MODELING APPROACH FOR NETWORKED EMBEDDED
SYSTEMS WITH HETEROGENEOUS COMMUNICATION
Modeling Common Gateway Functionalities for Interconnection
Johannes Kl¨ockner, Marcus M¨uller, Wolfgang Fengler and Yixin Huang
Computer Architecture Group, Ilmenau University of Technology, P.O. Box 100565, Ilmenau, Germany
Keywords:
Model based design, Fieldbus, FlexRay, CAN, Building blocks, Network simulation, MLDesigner, Gateway.
Abstract:
This paper presents a method to create system level models of gateway facilities to combine embedded field-
bus networks. Here automotive applications are used as representative with CAN and FlexRay. A strategy
is introduced to create gateway models, which provide high exibility in terms of reusability, replaceability,
extensibility and flexibility. Todays systems often incorporate a very complex heterogeneous communication
structure with real time requirements. These characteristics have an influence on the system design process.
The goal of avoiding design errors in early development stages requires the validation on hierarchical func-
tional models of various abstraction levels using simulation based analysis prior to hardware design. A current
approach includes problem oriented system analysis. This approach is extended to analyze the system behav-
ior of heterogeneous embedded systems focusing on switching mechanisms. Here lies the challenge on the
connection of subsystems with distinct fieldbuses realizing different paradigms. Regarding real time aspects
paradigm switches can be a source of errors in the system design.
1 INTRODUCTION
As a result of the increase in both system complexity
and the amount of incorporated functionality a single
embedded system has to be considered as a compo-
sition of several heterogeneous subsystems. Due to
this the intra-system networking of an embedded sys-
tem provides new challenges in system design. De-
veloping such complex systems requires a lot of plan-
ning and design decisions on a profound knowledge
of system behavior. The heterogeneity in a compound
system stems from differences in hardware, software
and subsystem architectures. In networked embedded
systems the communication is an important aspect.
Based on the complexity of systems and distributed
applications the amount of communication between
individual subsystems grows. Using communication
technologies with different paradigms, e.g. event or
time triggered, influence the system behavior.
The heterogeneity of large systems represents a
great challenge in the system design. This high grade
of variability complicates the development process.
Mistakes in early system design stages can negatively
influence the performance and development costs.
Avoiding these errors requires the validation on func-
tional models of various abstraction levels using sim-
ulation based analysis prior to hardware design.
To analyze the performance on system level
(Henia et al., 2005) use an approach based on a
scheduling analysis. A model based design approach
(Salzwedel, 2004) enables an efficient top down de-
velopment of a complete system using hierarchically
composed building blocks, thus providing a high
grade of reusability and exchangeability.
Aim of the presented work is the extension of an
existing modeling approach studying networked em-
bedded systems to allow the modeling of heteroge-
neous networked systems in a common way and the
analysis of system behavior. It is based on the tool
MLDesigner (MLDesign Technologies Inc., 2007)
and fulfills the requirements of a model based de-
sign approach allowing system analysis and devel-
opment. An example for a complex heterogeneous
networked embedded system is an automobile. It
is a system with a large amount of functionality,
real time requirements and distributed characteris-
tics. The complex communication infrastructure con-
tains several protocols with different paradigms, e.g.
FlexRay (FlexRay Consortium, 2005) and Controller
Area Network (CAN). The mentioned properties can
be found in a lot of systems belonging to areas en-
gaged with automation and control. This paper is
230
Klöckner J., Müller M., Fengler W. and Huang Y. (2009).
MODELING APPROACH FOR NETWORKED EMBEDDED SYSTEMS WITH HETEROGENEOUS COMMUNICATION - Modeling Common Gateway
Functionalities for Interconnection.
In Proceedings of the 6th International Conference on Informatics in Control, Automation and Robotics - Signal Processing, Systems Modeling and
Control, pages 230-233
DOI: 10.5220/0002207902300233
Copyright
c
SciTePress
organized as follows. Section 2 gives a short descrip-
tion of the used modeling tool. Section 3 introduces
related work and the basic modeling strategy. Section
4 describes the modeling concept. Section 5 presents
the drawn conclusions and a brief overview of further
development steps.
2 MODELING TOOL
Currently many tools exist, that realize model-based
design to support the development process. In this
work the tool MLDesigner by MLDesign Technolo-
gies, Inc. is used. The tool is dedicated to im-
proving the design process from early concepts to
implementation with mission and system level de-
sign. MLDesigner extends the Ptolemy project of
UC Berkeley (The Ptolemy Project, 2009) with mod-
eling paradigms. Different models of computation
so called domains are provided, e.g. discrete event
domain (DE) and finite state machines (FSM), syn-
chronous data flow domain (SDF). In this work the
discrete event domain and finite state machines are
used. These two are well suited to modeling net-
worked embedded systems.
Now a short introduction to the terms of MLDe-
signer. The System is the top level element in the
modeling hierarchy. The building blocks can be
atomic blocks called Primitives, specified as FSM or
in C/C++ code, or hierarchical Modules. To commu-
nicate with their environment building blocks can use
linked variables or ports. Ports are represented as ar-
rows on the bounding box of a building block and are
interconnected by signal paths. So called Wormholes
allow the embedding of building blocks belonging to
different domains.
3 RELATED WORK
In (Kl¨ockner et al., 2008) a modeling strategy for net-
worked embedded systems is discussed exemplified
by a top-down developed generalized FlexRay proto-
col model. The communication system is defined as a
composition of three different elements: Host, Com-
munication Controller (CC) and Channel as shown
in Figure 1. The combination of a host and a CC is
called node. The corresponding protocol can be im-
plemented within the CC, e.g. synchronization, error
detection, message transmission and reception. In ad-
dition, the CC provides several services to the host
to be interfaced by its application. The application
of a node can be described within the host. The host
Figure 1: Basic Model Structure for networked embedded
systems (Kl¨ockner et al., 2008).
uses CC provided services to configure the CC, initi-
ate the sending and receiving operations and process
the received data. According to this the host contains
a specialized sublayer realizing the protocol related
access to the CC. At this point the division into differ-
ent functional layers is visible, the host describes the
functionality and the CC describes the type of com-
munication. Nodes are grouped to communication
clusters by connecting them to a channel.
The physical characteristics of the connection be-
tween nodes is described and modeled within a chan-
nel. On this level the physical delay of a transmission
is determined by the respective message data length
and a fault injection model simulates the transmission
errors caused by physical medium and environmental
influences.
The focus of the approach (Kl¨ockner et al., 2008)
lies on homogeneous communication systems. The
model structure intends the extension of the approach
to allow the modeling of heterogeneous systems.
4 GATEWAY MODEL
In order to provide concepts to monitor both the be-
havior of networked systems and the communica-
tion across different types of networks with differ-
ent specifications and protocols, a concept for a gate-
way model which provides high flexibility in terms of
reusability, replaceability, extensibility and flexibility
to connect different communication technologies in a
common way, is needed. Based on the model struc-
tures presented in (Kl¨ockner et al., 2008) (M¨uller,
2007) and the libraries developed in these works, the
creation of a customizable gateway module is possi-
ble. This allows the combination of heterogeneous
subsystems in order to analyze the overall system be-
havior. Regarding the system architecture presented
in Figure 1 a gateway is a special type of node. Nor-
mally a node is a combination of a host and a CC. A
node realizing gateway functionality is now a combi-
nation of several hosts, associated CCs and a Gate-
way Core. The basic gateway structure is shown in
Figure 2. The CCs contain the identical functionality
as described before. A common strategy to model an
MODELING APPROACH FOR NETWORKED EMBEDDED SYSTEMS WITH HETEROGENEOUS
COMMUNICATION - Modeling Common Gateway Functionalities for Interconnection
231
CC
Host
Host
(CC-Access)
Host
(Gateway Core)
Channel
(Network 1)
Channel
(Network 2)
Node_GW
Node_NW2Node_NW1
CC CC CC
Host
Host
(CC-Access)
Figure 2: Extended Model Structure for networked embed-
ded systems.
application is not yet designed. Therefore the hosts
realize the specific sublayers to access the CCs. The
Gateway Core can be seen as an application of the
node providing the gateway base function. This con-
cept fits the described modeling strategy. Anticipating
a more detailed model containing different applica-
tions in a node divided in several tasks it is possible
to extend this architecture to enable the analysis of the
system behavior including other effects, e.g. proper-
ties of task scheduling.
Gateway facilities handle the data exchange be-
tween heterogeneous computer networks. Normally,
they are designed to combine specified networks, and
fulfill specified needs, which are based on the appli-
cation. Hence, there is no predefined specification of
a gateway. Although there are no fixed design ap-
proaches for gateway design, certain common func-
tions can still be assumed for any specialized com-
munications gateway: routing, protocol conversion,
dataflow control, real time constraint checking for
message transmission and message filtering or seg-
mentation. The purpose of this paper is to develop
a universal gateway model, which facilitates the com-
munications between different networks. The design
of the presented approach is similar to the idea of a
time triggered gateway described in (Shaheen et al.,
2007).
As described in (Fahmy, 1995), the gateway
model is based on a shared medium approach, which
uses an internal network to exchange uncommitted
messages between Host and Gateway Core. Received
frames are converted to an uncommitted protocol, af-
ter which the information is handled by the gateway
module. Finally the information is converted to match
the destination protocol. The uncommitted protocol
contains three values: Source Identifier, Destination
Identifier and Datafield. The routing information of
each message can be identified by Source Identifier
and Destination Identifier. The data of the received
message is stored in the Datafield of the uncommit-
ted message. To fulfill the flexibility requirement the
gateway model is a composition of two kinds of mod-
ules: protocol related and unrelated modules. Only
uncommitted gateway messages can be processed in-
side the protocol unrelated modules, and the proto-
col related modules process the specific protocol mes-
sages. In order to extend the gateway model to sup-
port a larger amount of protocols, only the protocol
related modules need to be created, while the proto-
col unrelated modules remain unchanged. This ap-
proach provides enormous reusability, replace ability
and extensibility. It allows an easy generation of a
customized gateway to realize relevant system char-
acteristics.
An important aspect of the gateway functionality
refers to (H¨orner, 2007). A gateway should provide
at least two methods of data exchange - PDU based
and signal based gateway functions. A PDU gate-
way routes the protocol data units (PDUs) unchanged
between two networks. In this case the data carried
on both networks, source and destination, are identi-
cal regarding content and length. It is also possible,
that only signals, which are contained in the received
PDUs from source network, are needed on the other
network. In this case, the gateway does not trans-
fer the entire PDU, but sends the individual signals
to the corresponding destination network. To achieve
this, a single received PDU is disassembled in signals
according to the specification of the source network.
Afterwards signals are grouped together congruent to
configuration of the destination network and assem-
bled to PDUs, which are send across the destination
network.
The central unit of the gateway model is the mod-
ule Gateway Core shown in Figure 3. This module
contains protocol unrelated functions, e.g. buffering
of received messages, routing, message segmentation
and disassembling messages into signals. Based on
the specification of function blocks with standardized
interfaces a high adaptability of the architecture is
achieved allowing the extension and substitution of
single functions. Therefore it is possible to analyze
the system performance of several realizations differ-
ing in their buffer strategy. In contrast to the gateway
TransmitMessage
TransmitSignal
SplitFrameToSignal
SendMsg
RoutingUnitDelay
Delay
FIFOQueue
CoreControlUnit
CoreSwitch
Init_Gateway
Trigger
FromHost CtrlMsgInput
OverFlow
ToHost
Figure 3: Module Gateway Core.
core, the protocol related gateway functions, such as
protocol conversion, initialization of the communica-
ICINCO 2009 - 6th International Conference on Informatics in Control, Automation and Robotics
232
tion controller, message transmission and message re-
ception, etc. are realized within the respective Host
module. The host connected to the source network
receives a message via the CC and converts it into
an uncommitted message. Afterwards the message is
sent to the gateway core. Inside the gateway core, the
uncommitted message will be processed and routed to
the destination host. There it is converted into the cor-
responding standard. Each host has a distinct identi-
fier. This allows the connection of an unlimited num-
ber of networks to the gateway model.
The example scenario shown in Figure 4 contains
two CAN and two FlexRay networks which are con-
nected by a gateway. The system itself is an ex-
tension of an application presented in (Hedenetz and
Belschner, 1998). Each network contains two or more
nodes. The behavior of the used modules CCs realizes
the connection to CAN and FlexRay. The analysis of
the simulation results showed the correct behavior of
the created model.
Brake1 Brake2
Brake3
Brake4
BBWM1PM1
Cluster 1
PM2BBWM2
Cluster 2
Gateway
Port 1
Port 2
Port 3
Port 4
Cluster 3
CAN F
CAN E
CAN D
CAN C
CAN B
CAN A
Cluster 4
Figure 4: System used as reference for validation.
5 CONCLUSIONS
The modeling strategy and developed common gate-
way model presented in this paper allow the universal
design and analysis of heterogeneous communication
networks. Modularized components in the gateway li-
brary represent available functions of a gateway facil-
ity and allow an easy construction of a wide range of
systems. The basic model concept behind the gateway
model enables the analysis of communication and be-
havior of heterogeneous networked systems and sup-
ports the design process of such systems. In early
stages of the development process the evaluation and
verification of system properties can be provided prior
to hardware design.
Up to now there are only few communication pro-
tocol models available. To extend the ability for sys-
tem modeling it is necessary to extend the stock of
models. At present this limits the systems which can
be analyzed by the presented approach to CAN and
FlexRay systems. To gain access to real world exam-
ples from within the automotive application domain
the design of models realizing MOST (Media Ori-
ented Systems Transport) and LIN (Local Intercon-
nect Network) are required. Future work will also
deal with the automated import of gateway routing
information by using e.g. CAN Database and FIBEX
(ASAM, 2007). This can be supplemented by con-
cepts of automated model generation.
REFERENCES
ASAM (2007). FIBEX - Field Bus Exchange Format Ver-
sion 2.0.1.
Fahmy, S. (1995). A survey of ATM switching techniques.
CIS-788-95 Semester Report, Ohio State University,
CIS Dep.
FlexRay Consortium (2005). FlexRay Communications
Systems - Protocol Specification Version 2.1.
Hedenetz, B. and Belschner, R. (1998). Brake-by-
wire without mechanical backup by using a ttp-
communication network. page 2.
Henia, R., Hamann, A., Jersak, M., Racu, R., Richter, K.,
and Ernst, R. (2005). System level performance anal-
ysis - the symta/s approach. In IEE Proceedings Com-
puters and Digital Techniques.
H¨orner, H. (2007). Das universelle Gateway-Steuerger¨at.
Elektronik Automotive, Sonderausgabe AUTOSAR,
pages 33–35.
Kl¨ockner, J., K¨ohler, S., and Fengler, W. (2008). Model
based design of networked embedded systems. In
ICINCO-SPSMC, pages 253–259. INSTICC Press.
MLDesign Technologies Inc. (2007). MLDesigner Docu-
mentation, Version 2.7. http://www.mldesigner.com/.
M¨uller, M. (2007). Dokumentation framebasiertes can-
modell.
Salzwedel, H. (2004). Design technology development to-
wards mission level design. In 49. Internationales
Wissenschaftliches Kolloquium IWK’2004.
Shaheen, S., Heffernan, D., and Leen, G. (2007). A gate-
way for time-triggered control networks. Micropro-
cess. Microsyst., 31(1):38–50.
The Ptolemy Project (2009). http://ptolemy.eecs.berke-
ley.edu/.
MODELING APPROACH FOR NETWORKED EMBEDDED SYSTEMS WITH HETEROGENEOUS
COMMUNICATION - Modeling Common Gateway Functionalities for Interconnection
233