Conference PaperPDF Available

Offloading SLAM for Indoor Mobile Robots with Edge-Fog-Cloud Computing


Abstract and Figures

Indoor mobile robots are widely used in industrial environments such as large logistic warehouses. They are often in charge of collecting or sorting products. For such robots, computation-intensive operations account for a significant percentage of the total energy consumption and consequently affect battery life. Besides, in order to keep both the power consumption and hardware complexity low, simple micro-controllers or single-board computers are used as onboard local control units. This limits the computational capabilities of robots and consequently their performance. Offloading heavy computation to Cloud servers has been a widely used approach to solve this problem for cases where large amounts of sensor data such as real-time video feeds need to be analyzed. More recently, Fog and Edge computing are being leveraged for offloading tasks such as image processing and complex navigation algorithms involving non-linear mathematical operations. In this paper, we present a system architecture for offloading computationally expensive localization and mapping tasks to smart Edge gateways which use Fog services. We show how Edge computing brings computational capabilities of the Cloud to the robot environment without compromising operational reliability due to connection issues. Furthermore, we analyze the power consumption of a prototype robot vehicle in different modes and show how battery life can be significantly improved by moving the processing of data to the Edge layer.
Content may be subject to copyright.
Offloading SLAM for Indoor Mobile Robots
with Edge-Fog-Cloud Computing
V. K. Sarker1, J. Pe˜
na Queralta1, T. N. Gia1, H. Tenhunen2and T. Westerlund1
1Department of Future Technologies, University of Turku, Turku, Finland
2Department of Electronics, KTH Royal Institute of Technology, Stockholm, Sweden
Email: 1{vikasar, jopequ, tunggi, tovewe},
Abstract—Indoor mobile robots are widely used in industrial
environments such as large logistic warehouses. They are often
in charge of collecting or sorting products. For such robots,
computation-intensive operations account for a significant per-
centage of the total energy consumption and consequently affect
battery life. Besides, in order to keep both the power con-
sumption and hardware complexity low, simple micro-controllers
or single-board computers are used as onboard local control
units. This limits the computational capabilities of robots and
consequently their performance. Offloading heavy computation
to Cloud servers has been a widely used approach to solve this
problem for cases where large amounts of sensor data such as
real-time video feeds need to be analyzed. More recently, Fog
and Edge computing are being leveraged for offloading tasks
such as image processing and complex navigation algorithms
involving non-linear mathematical operations. In this paper, we
present a system architecture for offloading computationally
expensive localization and mapping tasks to smart Edge gateways
which use Fog services. We show how Edge computing brings
computational capabilities of the Cloud to the robot environment
without compromising operational reliability due to connection
issues. Furthermore, we analyze the power consumption of a
prototype robot vehicle in different modes and show how battery
life can be significantly improved by moving the processing of
data to the Edge layer.
Index Terms—Edge, Fog, Cloud, SLAM, Efficiency, Computa-
tion, Offloading, Energy, Performance, Mobile, Robots
Small indoor robots which work alongside humans have
become ubiquitous in different fields such as home-care,
restaurant business and manufacturing plants. For instance,
lightweight robots are used in exhibitions for guiding visitors
to a desired stall, and are the operational backbone in logistic
warehouses of the world’s top e-commerce companies. In
smart homes, small robots can be used as a pet which can
communicate, watch over or play with children and elderly
people and can notify or trigger alarm when abnormal situation
occurs [1]. Companion robots can remind elderly people to
take medicine on time. However, small robots have limited
resources such as low battery capacity and light processing
capability. In most cases, these robots cannot operate for a long
period of time or cannot make useful decisions in complex
situations due to these limitations.
Internet of Things (IoT) can be defined as a virtual platform
which allows both physical and virtual objects to be inter-
connected and communicate with each other. It consists of
advanced technologies such as sensor fusion, wireless sensor
networks and Cloud computing to help improve the quality of
services [2]–[6]. Besides, global storage allows users to access
data anywhere at any time [7]. Proper utilization of IoT can
help overcome some of the drawbacks of small robots. Partic-
ularly, real-time position of a small robot can be tracked and a
system administrator can send control commands in cases of
complex situations [8]. However, IoT-assisted small robots still
have some disadvantages. For example, traditional gateways in
IoT platform are mainly responsible for collecting data from
robots and forwarding the data to Cloud servers. When the data
volume is tremendous e.g., due to a large number of multi-
sensor equipped robots operating simultaneously, the network
bandwidth may be not adequate or the communication latency
become too high [9]. Hence, an appropriate architecture is
required which can overcome these limitations.
Edge and Fog computing can be described as extra layers
between sensor devices, gateways and Cloud servers for en-
hancing the quality of services. These bring Cloud computing
paradigms to the Edge of the network and help reduce the bur-
den of Cloud providing faster services unsupported by Cloud
computing. Edge and Fog computing help to reduce energy
consumption of sensor nodes and diminish overall latency
[10], [11]. The combination of Edge and Fog computing with
IoT can provide a suitable approach for enhancing overall
In this paper, we exploit the concept of IoT and Fog
along with Edge gateways to enhance energy efficiency and
operational performance of small indoor robots. We present
a proof of concept for a hybrid Edge-Fog-Cloud architecture
which facilitates use of Fog/Edge and Cloud computing for
offloading computationally expensive tasks from robots with
limited processing power. We discuss related works on how
intensive tasks can be offloaded to Edge, Fog or Cloud
services in Section II. Section III presents the proposed system
architecture applying Edge and Fog computing. Section IV
illustrates the results of different tests and discusses the advan-
tages in terms of power saving and performance boost. Section
V concludes the work and depicts on future improvements.
Several works focus on increasing the performance of dif-
ferent algorithms which cannot be fully exploited in resource-
constrained devices. Dey et al. analyzed the advantages of
leveraging Fog and Edge computing to offload computationally
heavy tasks within SLAM algorithms for mobile robots [12].
They provided extensive simulations for different Robot-Edge-
Fog-Cloud computational tasks distribution schemes. The au-
thors concluded that Edge and Fog computing can bring signif-
icant enhancement to computationally intensive tasks such as
localization and positioning tasks. Edge and Fog computing
are shown to provide the same computational performance
than Cloud offloading, with the benefit of reduced bandwidth
and thus faster response time. However, the authors did not
perform real-life tests where the communication layer can play
an important role and become a bottleneck due to the high data
rates between mobile robots and gateways.
Offloading Simultaneous Localization and Mapping
(SLAM) tasks to Edge and Fog layers has only been
considered recently while Cloud-based SLAM has been
studied for the last decade. Benavidez et al. deployed an
instance of the Robot Operating System in the Cloud to
increase the computational capabilities of constraint robots
for Visual SLAM (VSLAM) [13]. They pointed how this
helps to overcome traditional bottlenecks in robots with
limited computational resources with VSLAM performing
feature identification and matching which usually need
large databases. They demonstrated how parallel computing
can be exploited with a multi-node Cloud deployment to
facilitate multi-robot SLAM and with limited impact on
the on-board computers that control the robots. A similar,
more comprehensive work was carried out by Riazuelo et al.
who presented C2TAM, a Cloud framework for cooperative
tracking and mapping [14]. The authors used the unlimited
computational resources available in the Cloud to enable the
implementation of algorithms which fail to run on single
board computers or even more powerful CPUs when the
number of agents increases.
Similarly, Turnbull et al. utilized Cloud servers to enable ad-
vanced neural command for formation control in large groups
of robots [15]. However, the network connection between
robots and Cloud servers is unreliable and cannot be guaran-
teed. Therefore, a trade-off exists between the robustness of the
solution in terms of available computational resources, opera-
tional reliability and performance. This is further explored by
Salmeron-Garcia et al. in their work on Cloud offloading of a
vision-based navigation system [16]. The authors stress in their
paper that processing time can be matched to communication
period with parallel computation in the Cloud and inferred
of a bottleneck due to communication bandwidth and latency.
They concluded that Cloud is necessary if the system needs
to be easily scalable or precise operation is required. This
occurs where the localization or navigation requirements are
such that high quality images need to be processed. Extensive
works exist in this direction by other authors in the field
[17] [18], taking the trade-offs into account both in terms of
communication and computational performance.
Kumar et al. gathered computational offloading techniques
and algorithms into a survey where they accentuate how
shifting of intensive applications will be essential in the future
for battery powered devices [19] and explored the energy-
saving potential of offloading computation to the Cloud.
Extensive work has been carried out on leveraging Cloud
computing for offloading heavy computational tasks. However,
most of this work have been focused on leveraging virtually
unlimited computational resources of the Cloud. Fewer re-
search have targeted the Edge and Fog computing paradigms
as a solution to the high latency and unreliability of Cloud
services. Keeping these in mind, we focus in both energy
efficiency and performance improvement by placing the most
expensive computational tasks in Edge and Fog layers leaving
the Cloud only for monitoring and management and thus not
entirely relying on network link to the Cloud services for
normal operation.
The proposed system architecture consists of 4 layers: Robot
Layer,Edge Layer for data processing and analysis, Fog layer
for distributed storage and Cloud Layer for monitoring and
general mission control. As shown in Fig. 1, we have com-
bined the Edge and Fog layer because the Fog layer consists
of the smart Edge Gateways and they use the same physical
resources. The architecture we present is easily scalable,
modular and distributed by definition. This section thoroughly
describes the role of each layer and the distribution of the
computational load through the network. Energy consumption,
latency and computational power at different layers are the key
factors taken into account when deciding the role of a layer.
In general, data is acquired by robots where minimal anal-
ysis is performed. Real-time decisions are taken at the Edge
gateways to minimize the latency and share the computational
load of different connected robots. The Fog layer provides
additional services such as distributed storage and processing
to enable an efficient handover mechanism with minimal data
loss when a robot switches from one gateway to another.
The handover includes sharing previous map data acquired by
the robot switching gateways so that the result of the SLAM
algorithm remains constant through the connection change.
Edge gateways also play an important role in reducing the
energy consumption at the end nodes or robots by running
the most computational expensive operations. Both Fog and
Edge computing are essential in safety critical situations where
they provide a more robust situational awareness and overall
system control, and robot state knowledge. Finally, the Cloud
layer enables end-users or administrators to control the system
by giving general instructions and monitor the performance at
different layers.
A. Robot Layer
In the robot layer, sensor data is gathered and streamed
in real time to the smart Edge gateways. The control of the
robot is run online and movement instructions are given in
real-time. In the proposed system, the robot can be either
aware or unaware of its current state depending on how
extensively information is analyzed on-board and whether an
Robot Layer Edge-Fog Layer Cloud Layer
High Computation
Cloud Storage
Web Applications
Control Panel
Distributed Storage
Task Coordination
Gateway Management
Motor Control
Proximity Detection
Optional features
Robot 1
Robot 2
Edge / Smart Gateway 1
Edge / Smart Gateway 2
Robot N Edge / Smart Gateway N
Fig. 1. Proposed System Architecture
Edge gateway provides enhanced state information to the robot
or only movement instructions.
The robot relies on low power wireless communication tech-
nologies such as Bluetooth or nRF if the bandwidth require-
ments are met, or Wi-Fi for more bandwidth-intensive appli-
cations such as high-quality video transmission. By streaming
sensor data directly to the gateway without being processed by
the MCU, the robot gains in energy efficiency. Furthermore,
low-power techniques can be applied to the MCU in order to
further reduce the overall power requirements.
Compared to a traditional approach where sensing, obsta-
cle detection, path planning and localization are made on-
board the robot, we can significantly reduce the computing
requirements of the robot’s control system and downgrade
them from processor-based to MCU-based. Only the most
basic information is analyzed at the robot itself. This includes,
for instance, wheel odometry or inertial data so that the robot is
aware of its current acceleration, velocity and orientation. This
can be combined with state estimation performed online at the
Edge, Fog or Cloud layers to allow more accurate movement
and simplify the instructions given to the robot.
B. Smart Edge Gateway Layer
Edge gateways receive sensor data from one or multiple
robots in real-time. This includes, for example, images from
on-board cameras or data from range sensors such as Lidars
and radars, which is analyzed at the gateways to perform
obstacle detection and avoidance, path planning, localization
or mapping. Odometry data (e.g. wheel odometry, visual
odometry) and inertial data (e.g. magnetometer, accelerometer
and gyroscope data) is also transmitted from and processed at
the gateway.
Acting as a central element, Edge gateways ensure a low-
latency and robust solution and provide fast decision making
while relocating computationally expensive tasks from the
robots. Moreover, if multiple robots are connected to a single
gateway, data from different sensors and sources is aggre-
gated and analyzed in order to obtain a more comprehensive
understanding of the environment. For instance, two robots
operating in the same environment and connected to the same
gateway are able to obtain information of larger areas through
sensors of other robots and see through other agents if these
are nearby and blocking their field of view.
The role of the Edge layer is of superior importance
in safety-critical situations and scenarios such as industrial
environments where human share operational space with au-
tonomous or semi-autonomous robots. Compared to the more
traditional practice of moving complex tasks to the Cloud,
the Edge layer reduces the latency to the point where safe
operation is achievable even in the case of network connectiv-
ity failure. From an operational point of view, smart Edge
gateways receive information from sensors aboard multiple
robots and are in charge of decision-making in terms of
robot movement and task allocation. These instructions are
transmitted continuously and wirelessly to the robots.
C. Fog Layer
The Fog layer includes the smart Edge gateways as part
of it. Nonetheless, it is defined as a separate layer because
its role in the proposed system architecture differs from the
data analysis and instantaneous control role of the Edge layer.
The Fog layer represents the interconnection of different Edge
gateways, together with other services such as distributed
databases or location services. In particular, the Fog layer takes
care of the handover mechanisms when the robots disconnect
from one gateway and connect to another one together with
the procurement of additional services including distributed
storage, external tracking and monitoring or location services.
Detailed research has been performed to minimize latency
and data-loss during the handover [20] [21]. An alternative for
managing the handover is to deploy Edge gateways just behind
conventional gateways such as Wi-Fi routers so that the gate-
ways can directly store the received information in distributed
storage. Thus, the handover problem would be reduced to the
distribution of the computational load of individual robots into
the set of Edge nodes, since all nodes are able to access robot
data in real-time from the shared storage.
Localization algorithms are mostly run in the Edge layer
in which real-time sensor data is matched with an area of
an existing map. This is crucial in situations where the robot
operates in a partially or totally unknown environment and
simultaneous localization and mapping algorithms are run in
the smart Edge gateways. In these cases, it is required that the
local map stored in one gateway is shared to other gateways
before or during the handover. Accordingly, our proposed
system architecture includes a distributed storage deployment
in the Fog layer. This can be implemented, for instance, in the
form of a decentralized database where the maps and other
essential robot data are stored and available to all nodes in the
Edge/Fog layers at any time.
Other services included in the Fog layer are external sensor
management, collaborative processing and monitoring. For
example, while Edge gateways process the on-board sensor
information from the robots, external monitoring cameras
and other sensors can be used to increase the accuracy of
localization algorithms or assure collision avoidance in cases
of sensor failure or other local problems that might appear
in the robot and go undetected by the Edge gateway. This
results in an enhanced situational awareness [22] where both
externally and internally collected information about a robot is
available in the Fog layer. By introducing additional services
in the Fog layer, we can enhance overall system robustness
and its fault tolerance. For instance, if a gateway abruptly
disconnects or any kind of failure occurs, all the information
that it was handling will be available to other gateways due to
the shared storage resources. This allows the robot to reconnect
to the next available gateway and continue its operation with
minimal data loss, operational interruption and latency.
D. Cloud Services
Time-series data of the robot state, including basic infor-
mation such as position, velocity, acceleration, steering drift
factor, drive torque variation, instantaneous current consump-
tion and battery level is uploaded to the Cloud for monitoring,
control and visualization by users or administrators [23].
Instead of uploading raw sensor data, it is pre-processed
and compressed in the Edge/Fog layers and only sporadic
local map updates or critical information is sent. This can be
implemented in a dynamic way in which, for instance, the map
around a robot or of a given area in the operation environment
is updated and uploaded to the Cloud at different frequencies
depending on whether an end-user is actively monitoring the
operation or not.
The use of Cloud services facilitates generic control in-
structions to the robot and stable operation [24]. For example,
factory managers can easily have an overview of the posi-
tion of different robots around the factory floor, or override
autonomous roaming to involve more robots assigned within
a certain group or to prioritize specific tasks. Besides, the
Cloud allows global access for the administrators and mission
Fig. 2. Prototype car with Lidar running on top, used for the experiment
planners to have a collaborative overview in a cost-effective
However, due to comparatively higher end-to-end latency
between the robot layer and the Cloud, it should not be
used for primary control purpose as it can be subject to
variation in network performance. The Cloud provides access
to pertinently processed data from the robot layer which is
periodically backed up at a reasonable interval. This yields a
conclusive operation and data history which can be analyzed
using high performance computing in the Cloud for machine
learning, better prediction and big data analytics.
In order to test the feasibility and validate the proposed
architecture, we have constructed a car controlled via Blue-
tooth communication to produce a map of the perimeter it is
bounded by. The car shown in Figure 2 incorporates a 360o
Lidar on top of the chassis, two independent motors- one for
steering control and another for driving forward or backward
with appropriate motor control circuitry. The car is connected
to a gateway through a Bluetooth Classic module. An AVR
8-bit MCU is used to control acquisition of raw data from
the Lidar and send to gateway via the serial port profile of
the Bluetooth standard. A prototype is built using the chassis
from an Reely 1:10 Elektro-Monstertruck NEW1 RC car and
replacing the control circuit and motors with lower power
variants with an RPLIDAR A1 M8 on the top, a low cost
360o2D LASER scanner from Slamtec [25].
We have run experimental tests with a minimal setup where
only Lidar is used for both localization and mapping purposes.
Adding inertial sensors would provide more accurate position-
ing and control over the vehicle; however, that is not the main
objective of this work and we focus on the implementation of
the hybrid Edge-Fog-Cloud architecture and its benefits.
The MCU controlling the movement of the car does not
save or analyze the data from the Lidar sensor in any form
and is forwarded directly to the gateway as serial stream over
Bluetooth. The gateway then runs an adapted version of the
BreezySLAM algorithm by S. D. Levy [26]. The algorithm has
been modified to take into account the intermittent nature of
Fig. 3. Mapping and Localization at the Edge. (The position of the car is
marked with a blue pointer)
the connection between the robot and the gateway, and execute
accordingly when processing successive batches of data.
Figure 3 shows the result of applying the mapping and
localization algorithm in the Edge gateway. As we are using
an inexpensive Lidar and the SLAM algorithms run without
any inertial measurement unit such as accelerometer and mag-
netometer, accurate positioning is hard to achieve in complex
environments. However, with more input data from multiple
sensors would suffice for enabling operation in non-trivial
local environments.
The first map in Figure 3 shows the map obtained after
the first few Lidar frames have been analyzed. The RPLIDAR
A1 M8 used in the tests is a low-cost Lidar and therefore
several scans are processed before obtaining a good quality
map. The granularity in the first map is derived from the
relatively low scan speed of the device compared to more
advanced sensors. Also, while the position within the map is
accurate and consistent through the test, noise appears near the
distant walls from the Lidar due to the low accuracy. However,
Operation Average Current Average Power
/ Approach (mA) (W)
MCU Idle 4 19
MCU + Lidar 339 1695
MCU + Lidar + Streaming 378 1890
R.Pi Idle 301 1503
R.Pi + Lidar 622 3108
R.Pi + Lidar + Edge Processing 829 4152
Car (Drive / Steer) 122 607
Car (Drive + Steer) 184 918
the walls and corners in the local area around the robot are
with little to no noise.
In terms of performance, the recorded path elapsed a total
of 44.3 seconds in which 314 Lidar frames were processed.
This translates into 7 frames/s, which matches with the Lidar
specification for a maximum speed scan of 10 Hz with an
average car speed of 0.3 m/s. This frame rate can be compared
to 1 frame/s processed on a Raspberry Pi 3 Model B running
Ubuntu Mate Desktop with the graphical map display in real-
time. While this can be optimized running a more trimmed,
customized version of Linux on the Raspberry Pi, its compu-
tational capability limits the processing to about 5 frames/s.
Since a MCU is able to handle the vehicle control and
transmission of Lidar data, not only it significantly increases
the system performance by offloading the SLAM tasks to the
Edge gateway, but a considerable amount of energy is saved
by replacing a power-hungry single-board computer such as
the Raspberry Pi with a low-power MCU.
In particular, the specific power consumption and average
current drawn by the prototype in different states are docu-
mented in Table I. In idle state, the Lidar consumes 300 to 350
mA, representing over 90% of the total power consumption.
This can be replaced with low-power cameras or more en-
ergy efficient sensors for a particular SLAM implementation.
However, when the prototype is moving, the Lidar’s share
of the power consumption goes down to around 60%. When
choosing the Lidar, general purpose motors have been used
to show that even with out-of-the-shelf devices, the overall
energy efficiency can be improved by around 40% when a
single-board computer such as a Raspberry Pi in this test is
replaced with a MCU. This yields a significantly longer battery
life and an increase in operational performance due to larger
ratio of active-time with respect to idle or charging time.
Although Edge computing can bring advantages such as
reduction in latency, bandwidth conservation, improvement in
application robustness and security [11], [27], [28], there are
inherent challenges. As the processing of the data stays near
the Edge, specific data-oriented applications which require
comparatively higher user interaction will have insignificant
performance gain. In the worst case scenario, performance can
deteriorate due to the long path data has to travel.
As multiple robots can connect to the same smart gate-
way, communication latency and inter-unit interference can
be an issue. Depending on the application, an appropriate
communication medium must be chosen to ensure that specific
parameters such as bandwidth, channel spacing, protocol,
sampling frequency and data rate are properly considered for
an interruption-free data transfer. Also, for wireless units,
it is useful to examine the wake-up and sleep time of the
radio communication module so that transmitter and receiver
circuitry is not toggled unnecessarily and sleep modes are
used appropriately. This yields better energy efficiency and
communication at an optimal duty cycle.
Keeping the Edge gateways near the robot layer and a Fog
layer provides distributed and scalable computing. However,
as the number of robots and frequency of data transmission per
node significantly increases, the gateways may not handle all
requests within the highest permitted latency. In such a case,
having more powerful and redundant gateways along with
advanced load-sharing algorithm in the Fog and distributed
data storage can effectively improve performance.
In this paper, the comparative advantage of using Edge
computing for robots and nodes are discussed keeping the
focus on energy efficiency and operational latency. Instead
of directly transferring a large amount of data from robots
to the Cloud, we proposed an Edge-Fog-Cloud based system
for performance advantages. Run from the mains supply, the
powerful Edge layer can pre-process and analyze data easily,
implement advanced features and ensure data security by
applying complex encryption algorithms. The Fog layer can
intelligently manage the smart gateways of Edge layer.
This approach effectively takes the heavy computational
tasks near the Edge layer and thus reduces the required energy
for operation of the node or robot itself. The shorter path of
data for a complete cycle improves latency and frees up valu-
able network bandwidth, especially when a lot of robots are
connected to the same gateway. In addition, the reduced robot-
to-Cloud transferable data eliminates potential bottleneck in
the network and lowers the chance of system downtime due
to a single point computing failure as in centralized computing
architecture. We demonstrated an example of SLAM showing
how a car in the robot layer can take advantage of the power
of Edge computing for mapping a confined perimeter.
In future, we plan to implement a comprehensive system
consisting of several vehicles, multiple smart Edge computing
gateways constituting a Fog layer and an application for the
Cloud layer. In addition, we plan to develop an autonomous
roaming algorithm for the robot car so as to perform the
mapping operation in shortest possible time. Besides, some
stress tests to estimate the performance of smart gateways
will be evaluated to emulate the scenario when the number
of connected cars is sufficiently high.
[1] N. Mitabe and N. Shinomiya. An intelligent care support system for the
elderly with an autonomous robot in ambient sensors. In 2017 IEEE
GCCE, pages 1–4, 2017.
[2] M. Jiang et al. Iot-based remote facial expression monitoring system
with semg signal. In 2016 IEEE SAS, pages 1–6. IEEE, 2016.
[3] S. R. Moosavi et al. Session resumption-based end-to-end security for
healthcare internet-of-things. In IEEE CIT, pages 581–588. IEEE, 2015.
[4] M. Ali et al. Autonomous patient/home health monitoring powered by
energy harvesting. In IEEE GLOBECOM 2017, pages 1–7. IEEE, 2017.
[5] T. N. Gia et al. Iot-based continuous glucose monitoring system: A
feasibility study. Procedia Computer Science, 109:327–334, 2017.
[6] I. Tcarenko et al. Energy-efficient iot-enabled fall detection system with
messenger-based notification. In International Conference on Wireless
Mobile Communication and Healthcare, pages 19–26. Springer, 2016.
[7] R. Petrolo et al. Integrating wireless sensor networks within a city cloud.
In IEEE SECON Workshops, pages 24–27, 2014.
[8] T. Kitanouma et al. Cloud-based self-organizing localization with virtual
network topology for wireless sensor networks and its implementation.
In 2016 IEEE PIMRC, pages 1–7, 2016.
[9] W. Yu et al. A survey on the edge computing for the internet of things.
IEEE access, 6:6900–6919, 2018.
[10] B. Negash et al. Leveraging fog computing for healthcare iot. In Fog
Computing in the Internet of Things, pages 145–169. Springer, 2018.
[11] T. N. Gia et al. Exploiting fog computing in health monitoring. Fog
and Edge Computing: Principles and Paradigms, pages 291–318, 2019.
[12] S. Dey and A. Mukherjee. Robotic slam: a review from fog computing
and mobile edge computing perspective. In MobiQuitous, pages 153–
158. ACM, 2016.
[13] P. Benavidez et al. Cloud-based realtime robotic visual slam. In 2015
Annual IEEE Systems Conference (SysCon) Proceedings, pages 773–
777, 2015.
[14] L. Riazuelo et al. C2tam: A cloud framework for cooperative tracking
and mapping. Robotics and Autonomous Systems, 62(4):401 – 413, 2014.
[15] L. Turnbull and B. Samanta. Cloud robotics: Formation control of a
multi robot system utilizing cloud infrastructure. In 2013 Proceedings
of IEEE Southeastcon, pages 1–4, 2013.
[16] J. Salmer´
on-Garcı´a et al. A tradeoff analysis of a cloud-based robot
navigation assistant using stereo image processing. IEEE Transactions
on Automation Science and Engineering, 12(2):444–454, 2015.
[17] Y. Hong et al. Energy efficient content-based image retrieval for
mobile systems. In 2009 IEEE International Symposium on Circuits
and Systems, pages 1673–1676, 2009.
[18] J. Hauswald et al. A hybrid approach to offloading mobile image
classification. In 2014 IEEE International Conference on Acoustics,
Speech and Signal Processing (ICASSP), pages 8375–8379, 2014.
[19] K. Kumar et al. A survey of computation offloading for mobile systems.
Mobile Networks and Applications, 18(1):129–140, 2013.
[20] X. Wang et al. Offloading in internet of vehicles: A fog-enabled
real-time traffic management system. IEEE Transactions on Industrial
Informatics, 14(10):4568–4578, 2018.
[21] T. N. Gia et al. Fog computing approach for mobility support in internet-
of-things systems. IEEE Access, 6:36064–36082, 2018.
[22] C. V. Smith et al. Enhanced situational awareness in autonomous mobile
robots using context-based mapping. In 2013 IEEE CogSIMA, pages
134–138, 2013.
[23] B. Zhou et al. Remote control system of mobile robot based on
cloud platform. In 2017 2nd International Conference on Robotics and
Automation Engineering (ICRAE), pages 94–98, 2017.
[24] C. Briese et al. Cloud-based active disturbance rejection control for
industrial robots. In 2018 IEEE ETFA, pages 559–565, 2018.
[25] Slamtec. RPLIDAR A1. [Online] Available:
en/lidar/a1. Accessed: 20 Feb. 2019.
[26] S. D. Levy. Breezyslam. [Online] Available:
simondlevy/BreezySLAM. Accessed: 20 Feb. 2019.
[27] T. N. Gia et al. Fog computing in healthcare internet of things: A case
study on ecg feature extraction. In 2015 IEEE CIT, pages 356–363.
IEEE, 2015.
[28] N. Abbas et al. Mobile edge computing: A survey. IEEE Internet of
Things Journal, 5(1):450–465, 2018.
... The MEC and ICC enabled implementations allow the co-location of independent applications on a shared edge/fog node through virtualized abstraction. As illustrated in Ta- ble 1, some of the existing studies where the main focus is on the low-latency and computation offload [9][10][11][12][13][14], [16], [18], [19], [21], [23], [24], [27][28][29] also adapt the concept of virtualization that allows robotic services to reuse the surrounding hardware and deploy applications on demand. ...
... The close proximity of MEC and IIC has been studied in some recent existing experimental studies to offload the location-based robotics services from mobile robots with limited computational and low-energy resources [17], [25], [29][30][31]. Additionally, [12][13][14], [22], [24] elaborate on the reduced network pressure and improved security and privacy of the robot sensor data that can be offered by restricting the access within a trusted private infrastructure. ...
Full-text available
Cloud-based robotics systems leverage a wide range of Information Technologies (IT) to offer tangible benefits like cost reduction, powerful computational capabilities, data offloading, etc. However, the centralized nature of cloud computing is not well-suited for a multitude of Operational Technologies (OT) nowadays used in robotics systems that require strict real-time guarantees and security. Edge computing and fog computing are complementary approaches that aim at mitigating some of these challenges by providing computing capabilities closer to the users. The goal of this work is hence threefold: i) to analyze the current edge computing and fog computing landscape in the context of robotics systems, ii) to experimentally evaluate an end-to-end robotics system based on solutions proposed in the literature, and iii) to experimentally identify current benefits and open challenges of edge computing and fog computing. Results show that, in the case of an exemplary delivery application comprising two mobile robots, the robot coordination and range can be improved by consuming real-time radio information available at the edge. However, our evaluation highlights that the existing software, wireless and virtualization technologies still require substantial evolution to fully support edge-based robotics systems.
... Normally, the real-time position of robots can be tracked and a system administrator can send the control commands in case of complex situations. Some computation-related tasks such as complex navigation, natural language processing, and machine learning-based intelligence applications can be forwarded to remote cloud servers to facilitate these IoT devices for performing smooth operations [3]. It may be achieved by a proxy server that is responsible for collecting data from these devices, applying some desired processing, and forwarding it to cloud servers. ...
Full-text available
Smart applications and intelligent systems are being developed that are self-reliant, adaptive, and knowledge-based in nature. Emergency and disaster management, aerospace, healthcare, IoT, and mobile applications, among them, revolutionize the world of computing. Applications with a large number of growing devices have transformed the current design of centralized cloud impractical. Despite the use of 5G technology, delay-sensitive applications and cloud cannot go parallel due to exceeding threshold values of certain parameters like latency, bandwidth, response time, etc. Middleware proves to be a better solution to cope up with these issues while satisfying the high requirements task offloading standards. Fog computing is recommended middleware in this research article in view of the fact that it provides the services to the edge of the network; delay-sensitive applications can be entertained effectively. On the contrary, fog nodes contain a limited set of resources that may not process all tasks, especially of computation-intensive applications. Additionally, fog is not the replacement of the cloud, rather supplement to the cloud, both behave like counterparts and offer their services correspondingly to compliance the task needs but fog computing has relatively closer proximity to the devices comparatively cloud. The problem arises when a decision needs to take what is to be offloaded: data, computation, or application, and more specifically where to offload: either fog or cloud and how much to offload. Fog-cloud collaboration is stochastic in terms of task-related attributes like task size, duration, arrival rate, and required resources. Dynamic task offloading becomes crucial in order to utilize the resources at fog and cloud to improve QoS. Since this formation of task offloading policy is a bit complex in nature, this problem is addressed in the research article and proposes an intelligent task offloading model. Simulation results demonstrate the authenticity of the proposed logistic regression model acquiring 86% accuracy compared to other algorithms and confidence in the predictive task offloading policy by making sure process consistency and reliability.
... The principle behind exploiting the gain formula is to make each in-group overlapping degree as large as possible, while the out-group is as small as possible. Again, we assign gain max to the current group and remove it from , and the neighboring vertices N ( gain max ) will be added to the buffer (Line [17][18][19][20]. In this step, we still need to check whether | | ≤ is met (Line 21-23). ...
With the wide penetration of smart robots in multifarious fields, Simultaneous Localization and Mapping (SLAM) technique in robotics has attracted growing attention in the community. Yet collaborating SLAM over multiple robots still remains challenging due to performance contradiction between the intensive graphics computation of SLAM and the limited computing capability of robots. While traditional solutions resort to the powerful cloud servers acting as an external computation provider, we show by real-world measurements that the significant communication overhead in data offloading prevents its practicability to real deployment. To tackle these challenges, this paper promotes the emerging edge computing paradigm into multi-robot SLAM and proposes RecSLAM, a multi-robot laser SLAM system that focuses on accelerating map construction process under the robot-edge-cloud architecture. In contrast to conventional multi-robot SLAM that generates graphic maps on robots and completely merges them on the cloud, RecSLAM develops a hierarchical map fusion technique that directs robots' raw data to edge servers for real-time fusion and then sends to the cloud for global merging. To optimize the overall pipeline, an efficient multi-robot SLAM collaborative processing framework is introduced to adaptively optimize robot-to-edge offloading tailored to heterogeneous edge resource conditions, meanwhile ensuring the workload balancing among the edge servers. Extensive evaluations show RecSLAM can achieve up to 39% processing latency reduction over the state-of-the-art. Besides, a proof-of-concept prototype is developed and deployed in real scenes to demonstrate its effectiveness.
... Authors in [26] state that the edge should provide on-demand services, while the cloud should be invoked only when it is necessary. An illustration of system offloading tasks at the design stage can be found in [27], where a simultaneous localization and mapping for indoor mobile robots employ edge-fog-cloud computing architecture designed to segregate tasks among three layers. The lowest layer, called "the robot layer", collects and forwards data towards the upper layers. ...
Full-text available
The utilization of edge-enabled cloud computing in unmanned aerial vehicles has facilitated advances in autonomous control by employing computationally intensive algorithms frequently related to traversal among different locations in an environment. A significant problem remains in designing an effective strategy to offload tasks from the edge to the cloud. This work focuses on creating such a strategy by employing a network evaluation method built on the mean opinion score metrics in concoction with machine learning algorithms for path length prediction to assess computational complexity and classification models to perform an offloading decision on the data provided by both network metrics and solution depth prediction. The proposed system is applied to the A* path planning algorithm, and the presented results demonstrate up to 94% accuracy in offloading decisions.
It is well known that power plants worldwide present access to difficult and hazardous environments, which may cause harm to on-site employees. The remote and autonomous operations in such places are currently increasing with the aid of technology improvements in communications and processing hardware. Virtual and augmented reality provide applications for crew training and remote monitoring, which also rely on 3D environment reconstruction techniques with near real-time requirements for environment inspection. Nowadays, most techniques rely on offline data processing, heavy computation algorithms, or mobile robots, which can be dangerous in confined environments. Other solutions rely on robots, edge computing, and post-processing algorithms, constraining scalability, and near real-time requirements. This work uses an edge-fog computing architecture for data and processing offload applied to a 3D reconstruction problem, where the robots are at the edge and computer nodes at the fog. The sequential processes are parallelized and layered, leading to a highly scalable approach. The architecture is analyzed against a traditional edge computing approach. Both are implemented in our scanning robots mounted in a real power plant. The 5G network application is presented along with a brief discussion on how this technology can benefit and allow the overall distributed processing. Unlike other works, we present real data for more than one proposed robot working in parallel on site, exploring hardware processing capabilities and the local Wi-Fi network characteristics. We also conclude with the required scenario for the remote monitoring to take place with a private 5G network.
Nowadays, numerous facilities operate with little or even no human supervision. Considering electrical power substations, they provide essential services, can be entirely autonomous or remotely operated, and may be difficult to access. This scenario increases the need of remote monitoring for predictive maintenance and problem analysis. 3D reconstruction processes are concentrated in academic and commercial applications, but they are non scalable and fail to provide remote real time data analysis. This research proposes a novel color 3D scanner architecture environment for remote real time multiple sensor processing and reconstruction. The goal is to present a more efficient system based upon scalability and low latency applications, where multiple sensors can be added without losing the overall processing capability. For this purpose, the present study approaches this problem in two ways. First, it improves 3D reconstruction algorithms by enhancing individual performance. Second, to optimize the entire system, it distributes the running processes in individual layers interconnected through an edge-fog architecture. It enables the use of multiple devices by optimizing payload distribution, latency, and throughput in the network. Unlike previous studies, the results present a thorough analysis of architecture efficiency when multiple sensors are operating in parallel unlike the traditional centralized architecture. Finally, this work provides the basis for real-time remote presence applications.
Full-text available
The agnostic robotic paradigm (ARP) represents a recent development as the use of robots becomes more common, and there is a need for agnostic robots to cope with rich artificial objects environments. All parties and stakeholders need to seize the imminent opportunity and act on ushering in the revolutionary changes of contemporary robotic and facility control solutions. The scalability and effectiveness of robotic enterprise solutions depend primarily on the availability of operational information, robotic solutions, and their information infrastructure. However, different functions and software of robotics and facilities are being launched in the market. Therefore, this paper investigates the implementation of the emerging ARP for the Industrial Internet of Things (IIoT) and resource synchronisation flexible robotic and facility control system to address this challenge. We propose an Artificial Intelligence (AI) edge intelligence and IIoT-based agnostic robotic architecture for resource synchronisation and sharing in manufacturing and robotic mobile fulfillment systems (RMFS). We adopted simultaneous localisation and mapping (SLAM) as one of the edge intelligence, provided the simulation results, and tested with multiple parameters under different conflicts. Our research suggests that purposely developing an ARP for flexible robotic and facility control system via IIoT assisted with AI-edge intelligence are a good solution for both operational and management level under a cloud platform.
Conference Paper
Simultaneous Localization and Mapping plays a key role in different Augmented and Mixed Reality applications to determine the pose, that is, the position and orientation of the AR user in a 3D coordinate system, in relation with the rendered digital objects. To meet the ever-growing resource demands of these SLAM algorithms, the localization task can be offloaded to edge cloud platforms. Although pose calculation techniques are optimized for resource-limited robotic environments, the volatile nature of cloud platforms with the strict requirements of realtime AR applications can still lead to deteriorated performance. In this paper, we propose an LSTM-driven overload control mechanism that can effectively improve the worst-case response time of an edge-assisted SLAM by predicting overloaded periods in advance. Our main contribution is threefold. First, we identify factors influencing the response time of edge-assisted SLAMs fed by real-time AR applications and propose two applicable control actions. Second, we present our control architecture including an encoder-decoder LSTM model that can forecast response time degradation by using a specific image quality metric. Third, we demonstrate the applicability and performance of our proposed control methods along with their effects on the pose accuracy by performing dedicated experiments with challenging motion patterns and the widely-known EuRoC benchmarking dataset.
Full-text available
Handover mechanism for mobility support in a remote real-time streaming IoT system was proposed in this paper. The handover mechanism serves to keep the connection between sensor nodes and a gateway with a low latency. The handover mechanism also attentively considers oscillating nodes which often occur in many streaming IoT systems. By leveraging the strategic position of smart gateways and Fog computing in a real-time streaming IoT system, sensor nodes’ loads were alleviated whereas advanced services, like push notification and local data storage, were provided. The paper discussed and analyzed metrics for the handover mechanism based on Wi-Fi. In addition, a complete remote real-time health monitoring IoT system was implemented for experiments. The results from evaluating our mobility handover mechanism for mobility support shows that the latency of switching from one gateway to another is 10% - 50% less than other state-of-the-art mobility support systems. The results show that the proposed handover mechanism is a very promising approach for mobility support in both Fog computing and IoT systems.
This chapter exploits fog computing in health‐monitoring Internet‐of‐Things (IoT) systems for enhancing the quality of healthcare service. It shows an overview of the architecture of an IoT‐based system with fog computing. Fog computing services locating in a fog layer of smart gateways are diversified for serving IoT applications. The chapter discusses the fog computing services in smart e‐health gateways. The health‐monitoring IoT system consists of several wearable sensor nodes, smart gateways with fog services, cloud servers, and terminals. The chapter discusses detailed implementations of these components. It provides a case study, experimental results, and evaluation related to heart rate variability (HRV) analysis. The chapter presents the related applications in fog computing and discusses future research directions. Fog computing demonstrates that it is one of the most suitable candidates for augmenting IoT systems in healthcare and other domains.
Fog computing has been merged with Internet of Vehicle (IoV) systems to provide computing resource for end users, by which low-latency and high computational capacity can be guaranteed. In this work, we propose a feasible solution that enables offloading for real-time traffic management in fog-based IoV systems, aiming at minimizing the average response time for events reported by vehicles. Specially, we construct a distributed city-wide traffic management system, in which vehicles close to RSUs can be utilized as fog nodes. Then, we model parked and moving-vehicle based fog nodes according to queueing theory, and draw the conclusion that moving-vehicle based fog nodes can be modeled as an M/M/1 queue. An approximate approach is developed to solve the formulated optimization problem by decomposing it into two subproblems. Performance analyses based on a real taxi-trace data set are conducted to demonstrate the superiority of our method.
The Internet of Things (IoT) now permeates our daily lives, providing important measurement and collection tools to inform our every decision. Millions of sensors and devices are continuously producing data and exchanging important messages via complex networks supporting machine-to-machine communications and monitoring and controlling critical smart-world infrastructures. As a strategy to mitigate the escalation in resource congestion, edge computing has emerged as a new paradigm to solve IoT and localized computing needs. Compared with the well-known cloud computing, edge computing will migrate data computation or storage to the network “edge”, near the end users. Thus, a number of computation nodes distributed across the network can offload the computational stress away from the centralized data center, and can significantly reduce the latency in message exchange. In addition, the distributed structure can balance network traffic and avoid the traffic peaks in IoT networks, reducing the transmission latency between edge/cloudlet servers and end users, as well as reducing response times for real-time IoT applications in comparison with traditional cloud services. Furthermore, by transferring computation and communication overhead from nodes with limited battery supply to nodes with significant power resources, the system can extend the lifetime of the individual nodes. In this paper, we conduct a comprehensive survey, analyzing how edge computing improves the performance of IoT networks. We categorize edge computing into different groups based on architecture, and study their performance by comparing network latency, bandwidth occupation, energy consumption, and overhead. In addition, we consider security issues in edge computing, evaluating the availability, integrity, and confidentiality of security strategies of each group, and propose a framework for security evaluation of IoT networks with edge computing. Finally, we compare the performance of various IoT applications (smart city, smart grid, smart transportation, etc.) in edge computing and traditional cloud computing architectures.
Mobile Edge Computing (MEC) is an emergent architecture where cloud computing services are extended to the edge of networks leveraging mobile base stations. As a promising edge technology, it can be applied to mobile, wireless and wireline scenarios, using software and hardware platforms, located at the network edge in the vicinity of end-users. MEC provides seamless integration of multiple application service providers and vendors towards mobile subscribers, enterprises and other vertical segments. It is an important component in the 5G architecture which supports variety of innovative applications and services where ultra low latency is required. This paper is aimed to present a comprehensive survey of relevant research and technological developments in the area of MEC. It provides the definition of MEC, its advantages, architectures, and application areas; where we in particular highlight related research and future directions. Finally, security and privacy issues and related existing solutions are also discussed.
Developments in technology have shifted the focus of medical practice from treating a disease to prevention. Currently, a significant enhancement in healthcare is expected to be achieved through the Internet of Things (IoT). There are various wearable IoT devices that track physiological signs and signals in the market already. These devices usually connect to the Internet directly or through a local smart phone or a gateway. Home-based and in hospital patients can be continuously monitored with wearable and implantable sensors and actuators. In most cases, these sensors and actuators are resource constrained to perform computing and operate for longer periods. The use of traditional gateways to connect to the Internet provides only connectivity and limited network services. With the introduction of the Fog computing layer, closer to the sensor network, data analytics and adaptive services can be realized in remote healthcare monitoring. This chapter focuses on a smart e-health gateway implementation for use in the Fog computing layer, connecting a network of such gateways, both in home and in hospital use. To show the application of the services, simple healthcare scenarios are presented. The features of the gateway in our Fog implementation are discussed and evaluated.
Health monitoring systems based on Internet-of-things (IoT) have been recently introduced to improve the quality of health care services. However, the number of advanced IoT-based continuous glucose monitoring systems is small and the existing systems have several limitations. In this paper we study feasibility of invasive and continuous glucose monitoring (CGM) system utilizing IoT based approach. We designed an IoT-based system architecture from a sensor device to a back-end system for presenting real-time glucose, body temperature and contextual data (i.e. environmental temperature) in graphical and human-readable forms to end-users such as patients and doctors. In addition, nRF communication protocol is customized for suiting to the glucose monitoring system and achieving a high level of energy efficiency. Furthermore, we investigate energy consumption of the sensor device and design energy harvesting units for the device. Finally, the work provides many advanced services at a gateway level such as a push notification service for notifying patient and doctors in case of abnormal situations (i.e. too low or too high glucose level). The results show that our system is able to achieve continuous glucose monitoring remotely in real-time. In addition, the results reveal that a high level of energy efficiency can be achieved by applying the customized nRF component, the power management unit and the energy harvesting unit altogether in the sensor device.