Model to Hardware: System-level Modeling for Wearable Devices

Daniela Genius and Roselyne Chotin

Sorbonne Université, LIP6, CNRS UMR 7606, Paris, France

Keywords:

Embedded Systems, Analog/Mixed Signal Design, Virtual Prototyping.

Abstract:

Wearable devices which capture and analyze physiological data in a non-invasive and not restraining manner

are increasingly popular, but usually employ specialized hardware. Software is often limited, speciﬁc, and

expensive. This paper shows how system-level modeling, virtual prototyping and electronic design can interact

to help designing low-cost wearable bio-medical devices. We propose a design method which allows validation

against actual electronic/mechanical prototypes and present a ﬁrst case study, a daytime sleepiness detector.

1 INTRODUCTION

Cyber-physical systems (CPS) integrate sensors,

computation and control into physical objects, con-

necting them among each other and to the external

world. The medical domain is a typical application

area of such systems (Dey et al., 2018). The spec-

trum of devices is very large, ranging from patches

for capturing biomedical signals in home hospitaliza-

tion (Silva and Tavakoli, 2020) to hospital beds for

intensive care.

Similar from the technical point of view, but much

less safety critical and subject to lower requirements

for certiﬁcation, a large number of products are pro-

posed in sports and for leisure activities, where the

collection of bio-physical data is used to monitor per-

formance and well-being of the human body.

Wearable devices capture and analyze physiologi-

cal data in a non-invasive and not restraining manner.

Smart watches and bracelets are increasingly popular,

but there are also shorts and stockings equipped with

vibration sensors, clothing with temperature and rain

sensors, smart glasses etc. Industrial wearables are

almost always based on specialized hardware focused

on a (set of) speciﬁc functionalities such as heart rate,

running speed, temperature etc. From an electronics

point of view, they contain ASICs (Application spe-

ciﬁc integrated circuits) which are produced at low

cost in large numbers, and cannot be modiﬁed, which

raises an obstacle to the exploration of new designs.

In early experimentation, during the process of de-

signing wearable devices for new purposes, designers

wish to check out combinations of several sensors, ex-

plore the best (cheapest, most energy efﬁcient, most

secure) way of communicating between devices and

to the external world (data base, physician, ...) with-

out investing in –often non-reusable– material. Also,

the software is suject to frequent changes and the de-

signer should ideally be able to test it on an as generic

as possible platform, using standard processor cores.

Thus, the idea here is to validate simulations gen-

erated from system-level models against actual ma-

terial implementations. Using an adequate abstrac-

tion level, we propose a system-level "construction

kit" for designing wearable devices, based on combi-

nations of models for sensors, actuators, communica-

tion and control. We extend an existing SysML-based

modeling tool, which already contains analog design

features. From the models, we automatically gener-

ate virtual prototypes for both the digital and analog

part, interface them in order to perform full-system

co-simulation. We ﬁnally validate simulation results

against measurements taken from the physical proto-

types, built from electronic and textile parts.

We discuss related work in Section 2, basic con-

cepts are introduced in Section 3. Section 4 describes

the design space: we evaluate abstractions of sensors

wireless networks, and control. Section 5 shows a val-

idation for a wearable bio-monitoring device destined

for non-invasive daytime sleepiness detection.

2 RELATED WORK

In this paper, we bring together experience from

two domains: SysML-based virtual prototyping and

analog/mixed signal hardware design.

2.1 Model-based Design for

Cyber-physical Systems

UML/SysML based modeling techniques such as

MARTE (Demathieu et al., 2008) have been em-

Genius, D. and Chotin, R.

Model to Hardware: System-level Modeling for Wearable Devices.

DOI: 10.5220/0011005400003119

In Proceedings of the 10th International Conference on Model-Driven Engineering and Software Development (MODELSWARD 2022), pages 387-394

ISBN: 978-989-758-550-0; ISSN: 2184-4348

387

ployed to model cyber-physical systems (Selic and

Gérard, 2013).

(Fitzgerald et al., 2013) uses model-based formal

methods by integrating discrete-event models of con-

trollers with continuous-time models of their environ-

ments. Starting from an initial discrete-event model,

approximations of continuous-time behavior are sub-

sequently replaced by couplings to continuous-time

models.

Modelica (Fritzson and Engelson, 1998) is an

object-oriented modeling language for component-

oriented systems containing e.g. mechanical, electri-

cal, electronic and hydraulic components. Yet, since

time synchronization is not predeﬁned, the simulation

engine must manipulate objects in a symbolic way in

order to determine an execution order between com-

ponents of different MoCs.

Linking simulations with different Models of

Computation can be done by using e.g. the Functional

Mock-up Interface (Blochwitz et al., 2011), closely

related to the Modelica tools.

TTool (Apvrille, 2011), an open-source modeling

and veriﬁcation framework offering extensive formal

veriﬁcation features, has recently been extended with

features which allow analog/mixed signal modeling

and co-simulation (Genius et al., 2019). The exten-

sion is based on timed Synchronous Data Flow (SDF).

Another extension to the SDF formalism, called

Polygraph, which includes frequency constraints and

adjustable communication rates and ensures synchro-

nization, is shown in (Dubrulle et al., 2019).

2.2 Multi-domain Modeling

The following tools target analog/mixed signal or

multi-domain design and co-simulation. Models of

Computation (MoCs) are an important notion, as they

basically differ depending on whether modeling digi-

tal or analog behavior is intended. Thus, the question

of synchronization between two or more parts of a

simulation poses itself.

Ptolemy II (Ptolemy.org, 2014) is based upon the

data-ﬂow model. It addresses digital/analog systems

by deﬁning several sub domains, however leaving

time synchronization to the designer.

Metro II (Davare et al., 2007) is based on hier-

archical high level models. So-called Adapters are

used for data synchronization between components

belonging to different MoC, yet the model designer

has to implement time synchronization; a common

simulation kernel handles the entire execution.

Modelica (Fritzson and Engelson, 1998) is an

object-oriented modeling language for cyber-physical

systems. Time synchronization is not predeﬁned and

the simulation engine must manipulate objects con-

taining sets of equations in a symbolic way in order

to determine an execution order between components

of different MoCs.

Linking simulations with different Models of

Computation can also be done by using the Functional

Mock-up Interface (Blochwitz et al., 2011), which is

closely related to the Modelica tools.

3 BASICS OF SystemC-AMS

MODELING

TTool-AMS (Genius et al., 2018), which we use in

our work, is an extension of TTool which in particu-

lar allows the generation of virtual prototype from a

SysML-like representation. It generates a SystemC

speciﬁcation for the digital part together with soft-

ware and operating system for full-system simula-

tion, as well as a SystemC-AMS speciﬁcation from

the analog part of the virtual prototype, and runs both

parts together in a co-simulation.

SystemC (IEEE, 2011) is a collection of C++

classes, while SystemC-AMS (Vachoux et al., 2003)

is an extension based on SystemC (IEEE, 2011), pro-

viding analog and mixed-signal (AMS) features. A

proof-of-concept simulator has been developed (Ein-

wich, 2016).

In SystemC-AMS, digital components are de-

scribed with a Discrete Event (DE) MoC, while ana-

log components are described with the Timed Data

Flow (TDF) MoC, based on the timeless Synchronous

Data Flow semantics (Lee and Messerschmitt, 1987).

A TDF module has input and output ports and

a processing function, which describes the module’s

functionality. TDF clusters are composed of several

TDF modules which are connected by signals. TDF is

the most abstract MoC proposed in SystemC-AMS to

describe analog components, as continuous functions

are sampled at regular intervals.

4 METHOD

Figure 1 shows the design ﬂow and validation. Start-

ing from a functional model of the application, af-

ter partitioning, software (left hand side) and hard-

ware (right hand side) are designed in the SysML-like

representation proposed in (Pedroza et al., 2011) for

software; hardware is also described in a SysML-like

style (Genius and Apvrille, 2016).

A so-called deployment diagram, shown in the

center of the ﬁgure, describes the mapping of soft-

MODELSWARD 2022 - 10th International Conference on Model-Driven Engineering and Software Development

388

Virtual Prototype

Co-Simulation

Deployment

Hardware

design

Digital

Hardware

Model

Validation

of Results

Hardware

prototype

Analog

Hardware

Model

Software

design

A B Y

R= 1

D= 1

Tm= 6 ms Tm= 4 ms

Tp= 4 ms

R= 3

Tp= 2 ms

D= 0

R= 2

D= 0

Tp= 2 ms

TDF Cluster

HW/SW Partitioning

Figure 1: Integration in the Design Flow.

ware onto the hardware. From this diagram, the tool

generates a platform containing a top cell instantiating

hardware components from the SoCLib library (So-

cLib consortium, 2003), a lightweight operating sys-

tem and software in the form of Posix threads.

The analog part is generated in the form of

SystemC-AMS clusters (Accellera Systems Initiative,

2010) connected to the digital platform by speciﬁc in-

terfaces. This reﬂects the idea that analog parts are

essentially sensors, whose task it is to supply infor-

mation to the digital part, charged with control and

running the software. Both parts are co-simulated, re-

specting causality between MoCs (Porto et al., 2021),

as shown on the lower center of the ﬁgure. The lower

right of the ﬁgure shows the actual hardware pro-

totype implementation which is used to validate the

simulations on SystemC/SystemC-AMS level.

In order to reach our aim, a toolkit for fast design

of lightweight biomedical devices, required four con-

tributions that extend the existing toolkit.

The ﬁrst new contribution is an abstraction of sen-

sors: after analyzing common features of usual analog

modules, a library of SysML modules emerges, con-

taining parameterizable models not only of sensors

but also of ﬁlters, Digital/Analog and Analog/Digital

converters (DAC and ADC, respectively), and other

typical analog and mixed-signal modules.

The second contribution is the modeling of several

means of wireless communication in TTool-AMS.

The third contribution is the integration of (lim-

ited) software, destined to run on a microcontroller

but tested beforehand on a general purpose proces-

sor platform, to obtain an as-realistic-as-possible full-

system co-simulation.

The ﬁnal contribution is the validation against the

actual electronic prototype, consisting of microcon-

trollers, eventually microprocessors and FPGA (Field

Programmable Gate Arrays) for the digital part, net-

work components (WIFI, Bluetooth etc.), and sen-

sors.

The advantage of using SystemC-AMS in the vir-

tual prototype is that its interaction with the DE MoC

is well understood (Porto et al., 2021). Another ad-

vantage of using SystemC-AMS is that models can

be successively reﬁned: the processing function of a

TDF module could thus initially only contain the right

number and type of input and output ports, which

are only read on the input side and written on the

output side of the module, then adding a table de-

scribing the output of the sensor for the respective

input, ﬁnally using mathematical functions and trans-

formations (Laplace transfer functions are available in

SystemC-AMS).

4.1 Sensor Abstraction

Our ﬁrst aim is to ﬁnd an abstract representation for

sensors. Biomedical sensors are destined to capture

signals emitted by the human body. Figure 3 shows a

cluster containing a sensor –here a ﬂex sensor– with

ADC and clock.

All in all we determine the common characteris-

tics of ﬁve sensor types. Apart from alimentation,

they have one single output pin, providing analog data

in most cases, digital via an ADC in some. Data out-

put from the TDF cluster is a timed, synchronous ﬂow

of ﬂoating point (integer, bit vector) data correspond-

ing to the data types available for TDF ports.

The ﬂex sensor or ﬂeximeter captures the level of

bending by a variable resistor disposed throughout

its length. Accelerometers and gyroscopes are very

common in everyday life (joystick, smartphones, etc.)

Model to Hardware: System-level Modeling for Wearable Devices

389

and, as they name indicates, capture acceleration and

orientation wrt. a referential. An electromyogram

measures the electrical activity of nerves and muscle

contraction. Non-invasive versions, the only ones of

interest in our context, use electrodes glued to the skin

and usually contain signal ampliﬁers. The pulse sen-

sor, based on an infrared LED and a phototransistor

measures variations of the blood ﬂow. All sensors are

modeled as look up tables in the processing function,

as can be seen in the lowest part of Figure 2.

Figure 2: Abstraction of ﬂex sensor as TDF block: (top) pa-

rameters, (middle) attributes, (bottom) processing function.

4.2 Wireless Network Abstraction

Currently, we provide models of WIFI for mid/short

range (body to server) and bluetooth for very short

range (close to the body) communication. Details are

out of the scope of this paper, there are fourteen TDF

blocks forming a quite detailed, yet still abstracted

model. The Wiﬁ module contains nearly twice as

many (twenty-six) TDF blocks.

Figure 3: Model of a generic sensor.

4.3 Code Generation and Co-simulation

Software is abstracted by a SysML-like representa-

tion of communicating blocks with an underlying se-

mantics of communicating Finite State Machines (Pe-

droza et al., 2011). Blocks can contain additional

entry code to capture behavior that cannot be ex-

pressed in that representation, in particular the treat-

ment of (sampled) ﬂoating point values typically ar-

riving from the analog domain. A virtual cycle-bit-

accurate prototype for full-system simulation is gen-

erated automatically (top cell, operating system and

deployed software).

Analog clusters are instantiated together with the

digital components in a common top cell. This top

cell is generated by TTool-AMS, from an extended

Deployment Diagram (Genius et al., 2019).

During co-simulation, the SystemC event-based

simulation kernel controls the AMS simulation. The

difﬁculty here is synchronization; errors are usually

detected late, often only during simulation, causing

an emergency stop. Recent work proposes a method

to ensure correct synchronization of the DE and TDF

MoCs of SystemC-AMS which is directly applicable

at System Design level (Porto et al., 2021). Synchro-

nization is ensured by detecting causality problems at

converter ports and proposing adequate modiﬁcation

to the concerned TDF modules. General Purpose I/O

(GPIO) serve interfaces between the analog clusters

and the central interconnect of the digital platform.

4.4 Validation

We validate the resulting traces of the co-simulation

against an actual implementation using electronic and

mechanical parts, which is typically accomplished by

master’s students with off-the-shelf electronics com-

ponents (Celik, 2021; Broux et al., 2019) and hand-

crafted or off-the shelf mechanical/textile compo-

nents. The sensors, whose parameters are identical to

MODELSWARD 2022 - 10th International Conference on Model-Driven Engineering and Software Development

390

Figure 4: Wearable daytime sleepiness detector (prototype).

those used in the model, are connected to one or sev-

eral microcontrollers, which communicate with each

other and with a server running a small data base.

Figure 4 shows a prototype resulting from a recent

study, made of textile parts, in which electronics and

mechanical devices, wireless transmitter/receiver and

power supply can be hidden (Celik, 2021).

5 CASE STUDY: WEARABLE

DAYTIME SLEEPINESS

DETECTOR

Daytime sleepiness, possible cause of trafﬁc accidents

when occurring through driving, is caused by prob-

lems during nighttime sleep (snoring, ...). It is very

current and also concerns a population that, even if

often older and more overweight, is still active in pro-

fessional life.

In the domain of sleep disorder detection, usual

polysomnography devices are most often very expen-

sive and require overnight or several days’ hospital-

ization (Douglas et al., 1992). On the initiative of

the ICAN (cardiometabolics and nutrition) institute,

part of the Pitié-Salpêtrière University Hospital com-

plex, the issue was raised to develop a pre-diagnostic

device, to be worn during at least two or three days,

collected data then analyzed by physicians off-line.

Such a device requires a certain, if limited, bat-

tery autonomy. It should also be easy to use (er-

gonomic user interface, out of the scope of this pa-

per), lightweight and ideally insconspicious (mini-

mization) and, if wireless data transmission is chosen,

satisfy security requirements (out of the scope of the

current paper but supported by TTool).

A wearable daytime sleepiness detection equip-

ment stemming from several master student’s projects

(Broux et al., 2019; Celik, 2021) serves as our case

study. It was decided beforehand that, in order to be

inconspicious, the device consists of two parts hidden

in the clothing: a headgear and a bracelet.

Figure 5: Overview of the client-server architecture.

We wish to evaluate combinations of different sen-

sors. The following sensors have been employed in

our experimentation:

• A ﬂex sensor FS7954 to monitor the bending an-

gle of the neck. It measures 7.4 cm and requires

0.5 Watts and 3.3 V thus 150 mA.

• A gyroscope/accelerometer combination Grove

101020584 which detects sudden movements

(head falling brusquely forward at onset of sleep).

The tension is 3.3V for the aaccelerometer which

consumes 150µ A and the gyroscope consumes 5

mA. This sensor has an integrated ADC.

• An EMG-Sensor SEN0240 (electromyograph) to

monitor muscle tension supplied by 3.3V and

minimum 20mA.

• A pulse sensor KY039HS used to determine the

heart rate. Its received signal varies in function

of the blood ﬂow present in the ﬁnger, i.e. the

contraction and decontraction of the heart. Each

amplitude variation between two extreme values

corresponds to a heart beat.

Only the last two are exclusively used in biomedical

sensing. With exception of the last one, all sensors

are located in the upper part of the device, in prox-

imity of the neck. The generic sensor discussed in 4

and shown in Figure 3 can be used as part of this de-

vice, parameterized to represent each of the ﬁve sen-

sors. The combination of sensors allows to detect in-

dications for daytime sleepiness: for example whether

a movement of the head is involuntary, for example

when accompanied by a slowdown of the heart rate.

Three ESP-32 micro-controllers implement con-

trol and communication. Each micro-controller is

connected to two sensors (ﬂex sensor and gyro-

scope/accelerometer in the neck area, EMG and pulse

sensor on arm/hand).

Model to Hardware: System-level Modeling for Wearable Devices

391

<<CPU>>

CPU0

Design::Application

<<VGMN>>

Interconnect

<<SystemC-AMS Cluster>>

WIFI_receiver

<<RAM>>

Memory0

<<TTY>>

TTY0

<<SystemC-AMS Cluster>>

sensor

<<SystemC-AMS Cluster>>

bluetooth

<<CPU>>

CPU0

Application

<<VGMN>>

Interconnect

<<SystemC-AMS Cluster>>

WIFI_receiver

<<RAM>>

Memory0

<<TTY>>

TTY0

<<SystemC-AMS Cluster>>

sensor

<<SystemC-AMS Cluster>>

bluetooth

Figure 6: TTool Deployment Diagram for the wearable daytime sleepiness detector.

Bluetooth is used for the communication between

the different parts of the device ﬁxed to the body

(short distance), ﬁnally WIFI for communicating with

the server (mid/short distance). We represent WIFI

and Bluetooth, ensuring the communication, modeled

as TDF clusters.

The server and software (among others a simple

database SQLite3) are simulated by a SoCLib mono-

processor platform. Figure 5 gives an overview: each

sensor pair worn by the patient is equipped with blue-

tooth and WIFI and an optinal ADC. Figure 6 shows

the deployment of software (Application task) on the

processor, and the three analog clusters attached.

The user interface for co-simulation is currently

very basic, essentially textual output on a terminal.

Traces are stored in a text ﬁle, and compared to mea-

surements taken from the electronic prototype by a

Python script. Figure 7 shows some results for one of

the sensors, a ﬂex sensor: output tension in mV (Y

axis) in function of time in seconds (X axis): at 150

seconds, a brusque movement is detected.

Figure 8 is a screenshot of the co-simulation: the

upper part shows the instantiation of the top cell com-

ponents of the SystemC part of the platform (proces-

sor with cache, TTY, memory segments, interconnect,

three GPIO interfaces), the lower part the SystemC-

AMS simulation (three clusters of different complex-

ity and number of modules: WIFI, by far the most

complex with more than 20 TDF modules, an exam-

ple sensor, and bluetooth).

The current prototype of the sleepiness detection

device, already shown earlier in Figure 4, is integrated

in a cap and a bracelet. An alternative version, less

targeted towards mid-age male patients, would inte-

grate the sensors monitoring the neck (ﬂex sensor, gy-

roscope/accelerometer) into a scarf. The most expen-

Figure 7: First results for the ﬂex sensor.

sive, and most heavy, part, are the batteries, which

must maintain alimentation during two or three days:

we require 2 batteries providing 9.600mAh. Two pos-

sibilities are foreseen for data transfer to the server:

either by SIM card or by WIFI, and fed into the

database. The cost of the prototype is less than 100$.

6 FUTURE WORK

The current paper presents work in progress. Co-

simulation results shown here stem from a device in-

cluding a single sensor cluster, while the current elec-

tronic prototype already contains up to three clusters

carrying ﬁve sensors. It is possible to reﬁne measure-

ments by adding more sensors. Nevertheless, the sim-

ulated results are close to those obtained with the ac-

tual implementation, which is clearly encouraging.

The students tested the device by wearing it them-

selves during several hours; due to longer-term legal

MODELSWARD 2022 - 10th International Conference on Model-Driven Engineering and Software Development

392

Figure 8: Co-simulation.

Model to Hardware: System-level Modeling for Wearable Devices

393

questions and the sanitary situation, currently larger-

scale tests were not yet possible.

Currently the AMS part is restricted to TDF,

which is a rather coarse abstraction of analog behav-

ior. We plan to reﬁne analog models by integrating

ELN (Electrical Linear Network), another MoC avail-

able in SystemC-AMS.

Most crucial is the optimization of battery usage;

power consumption estimation should be added to the

tool. Finally, even if the method was developed in a

speciﬁc domain, it might be generalized to other IoT

devices equipped with multiple sensors.

REFERENCES

Accellera Systems Initiative (2010). SystemC AMS exten-

sions Users Guide, Version 1.0.

Apvrille, L. (2011). TTool, an open-source toolkit for the

modeling and veriﬁcation of embedded systems.

Blochwitz, T. et al. (2011). The functional mockup interface

for tool independent exchange of simulation models.

In 8th Int. Modelica Conference, Dresden, Germany,

pages 105–114.

Broux, P.-E., Lavalade, P., Sylla, S., and Zhuang, K. (2019).

Somnolence diurne sodi.

Celik, H. O. (2021). Minimization, deployment and evalua-

tion with ai methods of a portable sleepyness detection

device, master’s thesis (French), Sorbonne Université.

Davare, A., Densmore, D., Meyerowitz, T., Pinto, A.,

Sangiovanni-Vincentelli, A., Yang, G., Zeng, H., and

Zhu, Q. (2007). A next-generation design framework

for platform-based design. In DVCon, volume 152.

Demathieu, S., Thomas, F., André, C., Gérard, S., and Ter-

rier, F. (2008). First experiments using the uml proﬁle

for marte. pages 50–57. IEEE.

Dey, N., Ashour, A. S., Shi, F., Fong, S. J., and Tavares,

J. M. R. (2018). Medical cyber-physical systems: A

survey. J. of medical syst., 42(4):74.

Douglas, N. J., Thomas, S., and Jan, M. A. (1992).

Clinical value of polysomnography. The Lancet,

339(8789):347–350.

Dubrulle, P., Gaston, C., Kosmatov, N., Lapitre, A., and

Louise, S. (2019). A data ﬂow model with frequency

arithmetic. In International Conference on Funda-

mental Approaches to Software Engineering, pages

369–385. Springer, Cham.

Einwich, K. (2016). SystemC AMS PoC2.1 Library,

COSEDA, Dresden.

Fitzgerald, J. S., Larsen, P. G., Pierce, K. G., and Verhoef,

M. H. G. (2013). A formal approach to collabora-

tive modelling and co-simulation for embedded sys-

tems. Mathematical Structures in Computer Science,

23(4):726–750.

Fritzson, P. and Engelson, V. (1998). Modelica—a uni-

ﬁed object-oriented language for system modeling

and simulation. In European Conference on Object-

Oriented Programming, pages 67–90. Springer.

Genius, D. and Apvrille, L. (2016). Virtual yet precise pro-

totyping: An automotive case study. In ERTSS’2016,

Toulouse.

Genius, D., Cortés Porto, R., Apvrille, L., and Pêcheux,

F. (2019). A tool for high-level modeling of

analog/mixed signal embedded systems. In MODEL-

SWARD.

Genius, D., Cortés Porto, R., and Apvrille, L. (2018). TTool

AMS.

IEEE (2011). SystemC. IEEE Standard 1666-2011.

Lee, E. A. and Messerschmitt, D. G. (1987). Synchronous

data ﬂow. Proceedings of the IEEE, 75(9):1235–1245.

Pedroza, G., Knorreck, D., and Apvrille, L. (2011).

AVATAR: A SysML environment for the formal veri-

ﬁcation of safety and security properties. In 11th IEEE

Conference on Distributed Systems and New Tech-

nologies, Paris, France.

Porto, R. C., Genius, D., and Apvrille, L. (2021). Handling

causality and schedulability when designing and pro-

totyping cyber-physical systems. Software and Sys-

tems Modeling, pages 1–17.

Ptolemy.org, editor (2014). System Design, Modeling, and

Simulation using Ptolemy II.

Selic, B. and Gérard, S. (2013). Modeling and Analysis

of Real-Time and Embedded Systems with UML and

MARTE: Developing Cyber-Physical Systems. Else-

vier.

Silva, A. F. and Tavakoli, M. (2020). Domiciliary hospital-

ization through wearable biomonitoring patches: Re-

cent advances, technical challenges, and the relation

to covid-19. Sensors, 20(23):6835.

SocLib consortium (2003). The SoCLib project: An In-

tegrated System-on-Chip Modelling and Simulation

Platform, www.soclib.fr.

Vachoux, A., Grimm, C., and Einwich, K. (2003). Analog

and mixed signal modelling with SystemC-AMS. In

ISCAS (3), pages 914–917. IEEE.

MODELSWARD 2022 - 10th International Conference on Model-Driven Engineering and Software Development

394