Towards a New Generation of Smart Devices for Mobility Assistance:CloudWalker, a Cloud-Enabled Cyber-Physical System

Ricardo Mello, Mario Jimenez, Franco Souza,

Moises R. N. Ribeiro, and Anselmo Frizera-Neto, Member, IEEE

Abstract— The increased computational complexity de-manded by recent algorithms and techniques applied to health-care and social robotics, often limited by the robot’s embeddedhardware, coupled with advancements on networking and cloudcomputing enabled the so-called cloud robotics paradigm. Thiswork explores cloud robotics concepts pointing at opportunitieson the design and development of robotic platforms used forpatient mobility assistance. Moreover, we present CloudWalker,a cloud-enabled cyber-physical system to assist mobility im-paired individuals. The conception of such system envisionsthe integration of smart walkers and remote cloud computingplatforms, aiming at expanding the range of features thesedevices can offer to users, patients, healthcare professionals,and family members. Results from validation experimentspoint to the emergence of a new generation of smart walkersand assistive devices in general, designed to leverage cloudcomputing concepts to provide an extended range of servicesto users, relatives, and healthcare professionals.

I. INTRODUCTION

Mobility is an important human faculty and impacts the

capacity to freely move through multiple environments and

to perform daily chores with ease [1]. Besides decreasing

with age, neurological diseases also affect human mobility on

different levels. Studies have found that mobility difficulties

can be linked with psychosocial disorders and increased

mortality [2].

Assistive devices can be employed to mitigate mobility

impairments. The use of such devices is recommended ac-

cording to the type of condition affecting mobility and taking

into account the level of impairment [1]. Technological

developments lead to the possibility of integrating robotics

concepts, sensors, and circuitry on such devices, giving

birth to a “smart” generation of mobility assistive devices.

Smart wheelchairs, canes, and walkers have been developed

by multiple research groups over the last few decades to

provide new or improved functionalities of physical support,

and sensorial and cognitive assistance [3]. Smart walkers,

for instance, have been designed to improve pathological

gait while exploiting users’ residual locomotion capabilities.

Those devices often make use of support platforms for

*The research leading to these results received funding from the Eu-ropean Commission H2020 program under grant agreement no. 688941(FUTEBOL), as well from the Brazilian Ministry of Science, Technology,Innovation, and Communication (MCTIC) through RNP and CTIC. Thiswork was partially supported by CAPES (88887.095626/2015-01), FAPES(80709036 & 72982608), and CNPq (304192/2016-3).

Authors are with the Electrical Engineering Graduate Program, FederalUniversity of Esprito Santo, Av. Fernando Ferrari, 514, Vitoria, Brazil.Corresponding author: ricardo.mello@ieee.org

upper limbs and offer a variety of features, such as health

monitoring and cognitive assistance [4].

The concept of cyber-physical systems (CPS) is often

applied to enhance the capacities of standalone systems.

CPS are systems that combine sensing, communication,

control, and computing to interact with a physical entity

[5]. CPS concepts encloses multiple systems from various

fields, encompassing systems that integrate physical, control,

computation and communication elements [6].

The evolution of the control algorithms within the robotics

field in general, relying on great amounts of data and com-

putational capabilities, combined with recent developments

on communication and cloud computing technologies gave

rise to the so-called cloud robotics paradigm [7]. On such

paradigm, the individual robot does not rely solely on its own

sensing and processing capabilities. Instead, it communicates

with remote computing platforms to acquire and send data,

instructions, and even control signals, as the cloud should

able to provide nearly unlimited amounts of storage space

and processing power. This allows for the exploitation of

techniques based on recent fields, such as big data and deep

learning, on a larger range of service robots, also allowing

for the use of computationally expensive control algorithms

based on simultaneous localization and mapping (SLAM),

computer vision, trajectory planning, among others [8].

At the light of a paradigm shift, a new generation of

robotic assistive devices is rising. By leveraging the cloud

capabilities, new control techniques can be applied on multi-

robot environments, in which assistive devices should be able

to learn in a faster pace and to adapt to the ever-evolving

impairment condition of individual users. Over such scenario,

mobility assistive devices should evolve to offer new features

to users, patients, therapists, and caretakers.

In this context, this work aims at introducing a cloud-

enabled CPS for mobility assistance, the CloudWalker. This

system envisions the integration of smart walkers and cloud

computing platforms. CloudWalker is in its early stages of

development and the results obtained from validation exper-

iments argue for the feasibility of employing such system.

This opens a door for the development of a new generation of

robotic assistive devices, in which the devices are enabled by

the cloud and able to cope with future connected healthcare.

II. CYBER-PHYSICAL SYSTEMS AND CLOUD

ROBOTICS

Many research works have been conducted on CPS archi-

tectures, and Liu et al. [5] separate CPS’ closed loop control

2018 7th IEEE International Conference on BiomedicalRobotics and Biomechatronics (Biorob)Enschede, The Netherlands, August 26-29, 2018

978-1-5386-8183-1/18/$31.00 ©2018 IEEE 439

in three layers. The first layer comprises the physical system

(i.e., the hardware of the device, including sensors, actuators,

among others). The second comprises the information system

(i.e., real-time control and other data analytics elements), and

the last layer is the user layer, which comprises the way the

user interacts with the CPS.

The usual definitions for CPS encompasses a variety

of systems throughout different engineering fields, from

cochlear implants to industrial plants, and the integration

of the different CPS components intensified recently [6].

In other words, multiple systems that relied solely on em-

bedded hardware have been migrated to distributed systems

employing CPS’s concepts [9]. Computational tasks such as

monitoring and control can be performed by remote plat-

forms, which makes it possible to diminish the computational

capacity available on the physical local device.

CPS’ concepts are often applied to enhance the capacities

of standalone systems. Despite the term not being directly

linked to robotics, the concept is applied in networked

and distributed robotics, in applications as teleoperation and

telemedicine [5]. Smart devices for mobility assistance are

no different, and those concepts are leveraged for remote

health monitoring or storing patient status and history [10].

A. Cloud Robotics in Healthcare

Robotics has been applied in healthcare for some time

and future medical facilities will likely rely on autonomous

devices such as robotic caretakers and vehicles. It is foreseen

that cloud computing can be exploited in healthcare initially

in simple tasks, as in multiple networked sensors for remote

patient status monitoring, moving forward to more advanced

tasks, that can be as demanding as remotely controlled

robotic surgery.

In current healthcare literature, most cloud-enabled so-

lutions explore networked sensors, wearable sensors and

clothing, and wireless body area network concepts. In [10],

smartphones are used to perform patient localization in a

cloud-supported CPS for patient monitoring. In [11], a CPS

for patient-centric healthcare applications and services is

proposed and a similar approach of gathering patient data

and connecting stakeholders through a cloud platform is

presented in [12].

With implications to rehabilitation robotics, data from

an exoskeleton is monitored and stored in a cloud-enabled

application presented in [13]. In [14], a cloud platform

is used by a upper-limb rehabilitation robot to allow the

physician to remotely monitor therapy and adjust parameters

when needed.

Initiatives in social robotics aim at designing cloud-

enabled caregiver and companion robots to assist the elderly.

In [15], a personal health management service is presented,

combining cloud computing services, a domestic robot, an

Android app and a web portal. Services such as localization

and speech recognition are provided and the robot is able to

interact with the user.

Devices used for mobility assistance are beginning to

benefit from the use of cloud computing. In [16], naviga-

tional information from a wheelchair is transmitted to a

cloud storage service, and in [17], the cloud is employed

to share multiple wheelchair positions and the map of the

environment.

There are not many works exploring the integration of

cloud computing and smart walkers. In [18], a multi-robot

system integrated with cloud data center was proposed and

simulated. A thorough search in the literature regarding smart

walkers yielded ANG-MED robot [19] as the only smart

walker currently making use of cloud-based services. It is

a passive smart walker with independent brakes that can be

used to stop the walker or to correct its trajectory during

obstacle avoidance routines.

The ANG-MED was developed within the RAPP project

[20], which aims to provide cloud-based applications for

robots in general. The ANG-MED walker is supported by

a cloud platform and the caregiver was able to remotely

monitor and act upon the smart walker’s brakes. The ANG-

MED has been tested and evaluated by care professionals

and positive results were reported.

Cloud robotics can be further explored in robotic devices

used for mobility assistance to provide an extended range

of features and services. In the next section, we present

CloudWalker, a cloud-enabled CPS that arises from the

integration of smart walkers and remote clouds.

III. CLOUDWALKER: A CLOUD-ENABLED

MOBILITY ASSISTIVE DEVICE

CloudWalker is conceived from the integration of smart

walkers and remote cloud computing platforms, aiming at

expanding smart walkers capabilities and the features those

devices can offer to patients, medical staff, and family mem-

bers. The system explores cloud robotics to unleash smart

walkers capabilities despite possible hardware limitations,

leveraging cloud-based services.

A. System Architecture

The overall architecture of CloudWalker allows for the

conception of smart walkers as cloud-enabled CPS. Such sys-

tem encompasses both physical and cognitive human-robot-

environment interaction to assist individuals with mobility

impairments. CloudWalker’s architecture is depicted in Fig.

1(a).

To provide mobility assistance services, CloudWalker can

perceive user’s motion intents on the walker’s embedded sen-

sors and process such signals on a cloud platform. The smart

walker can send data to be processed on the cloud, which

can provide services such as storage and data processing. The

cloud platform can also generate control signals to command

actuators, in case control algorithms need to be remotely

processed. Moreover, CloudWalker envisions a connected

healthcare, in which data stored on database services can be

accessed by healthcare professionals, which should also be

able to remotely tune parameters regarding therapy routines.

Emergency services must also be directly triggered from

monitoring services when needed, which can potentially

shorten response times and improve rescue efficiency.

440

(a) (b)

Fig. 1. CloudWalker: a) overall architecture, linking robotic assistivedevices and cloud platforms to interact and provide mobility assistance andservices to patients and healthcare professionals; b) robotic platform usedin CloudWalker’s first implementation, the UFES Smart Walker.

CloudWalker architecture decouples the smart walker and

its embedded hardware from the features it can offer. The

centralization of services running on the cloud allows for the

use of multiple heterogeneous cloud-enabled smart walkers

working under the same CloudWalker architecture while

making use of subsets of the cloud-based services. The

gathering of large amounts of data may allow for the

implementation of algorithms able to automatically identify

the user and adapt to its specific needs. The cloud is also

able to offer elasticity and scalability to guarantee the proper

execution of real-time services and to allow the insertion of

new devices into the system.

B. Materials and Methods

CloudWalker is implemented for the first time by integrat-

ing an in-house developed smart walker, a cloud platform,

and a series of controllers processed on virtual machines on

the cloud.

A important building block in CloudWalker’s first imple-

mentation is the UFES Smart Walker (shown in Fig. 1(b)), a

device that has been used in several research works over the

last years (e.g., [4], [21], [22]), mainly focused on human-

robot interaction strategies.

The UFES Smart Walker possesses the kinematics of a

unicycle robot, with a front caster wheel and a pair of

differential rear wheels driven by DC motors. Two 3D

force sensors, MTA400 (Futek, USA), are positioned under

each forearm supporting platform to capture the physical

interaction forces between the user and the walker. The UFES

Smart Walker can also make use of two laser rangefinder

(LRF) sensors to capture user’s legs position and to obtain

environmental information. The embedded computer is based

on the PC/104-Plus standard and integrated into the Simulink

Real-Time (MathWorks, USA) environment.

A Raspberry Pi 3 model B is used as a wireless gateway

for communicating with the cloud platform. Communication

between the Raspberry and the smart walkers’ PC/104-Plus

uses datagram sockets in a strategy based on the one pre-

sented in [23]. In this first implementation, communication

between the Raspberry Pi and the cloud is also based on

datagram sockets. To minimize packet size and subsequent

latency and packet loss rates, only the fundamental infor-

mation (i.e., the controller’s inputs and packet identification

number) is sent to the cloud. Therefore, wireless communi-

cation is based on the exchange of 56 B packets directed

to the cloud platform and 24 B packets (generated control

signal) sent back to the Raspberry Pi.

Another core building block of CloudWalker is the cloud

platform. In this work, the cloud platform used is based

on OpenStack version Ocata. OpenStack is an open-source

cloud management software capable to provide and manage

network, processing, and storage resources in a data center.

By the joint work of several services, OpenStack is able to

provide the so-called Infrastructure as a Service (IaaS) cloud

level. Virtual machines (VM) are instantiated in the cloud

to provide services for CloudWalker. Different computing

resources can be set before VM instantiation, and OpenStack

can be programmed to automatically manage life cycle and

resources to provide elasticity and scalability if necessary.

The Robot Operating System (ROS) is installed in the

cloud’s VMs and in the Rapsberry Pi to manage data

processing, storage, and logging. ROS is an open-source

middleware that provides an abstraction layer to robot’s

hardware resources, treating physical data as streams on

a publisher-subscriber architecture [24]. Independent ROS

masters are instantiated in the cloud and in the Raspberry Pi

to decouple the local device and the remote platform.

IV. CLOUDWALKER: FIRST IMPLEMENTATION

In CloudWalker’s first implementation, a mobility as-

sistance service is instantiated in the cloud. Such service

makes use of an admittance-based controller to command

the physical interaction between user and smart walker.

The overall diagram is shown in Fig. 2. Sensor data is

stored at the Raspberry Pi and force signals are transmitted

to the cloud platform to be processed by the mobility

assistance service. The admittance-based controller is based

on a simplified version of the one presented in [22], and

all the parameters employed by the controller are fixed and

empirically established. Such controller takes the force sig-

nals as input to generate velocity commands to coordinate the

interaction between assistive device and user. In other words,

the service makes use of the interaction forces captured by

the force sensors to provide full control over the walker to

the user.

A. Validation Experiments

The proposed scenario for CloudWalker validation experi-

ments is based on a mobility assistive service in which all the

control algorithms are remotely executed. By delegating all

intelligence and processing tasks to a remote cloud platform,

the system is implemented in its most latency-sensitive con-

figuration, as network latency and packet loss rates impose

441

Fig. 2. CloudWalker validation experiments scenario: the smart walkeracquires force sensors’ signals and sends this data to be processed by thereal-time mobility assistance service at the cloud platform. Control signalsare remotely generated and sent back to be executed by the smart walker’sactuators.

direct influence on control performance. If the feasibility

of the system can be verified under such configuration,

CloudWalker can be further extended to leverage different

configurations, in which some processing power is retained

at the local device and local safety assurance algorithms are

deployed.

A set of experiments is performed to verify the feasibility

of the system when exploring mobile edge computing (MEC)

concepts. Edge clouds are of fundamental importance on

the remote real-time control of cloud-enabled robots, as

the physical proximity and reduced network complexity

mitigates latency issues and packet loss rates. Our cloud

platform runs on an edge data center located in the same

university campus in which the experimentation takes place.

Therefore, CloudWalker’s implementation is validated in a

non-simulated scenario built over a non-structured environ-

ment.

The user conducts CloudWalker on a pedestrian pathway

of approximately 9 m long and 1.8 m wide. An wireless

AP working on the 802.11n 2.4 GHz band is positioned at

approximately 10 meters away from the pathway midpoint

to provide connectivity to CloudWalker. Received signal

strength indicator (RSSI) levels along the pathway were

measured and average values ranging from -65 to -60 dBm

were observed.

Starting and finishing lines are draw to delimit the track

inside the pathway. The walker starts over the starting line,

the user is able to fully control CloudWalkers motion and

there is no specif path to be followed. The user interacts

with the walker heading towards the finishing line, stopping

only when reaching the tracks limit.

Six experiment realizations are conducted over the same

track. Each experiment realization starts after communication

between walker and cloud is established. Sensor data and

control signals are timestamped and stored in the walker for

Fig. 3. CloudWalker validation experiments: odometry recorded from onerealization. The walker starts on the lower right part of the image, over thestarting line of the track, and is guided by the user towards the mark on theleft upper part, representing the finishing line.

offline processing.Another experiment realization, the seventh one, is per-

formed until complete loss of connectivity is observed.

The user commands the walker’s displacement along the

pathway, moving away from the wireless AP, until there is

no communication between the smart walker and the cloud

platform.

B. Results and Discussion

During the first six realizations, no communication loss

was observed and network and cloud parameters impact over

the system could not be perceived by the user during most

of the time. On two occasions, the user described a feeling

of momentary disobedience or discomfort. An illustrative

odometry plot is shown in Fig. 3.Latency and packet loss measurements were performed

to assess the quality of the service provided. Latency was

measured as the round-trip time between the Raspberry Pi

sending information to the cloud and receiving the corre-

sponding control signal, and packet loss rates were calculated

based on the rate of received and sent packets between the

Raspberry Pi and the cloud.On the first six realizations, the average round-trip latency

was 6.8 ms and a variance-to-mean ratio (VMR) of 27.98

ms was observed. Maximum values observed for latency

were often larger than 100 ms, but only observed in eventual

spikes. This impacts latency variability, and the VMR values

show that there is considerable disorder in packet arrival. An

average packet loss rate of 0.47 % was also observed, and

such low values are not expected to impact the use of the

system.The quality of service verified in the first six realizations

of the experiment was sufficient to allow for a comfortable

interaction between user and CloudWalker. Even with all

control algorithms being executed on the cloud platform, the

user could guide the smart walker while making use of the

mobility assistance service. Interaction forces and subsequent

control signals and velocities from one realization of the

experiment are displayed in Fig. 4.The readings from both force sensors are repeated in Figs.

4(a)–(b) so that the impacts of the interaction forces over

442

(a) (b)

(c) (d)

(e) (f)

Fig. 4. Measured interaction forces and resulting velocities on one of the realizations: a,b) forward forces measured in each 3D force sensor, repeatedto ease visualization of interaction effects upon velocities; c) resulting total interaction forward force; d) resulting interaction torque; e) control signal andmeasured values for linear velocities; f) control signal and measured values for angular velocities.

forward force and torque, control signals, and velocities can

be better assessed. The oscillations on the measured force

values result mainly from user gait. During each stride, there

is a tendency to push or pull the walker, differing in each arm

according to the stage of the gait. The total forward force

and torque (Figs. 4(c)–(d)) are calculated as a result of the

measured forces in each sensor, and those values dictate how

CloudWalker must interact with the user. Figure 4 illustrates

the beginning of the movement, the moment of each turn, and

the moment in which the user attempts to stop the walker as

it approaches the finishing line.

The control signal values for linear and angular veloci-

ties (Figs. 4(e)–(f)) are generated by the admittance-based

controller. The control signal for linear velocity is limited to

a safe value for displacement and maximum values can be

observed during most of the experiment, decreasing mostly

when the user is turning the walker. This can be observed

on two periods of time: one ranging from 15 to 20 s, and

the other one ranging from 25 to 32 s, approximately. The

turning periods are indicated by the sign opposition in the

forces measured by the forces sensors, which results in the

largest torque values. CloudWalker could respond properly to

user’s desires and no abnormal behavior could be observed

in any of the first six realizations of the experiment.

The seventh experiment realization is related to connectiv-

ity loss. During this realization, latency, its variability, and

packet loss increased as the walker moved away from the

wireless AP. It was observed a mean end-to-end latency of

100.4 ms (VMR of 761.8 ms) and a ratio of 5.5 % packet

loss. The walker lost connectivity with the cloud platform

after reaching a zone in which RSSI was measured to be

around -85 dBm.

Network and cloud availability are concerns due to mobile

nature of the system. Wireless connectivity and seamless AP

migration are also necessary to guarantee sufficient RSSI

without downtimes. Such issues must be dealt with, and

CloudWalker should be employed taking into consideration

the wireless communication technology. The use of com-

mercial WiFi demands an adequate distribution of AP and

handover strategies to avoid downtimes during AP migration,

such as the one presented in [25].

CloudWalkers first implementation was completed and,

despite network simplicity and physical proximity of the

edge cloud to the walker, results from the validation ex-

periments indicates that CloudWalker can indeed be em-

ployed for mobility assistance. The analysis of the obtained

results points to CloudWalkers feasibility when exploring

MEC concepts. This is probably enhanced by retaining local

processing on the device, which should be able to perform

its own control algorithms whenever poor quality of service

is detected.

V. FUTURE WORK

CloudWalker must be extended to contemplate the full

range of sensors and interfaces that are already integrated

into the smart walker used in this first implementation.

Control and interaction strategies previously developed at

the research group should be replicated using the cloud to

process the necessary algorithms exploring different system

configurations, regarding which tasks must be locally pro-

cessed and which ones can be remotely deployed. Safety

mechanisms must also be developed and implemented to

assure the safe use of the system, which should lead to the

clinical validation of CloudWalker.

Computationally intensive algorithms can now be incor-

porated to CloudWalker. The LRF sensors can be used by

443

probabilistic techniques, such as SLAM, to perform self-

localization and create environmental maps, feeding map

libraries stored by cloud services. Path planning algorithms

can be implemented to assist in guidance tasks, interactively

creating routes to be followed by the user, based on voice

acquired destinations or previously constructed maps. A

camera system can also be installed to explore computational

vision in context-aware applications, exploring interaction

strategies in which facial recognition or the ability to discern

humans from other moving objects are important. Wearable

sensors can also be coupled with the use of CloudWalker

for patient health monitoring. Moreover, a cloud comput-

ing platform must be able to provide services connecting

information from multiple robots scattered among different

facilities and communicating with doctors and therapists via

web applications, always assuring for patient privacy and

personal data safety.

This work envisions a smarter, new generation of mobility

assistive devices. Future work shall encompass the design

of cost-effective robotics devices fully integrated with cloud

computing platforms, able to provide a wider range of ser-

vices for mobility impaired individuals, aiming at mitigating

independence restrictions and improving quality of life.

VI. CONCLUSIONS

A cloud-enabled CPS for mobility assistance, the Cloud-

Walker, was proposed and validated in this work. The system

was implemented by integrating a smart walker and a cloud

platform. CloudWalker was validated against a scenario

involving a real-time mobility assistance service provided by

the cloud. The implemented system proved itself capable of

providing assistive services even when all control algorithms

are remotely executed. These results point to the emergence

of a new generation of smart walkers, designed to leverage

cloud computing concepts to provide an extended range of

services to users, relatives, and healthcare professionals.

REFERENCES

[1] A. Frizera-Neto, R. Ceres, E. Rocon, and J. L. Pons, “Empowering andassisting natural human mobility: The simbiosis walker,” InternationalJournal of Advanced Robotic Systems, vol. 8, no. 3, pp. 34–50, 2011.

[2] M. Hirvensalo, T. Rantanen, and E. Heikkinen, “Mobility difficultiesand physical activity as predictors of mortality and loss of inde-pendence in the community-living older population,” Journal of theAmerican Geriatrics Society, vol. 48, no. 5, pp. 493–498, 2000.

[3] M. M. Martins, C. P. Santos, A. Frizera-Neto, and R. Ceres, “Assis-tive mobility devices focusing on Smart Walkers: Classification andreview,” Robotics and Autonomous Systems, vol. 60, no. 4, pp. 548–562, 2012.

[4] C. A. Cifuentes, C. Rodriguez, A. Frizera-Neto, T. F. Bastos-Filho, andR. Carelli, “Multimodal Human-Robot Interaction for Walker-AssistedGait,” IEEE Systems Journal, vol. 10, no. 3, pp. 933–943, 2016.

[5] Y. Liu, Y. Peng, B. Wang, S. Yao, and Z. Liu, “Review on cyber-physical systems,” IEEE/CAA Journal of Automatica Sinica, vol. 4,no. 1, pp. 27–40, jan 2017.

[6] S. C. Suh, U. J. Tanik, J. N. Carbone, and A. Eroglu, Applied cyber-physical systems, 2014, vol. 9781461473.

[7] J. Wan, S. Tang, H. Yan, D. Li, S. Wang, and A. V. Vasilakos, “Cloudrobotics: Current status and open issues,” IEEE Access, vol. 4, pp.2797–2807, 2016.

[8] B. Qureshi and A. Kouba, “Five traits of performance enhancementusing cloud robotics: A survey,” Procedia Computer Science, vol. 37,pp. 220–227, 2014.

[9] R. Rajkumar, I. L. I. Lee, L. S. L. Sha, and J. Stankovic, “Cyber-physical systems: The next computing revolution,” in Design Automa-tion Conference (DAC), 2010 47th ACM/IEEE, 2010, pp. 0–5.

[10] M. S. Hossain, “Cloud-supported cyber-physical localization frame-work for patients monitoring,” IEEE Systems Journal, vol. 11, no. 1,pp. 118–127, 2017.

[11] Y. Zhang, M. Qiu, C.-W. Tsai, M. M. Hassan, A. Alamri, S. Member,and C.-W. Tsai, “Health-CPS : Healthcare Cyber-Physical SystemAssisted by Cloud and Big Data,” IEEE Systems Journal, vol. 11,no. 1, pp. 88–95, mar 2017.

[12] M. Chen, Y. Ma, J. Song, C. F. Lai, and B. Hu, “Smart Clothing:Connecting Human with Clouds and Big Data for Sustainable HealthMonitoring,” Mobile Networks and Applications, vol. 21, no. 5, pp.825–845, 2016.

[13] C. Radu, C. Candea, and G. Candea, “Towards an Assistive System forHuman,” in Proceedings of the 9th ACM International Conference onPErvasive Technologies Related to Assistive Environments - PETRA’16, 2016, pp. 1–4.

[14] H.-J. J. Li and A.-G. G. Song, “Architectural Design of a CloudRobotic System for Upper-Limb Rehabilitation with Multimodal Inter-action,” Journal of Computer Science and Technology, vol. 32, no. 2,pp. 258–268, mar 2017.

[15] L. Fiorini, R. Esposito, M. Bonaccorsi, C. Petrazzuolo, F. Saponara,R. Giannantonio, G. D. Petris, P. Dario, and F. Cavallo, “Enablingpersonalised medical support for chronic disease management througha hybrid robot-cloud approach,” Autonomous Robots, vol. 41, no. 5,pp. 1263–1276, 2017.

[16] J. Fu, M. Jones, T. Liu, W. Hao, Y. Yan, G. Qian, and Y. K. Jan,“A novel mobile-cloud system for capturing and analyzing wheelchairmaneuvering data: A pilot study,” Assistive Technology, vol. 28, no. 2,pp. 105–114, 2016.

[17] K. Salhi, A. M. Alimi, P. Gorce, and M. M. Ben Khelifa, “Navigationassistance to disabled persons with powered wheelchairs using trackingsystem and cloud computing technologies,” Proceedings - Interna-tional Conference on Research Challenges in Information Science,vol. 2016-Augus, 2016.

[18] P. Benavidez, M. Kumar, S. Agaian, and M. Jamshidi, “Design ofa home multi-robot system for the elderly and disabled,” 2015 10thSystem of Systems Engineering Conference, SoSE 2015, pp. 392–397,2015.

[19] M. Iturburu, E. Goiburu, J. Yanguas, E. Andueza, E. Corral,C. Alderete, A. Orbegozo, D. Daney, V. Prunet, and J.-p. Merlet, “UserNeeds and Requirements for the Mobility Assistance and ActivityMonitoring Scenario within the RAPP Project,” vol. 351, pp. 105–117, 2015.

[20] E. G. Tsardoulias, A. M. Kintsakis, K. Panayiotou, A. G. Thallas, S. E.Reppou, G. G. Karagiannis, M. Iturburu, S. Arampatzis, C. Zielinski,V. Prunet, F. E. Psomopoulos, A. L. Symeonidis, and P. A. Mitkas,“Towards an integrated robotics architecture for social inclusion TheRAPP paradigm,” Cognitive Systems Research, vol. 43, pp. 157–173,jun 2017.

[21] C. Valadao, E. Caldeira, T. Bastos-Filho, A. Frizera-Neto, andR. Carelli, “A new controller for a smart walker based on human-robotformation,” Sensors (Switzerland), vol. 16, no. 7, pp. 1–26, 2016.

[22] M. F. Jimenez, M. Monllor, A. Frizera, T. Bastos, F. Roberti,and R. Carelli, “Admittance controller with spatial modulationfor assisted locomotion using a smart walker,” Journal ofIntelligent & Robotic Systems, May 2018. [Online]. Available:https://doi.org/10.1007/s10846-018-0854-0

[23] M. Wahba, R. Leary, N. Ochoa Lleras, J. Safi, and S. Brennan, “AROS-Simulink Real-Time Communication Bridge Using UDP Witha Driver-in-the-Loop Application,” in ASME 2016 Dynamic Systemsand Control Conference, 2016, pp. 1–8.

[24] A. Koubaa, “ROS As a Service : Web Services for Robot OperatingSystem,” Journal of Software Engineering for Robotics, vol. 1, no. 1,pp. 123–126, 2015.

[25] V. M. G. Martinez, R. Guimaraes, R. Mello, P. Hasse, M. R. N.Ribeiro, M. Martinello, and V. Frascolla, “Ultra reliable communi-cation for robot mobility enabled by sdn splitting of wifi functions,”in IEEE Symposium on Computers and Communication (ISCC), 2018.

444