algorithms, we utilize Robot Operating
System (ROS2) which runs as a container application as well.
ROS is the de-facto standard for robotic development in both
academia and industry. ROS2 focuses on distributed multi-
robot systems and real-time computing, and will allow us to
exploit container technologies for drones.
For network interfacing, there are no solutions that integrate
ROS2 and mesh networking so far. Our experiments will
utilize more traditional solutions at first, with all drones
connected to a single Wi-Fi access point. Nonetheless, we
will work towards the integration of ROS2 and a Bluetooth
5 mesh network. Another recent technology that can provide
significant advantages is ultra-wideband (UWB) [36]. UWB
enables accurate localization in multi-robot systems and, in
particular, in drones [37], and has the potential for simulta-
neous communication and localization. We will work on ex-
tending our current works on UWB-based mobile localization
systems [38], studying the integration of UWB as a network
interface between ROS2 nodes.
Finally, we leverage blockchain-powered consensus algo-
rithms suitable for multi-robot systems. Recent works [39],
[12], [40] have proposed design concepts for integrating next-
generation low-latency and scalable blockchains within het-
erogeneous multi-robot systems. Blockchains can be utilized
as a distributed framework to achieve consensus in a multi-
robot system, and also validate identities. This can be then
utilized by ASRE management services, which could, in turn,
be implemented as distributed Smart Contracts for resource
coordination. The idea of utilizing a blockchain-based frame-
work for managing edge resources has already been explored
in our previous works [41].
B. Edge computing algorithms for sensing
We classify the algorithms in the edge layer in three main
types: spatial coordination, collaborative sensing, and decision
making. In previous works, these are typically defined with
strong dependencies. For instance, depending on the sensing
variable robots might be required to be in a specific spatial
formation [42]. However, in our architecture these algorithms
are designed independently and abstracted as edge services.
This modular architecture brings multiple advantages. For
instance, spatial coordination algorithms take feedback from
the sensing algorithms regarding the location of regions of
interest that should be analyzed more closely. This feedback is
used to rearrange the drones in the proper shape and location.
At the same time, the role of each drone within the spatial
pattern is given by the collaborative decision making process.
Analogously, the spatial coordination algorithms give feedback
to other processes about the movement constrains of the swarm
C. Resource Management
The RM provisions and manages all resources, from
hardware to edge services. Sensing resources are abstracted
through ROS nodes (drivers) that produce data in standard
formats. Each edge service is broken down into sub-services
(for example, independent parts of an algorithm) and each
of these is abstracted as a ROS node that consumes and
produces different types of data (always in standard formats).
All these ROS nodes are containerized and provisioned by
the RM across the available the computing resources. The
computing resources are modeled based on the amount and
type of containers that they can run, and the performance for
each containers. The provisioning is an optimization process
that takes into account communication latency between data
producers and consumers, and execution latency. The RM
itself runs as a distributed and containerized application across
the swarm, and manages the resources with elastic techniques.
Each application that utilizes the SwaaS API must define a
Quality of Results (QoRs) requirement. We bring this concept
from elastic computing models in which a QoR is defined
in terms of performance, quality of data, type of output, and
other measurable information [43], which the swarm provides
to the controller through the SwaaS API. The QoR is essential
for elastic resource management to dynamically provision
the different resources taking the QoR requirement as an
optimization constraint. We will extend the work by Mariani et
al. on coordination-aware elasticity for developing primitives
and algorithms to control the elasticity of swarms [44].
Drones are already being utilized by fire departments [45],
police forces [46], and for creating ad-hoc networks in post-
disaster scenarios [47]. Fire and police departments are using
drones for monitoring and surveying emergency scenes and
disaster sites as well as finding missing people. Going beyond
perception-centric usage would be beneficial in all of these
use cases. This would require embedding intelligence and
information processing in the drones themselves, especially
when authorities are starting to deploy multiple drones in
parallel for the same applications. Deploying multiple drones
requires coordination and collaboration among their operators
binding valuable resources from the actual mission. Therefore,
it is essential to establish a collective intelligence that enables
autonomous coordination and collaboration among the drones
instead of passively receiving the tasks individually from a
centralized control system.
Another key application area for UAVs is real-time sur-
veying and monitoring. The same methods and technologies
presented in this paper can be applied in different tasks such as
environmental monitoring, land surveying or scientific study of
different parameters in remote locations. The main benefit of
utilizing reconfigurable drone swarms is the ability to perform
complex data analysis at the Swarm level through distributed
Edge computing. This allows for long-term autonomous oper-
ation when energy constraints allow, as well as reconnaissance
in remote areas with poor network connectivity.
We have proposed an architectural definition for reconfig-
urability in heterogeneous robotic swarms. This architecture is
based on a combination of concepts and techniques from the
robotics domain, multi-agent systems domain and edge-cloud
computing domain. This is, to the best of our knowledge, the
first work that proposes the abstraction and management of
both hardware (sensors, actuators, computation and commu-
nication) and software (distributed sensing, coordination and
decision making) as edge resources with elastic techniques. In
particular, we explain how we are designing a reconfigurable
drone swarm and what are the different hardware and software
that make reconfigurability and elasticity possible.
This work was partially supported by the Academy of
Finland’s AutoSOS project with grant number 328755.
[1] W. Shi, G. Pallis, and Z. Xu, “Edge computing [scanning the issue],”
Proceedings of the IEEE, vol. 107, no. 8, pp. 1474–1481, Aug 2019.
[2] A. J. Ferrer, J. M. Marques, and J. Jorba, “Towards the decentralised
cloud: Survey on approaches and challenges for mobile, ad hoc, and
edge computing,” ACM Comput. Surv., vol. 51, no. 6, pp. 111:1–111:36,
Jan. 2019. [Online]. Available:
[3] F. Jalali, O. J. Smith, T. Lynar, and F. Suits, “Cognitive iot
gateways: Automatic task sharing and switching between cloud and
edge/fog computing,” in Proceedings of the SIGCOMM Posters
and Demos, ser. SIGCOMM Posters and Demos ’17. New
York, NY, USA: ACM, 2017, pp. 121–123. [Online]. Available:
[4] D. Calvaresi, M. Marinoni, A. Sturm, M. Schumacher, and G. Buttazzo,
“The challenge of real-time multi-agent systems for enabling iot and
cps,” in Proceedings of the international conference on web intelligence.
ACM, 2017, pp. 356–364.
[5] J. Penders, L. Alboul, U. Witkowski, A. Naghsh, J. Saez-Pons, S. Her-
brechtsmeier, and M. El-Habbal, “A robot swarm assisting a human
fire-fighter,Advanced Robotics, vol. 25, no. 1-2, pp. 93–117, 2011.
[6] J. De Greeff, N. Smets, K. Hindriks, M. A. Neerincx, and I. Kruijff-
a, “Incremental development of large-scale human-robot team-
work in disaster response environments,” in 12th Annual ACM/IEEE In-
ternational Conference on Human-Robot Interaction, HRI 2017. ACM,
IEEE Computer Society, 2017, pp. 101–102.
[7] E. Masehian, M. Jannati, and T. Hekmatfar, “Cooperative mapping of
unknown environments by multiple heterogeneous mobile robots with
limited sensing,” Robotics and Autonomous Systems, vol. 87, pp. 188–
218, 2017.
[8] A. Kattepur, H. Dohare, V. Mushunuri, H. K. Rath, and A. Simha,
“Resource constrained offloading in fog computing,” in Proceedings of
the 1st Workshop on Middleware for Edge Clouds & Cloudlets. ACM,
2016, p. 1.
[9] Z. Hong, H. Huang, S. Guo, W. Chen, and Z. Zheng, “Qos-aware
cooperative computation offloading for robot swarms in cloud robotics,
IEEE Transactions on Vehicular Technology, vol. 68, no. 4, pp. 4027–
4041, April 2019.
[10] D. Palossi, F. Conti, and L. Benini, “An open source and open hardware
deep learning-powered visual navigation engine for autonomous nano-
uavs,” in 2019 15th International Conference on Distributed Computing
in Sensor Systems (DCOSS), 2019, pp. 604–611.
[11] L. Qingqing, J. Pe˜
na Queralta, T. N. Gia, Z. Zou, H. Tenhunen, and
T. Westerlund, “Visual Odometry Offloading in Internet of Vehicles with
Compression at the Edge of the Network,” in The 12th International
Conference on Mobile Computing and Ubiquitous Networking (ICMU),
[12] Jorge Pe˜
na Queralta and Tomi Westerlund, “Blockchain-powered
collaboration in heterogeneous swarms of robots,” arXiv preprint
arXiv:1912.01711, 2020, symposium on Blockchain for Robotic Sys-
tems, MIT Media Lab.
[13] X. Y. Deng, P. Trinder, and G. Michaelson, “Autonomous mobile pro-
grams,” in 2006 IEEE/WIC/ACM International Conference on Intelligent
Agent Technology. IEEE, 2006, pp. 177–186.
[14] Z. Zhang and X. Zhang, “A load balancing mechanism based on
ant colony and complex network theory in open cloud computing
federation,” in 2010 The 2nd International Conference on Industrial
Mechatronics and Automation, vol. 2. IEEE, 2010, pp. 240–243.
[15] S. Banerjee and J. P. Hecker, “A multi-agent system approach to load-
balancing and resource allocation for distributed computing,” in First
Complex Systems Digital Campus World E-Conference 2015. Springer,
2017, pp. 41–54.
[16] M. Elango, S. Nachiappan, and M. K. Tiwari, “Balancing task allocation
in multi-robot systems using k-means clustering and auction based
mechanisms,” Expert Systems with Applications, vol. 38, no. 6, pp.
6486–6491, 2011.
[17] A. Gautam, S. A. Ram, V. S. Shekhawat, and S. Mohan, “Balanced
partitioning of workspace for efficient multi-robot coordination,” in 2017
IEEE International Conference on Robotics and Biomimetics (ROBIO).
IEEE, 2017, pp. 104–109.
[18] H.-L. Truong, “Asre -towards application-specific resource ensembles
across edges and clouds,” 2020, preprint, working paper. [Online].
[19] J. Gregory, J. Fink, E. Stump, J. Twigg, J. Rogers, D. Baran, N. Fung,
and S. Young, “Application of multi-robot systems to disaster-relief
scenarios with limited communication,” in Field and Service Robotics.
Springer, 2016, pp. 639–653.
[20] M. Hu, W. Liu, K. Peng, X. Ma, W. Cheng, J. Liu, and B. Li, “Joint
routing and scheduling for vehicle-assisted multidrone surveillance,
IEEE Internet of Things Journal, vol. 6, no. 2, pp. 1781–1790, 2018.
[21] N. Mendes, M. Safeea, and P. Neto, “Flexible programming and or-
chestration of collaborative robotic manufacturing systems,” in 2018
IEEE 16th International Conference on Industrial Informatics (INDIN).
IEEE, 2018, pp. 913–918.
[22] V. Lomonaco, A. Trotta, M. Ziosi, J. d. D. Y. ´
Avila, and N. D´
ıguez, “Intelligent drone swarm for search and rescue operations at
sea,” arXiv preprint arXiv:1811.05291, 2018.
[23] H.-L. Truong, L. Berardinelli, I. Pavkovic, and G. Copil, “Modeling and
provisioning iot cloud systems for testing uncertainties,” in Proceedings
of the 14th EAI International Conference on Mobile and Ubiquitous
Systems: Computing, Networking and Services. ACM, 2017, pp. 96–
[24] V. Medel, O. Rana, J. A. Ba˜
nares, and U. Arronategui, “Modelling per-
formance amp; resource management in kubernetes,” in 2016 IEEE/ACM
9th International Conference on Utility and Cloud Computing (UCC),
Dec 2016, pp. 257–262.
[25] R.-A. Cherrueau, A. Lebre, D. Pertin, F. Wuhib, and J. M.
Soares, “Edge computing resource management system: a critical
building block! initiating the debate via openstack,” in USENIX
Workshop on Hot Topics in Edge Computing (HotEdge 18).
Boston, MA: USENIX Association, 2018. [Online]. Available:
[26] S. Dustdar, Y. Guo, B. Satzger, and H. L. Truong, “Principles of elastic
processes,” IEEE Internet Computing, vol. 15, no. 5, pp. 66–71, 2011.
[Online]. Available:
[27] J. Pe˜
na Queralta, C. McCord, T. N. Gia, H. Tenhunen, and T. Wester-
lund, “Communication-free and index-free distributed formation control
algorithm for multi-robot systems,” Procedia Computer Science, 2019,
the 10th ANT Conference.
[28] C. McCord, J. Pe˜
na Queralta, T. N. Gia, and T. Westerlund, “Distributed
Progressive Formation Control for Multi-Agent Systems: 2D and 3D
deployment of UAVs in ROS/Gazebo with RotorS,” in 2019 European
Conference on Mobile Robots (ECMR). IEEE, 2019.
[29] J. Pe˜
na Queralta, T. N. Gia, H. Tenhunen, and T. Westerlund, “Col-
laborative mapping with ioe-based heterogeneous vehicles for enhanced
situational awareness,” in IEEE Sensors Applications Symposium (SAS).
IEEE, 2019.
[30] K.-K. Oh, M.-C. Park, and H.-S. Ahn, “A survey of multi-agent
formation control,” Automatica, vol. 53, pp. 424–440, 2015.
[31] M. T. Nguyen, H. M. La, and K. A. Teague, “Collaborative and
compressed mobile sensing for data collection in distributed robotic
networks,” IEEE Transactions on Control of Network Systems, vol. 5,
no. 4, pp. 1729–1740, 2017.
[32] M. Mavrovouniotis, C. Li, and S. Yang, “A survey of swarm intelligence
for dynamic optimization: Algorithms and applications,” Swarm and
Evolutionary Computation, vol. 33, pp. 1–17, 2017.
[33] E. Kaufmann, M. Gehrig, P. Foehn, R. Ranftl, A. Dosovitskiy, V. Koltun,
and D. Scaramuzza, “Beauty and the beast: Optimal methods meet
learning for drone racing,” in 2019 International Conference on Robotics
and Automation (ICRA). IEEE, 2019, pp. 690–696.
[34] L. Qingqing, F. Yuhong, J. Pe ˜
na Queralta, T. N. Gia, Z. Zou, H. Ten-
hunen, and T. Westerlund, “Edge Computing for Mobile Robots: Multi-
Robot Feature-Based Lidar Odometry with FPGAs,” in The 12th Inter-
national Conference on Mobile Computing and Ubiquitous Networking
(ICMU), 2019.
[35] J. Pe˜
na Queralta, F. Yuhong, L. Salomaa, L. Qingqing, T. Gia, Z. Zou,
H. Tenhunen, and T. Westerlund, “Fpga-based architecture for a low-cost
3d lidar design and implementation from multiple rotating 2d lidars with
ros,” in 2019 IEEE SENSORS. IEEE, 2019, pp. 1–4.
[36] W. Shule, C. Mart´
ınez Almansa, J. Pe˜
na Queralta, Z. Zou, and T. West-
erlund, “Uwb-based localization for multi-uav systems and collabo-
rative heterogeneous multi-robot systems: a survey,arXiv preprint
arXiv:2004.08174, 2020.
[37] J. Pe˜
na Queralta, C. M. Almansa, F. Schiano, D. Floreano, and T. West-
erlund, “Uwb-based system for uav localization in gnss-denied environ-
ments: Characterization and dataset,” arXiv preprint arXiv:2003.04380,
[38] C. Mart´
ınez Almansa, W. Shule, J. Pe˜
na Queralta, and T. Wester-
lund, “Autocalibration of a mobile uwb localization system for ad-hoc
multi-robot deployments in gnss-denied environments,arXiv preprint
arXiv:2004.06762, 2020.
[39] V. Strobel, E. Castell´
o Ferrer, and M. Dorigo, “Managing byzantine
robots via blockchain technology in a swarm robotics collective decision
making scenario,” in Proceedings of the 17th International Conference
on Autonomous Agents and MultiAgent Systems. International Founda-
tion for Autonomous Agents and Multiagent Systems, 2018, pp. 541–
[40] E. C. Ferrer, “The blockchain: a new framework for robotic swarm sys-
tems,” in Proceedings of the Future Technologies Conference. Springer,
2018, pp. 1037–1058.
[41] J. Pe ˜
na Queralta, L. Qingqing, Z. Zou, T. Westerlund, “Enhancing
autonomy with blockchain and multi-acess edge computing in distributed
robotic systems,” in The Fifth International Conference on Fog and
Mobile Edge Computing (FMEC). IEEE, 2020.
[42] J. Pe˜
na Queralta, L. Qingqing, T. N. Gia, H. Tenhunen, and T. Wester-
lund, “Distributed progressive formation control with one-way commu-
nication for multi-agent systems,” in 2019 IEEE Symposium Series on
Computational Intelligence, 2019.
[43] H. L. Truong and S. Dustdar, “Principles of software-defined elastic
systems for big data analytics,” in Proceedings of the 2014 IEEE
International Conference on Cloud Engineering, ser. IC2E ’14.
Washington, DC, USA: IEEE Computer Society, 2014, pp. 562–567.
[Online]. Available:
[44] S. Mariani, H.-L. Truong, G. Copil, A. Omicini, and S. Dustdar,
“Coordination-aware elasticity,” in Proceedings of the 2014 IEEE/ACM
7th International Conference on Utility and Cloud Computing, ser.
UCC ’14. Washington, DC, USA: IEEE Computer Society, 2014, pp.
465–472. [Online]. Available:
[45] W. F. Price et al., “Public safety drones,” 2019.
[46] M. W. Miller et al., “The use of small unmanned aircraft systems (suas)
in law enforcement,” 2016.
[47] K. G. Panda, S. Das, D. Sen, and W. Arif, “Design and deployment of
uav-aided post-disaster emergency network,IEEE Access, vol. 7, pp.
102 985–102 999, 2019.
ResearchGate has not been able to resolve any citations for this publication.
Conference Paper
Full-text available
This conceptual paper discusses how different aspects involving the autonomous operation of robots and vehicles will change when they have access to next-generation mobile networks. 5G and beyond connectivity is bringing together a myriad of technologies and industries under its umbrella. High-bandwidth, low-latency edge computing services through network slicing have the potential to support novel application scenarios in different domains including robotics, autonomous vehicles, and the Internet of Things. In particular, multi-tenant applications at the edge of the network will boost the development of autonomous robots and vehicles offering computational resources and intelligence through reliable offloading services. The integration of more distributed network architectures with distributed robotic systems can increase the degree of intelligence and level of autonomy of connected units. We argue that the last piece to put together a services framework with third-party integration will be next-generation low-latency blockchain networks. Blockchains will enable a transparent and secure way of providing services and managing resources at the Multi-Access Edge Computing (MEC) layer. We overview the state-of-the-art in MEC slicing, distributed robotic systems and blockchain technology to define a framework for services the MEC layer that will enhance the autonomous operations of connected robots and vehicles.
Conference Paper
Full-text available
A recent trend in the IoT is to shift from traditional cloud-centric applications towards more distributed approaches embracing the fog and edge computing paradigms. In autonomous robots and vehicles, much research has been put into the potential of offloading computationally intensive tasks to cloud computing. Visual odometry is a common example, as real-time analysis of one or multiple video feeds requires significant on-board computation. If this operations are offloaded, then the on-board hardware can be simplified, and the battery life extended. In the case of self-driving cars, efficient offloading can significantly decrease the price of the hardware. Nonetheless, offloading to cloud computing compromises the system's latency and poses serious reliability issues. Visual odometry offloading requires streaming of video-feeds in real-time. In a multi-vehicle scenario, enabling efficient data compression without compromising performance can help save bandwidth and increase reliability.
Conference Paper
Full-text available
The cooperation of multiple robots towards a common goal requires a certain spatial distribution, or formation configuration, of the agents in order to succeed. Centralized controllers that have information about the absolute or relative positions of all agents, or distributed approaches using communication to share system-wide information between agents, are able to calculate optimal individual paths. However, this reserves important computational resources as the number of agents in the system increases. We address the problem of distributed formation control with minimal communication and minimal computational power required. The algorithm introduced in this paper progressively generates a directed path graph to uniquely assign formation positions to all agents. The benefits of the proposed algorithm compared to previous include the need for one-way communication only, low computational complexity, ability to converge without any a priori assignments under certain geometric conditions and need for limited sensing information only. The algorithm can be deployed to computationally constrained devices, enabling its deployment in robots with simpler hardware architectures. The ability to converge and distributively assign positions, or roles, with only one-way communication makes this algorithm robust during deployment, at which time all agents are equivalent and anonymous. Moreover, we account for limited communication and sensing range, and agents only need to have information about other agents in their vicinity in order to make decisions. Communication is simple and allows for scalability without an impact on performance or convergence latency, with a linear dependence on the number of agents. Agents only need to broadcast their status to other neighboring agents and do not reply any message. Finally, this algorithm enables almost-arbitrary configurations. The main limitation for the choice of formation configuration is that all pairs of points forming an edge in the polygonal line defining the boundary of the convex hull must be within sensing range.
Conference Paper
Full-text available
Offloading computationally intensive tasks such as lidar or visual odometry from mobile robots has multiple benefits. Resource constrained robots can make use of their network capabilities to reduce the data processing load and be able to perform a larger number tasks in a more efficient manner. However, previous works have mostly focused on cloud offloading, which increases latency and reduces reliability, or high-end edge devices. Instead, we explore the utilization of FPGAs at the edge for computational offloading with minimal latency and high parallelism. We present the potential for modelling feature-based odometry in VHDL and utilizing FPGA implementations.
Conference Paper
Full-text available
Autonomous micro aerial vehicles still struggle with fast and agile maneuvers, dynamic environments, imperfect sensing, and state estimation drift. Autonomous drone racing brings these challenges to the fore. Human pilots can fly a previously unseen track after a handful of practice runs. In contrast, state-of-the-art autonomous navigation algorithms require either a precise metric map of the environment or a large amount of training data collected in the track of interest. To bridge this gap, we propose an approach that can fly a new track in a previously unseen environment without a precise map or expensive data collection. Our approach represents the global track layout with coarse gate locations, which can be easily estimated from a single demonstration flight. At test time, a convolutional network predicts the poses of the closest gates along with their uncertainty. These predictions are incorporated by an extended Kalman filter to maintain optimal maximum-a-posteriori estimates of gate locations. This allows the framework to cope with misleading high-variance estimates that could stem from poor observability or lack of visible gates. Given the estimated gate poses, we use model predictive control to quickly and accurately navigate through the track. We conduct extensive experiments in the physical world, demonstrating agile and robust flight through complex and diverse previously-unseen race tracks. The presented approach was used to win the IROS 2018 Autonomous Drone Race Competition, outracing the second-placing team by a factor of two.
Conference Paper
Full-text available
Coordination of multiple robots in order to cooperatively perform a given task requires a certain distribution of the different units in space. Furthermore, individual robots, or agents, might have different tasks, or positions, assigned. Formation control algorithms might rely on a priori information, a centralized controller, or communication among the agents to assign roles. Distributed approaches that only need local interaction between agents have limited possibilities, such as flocks where agents actively control the distance to neighboring agents. Alternatively, two-way local communication has been applied to progressively assign roles and converge towards a given configuration. We propose a progressive assignment algorithm and formation control scheme that extends leader-follower formations in order to enable cooperation of multiple robots with minimal, one-way, local communication between agents. The proposed algorithm progressively generates a directed, locally convex, path graph to uniquely assign formation positions to all agents. The low computational complexity of our algorithm enables its implementation in resource-limited devices. Agents only require information about neighboring agents and be able to locally broadcast their status. This algorithm enables almost arbitrary two-dimensional configurations, with the only limitation being the sensing range enabling the definition of a series of convex hulls in a certain subset of agents such that agents sharing an edge in the hull are able to sense each other. Moreover, we propose a methodology for deploying agents to an arbitrary three-dimensional configuration after the assignment process is made on the plane.
Full-text available
Designing a reliable, resilient and quickly deployable emergency communication network is a key challenge for post-disaster management. In this paper, a UAV-assisted emergency Wi-Fi network is proposed to expedite the rescue operations by guiding the survivors to the nearest rescue camp location. Here, Raspberry PI (RPI) development board, mounted on UAV is considered to form a Wi-Fi chain network over the disaster region. During network set-up, the proposed solutions for the design challenges like UAV synchronization, avoid communication disruption and surveillance data management are the key contributions of this paper. The designed UAV network is capable of doing on-site surveillance and transmitting the data to the relief center for better rescue planning. One major challenge is to alert a survivor about the emergency network, which is addressed by designing a captive portal. Furthermore, to extend the Wi-Fi network, an Android-based application is developed by which each smartphone acts as a relay for its neighbor. Three types of field experiment are carried out to evaluate the performance of the designed prototype. It is found from the field results; Wi-Fi access point mode and user datagram protocol are more suitable for network design as compared to Ad-Hoc mode and transmission control protocol, respectively. It is also observed from the experiment that the maximum hop distance for the prototype is 280 meters and 290 meters for a Wi-Fi configuration following IEEE 802.11n and IEEE 802.11ac protocol, respectively.