ArtistDesign NoE

45. Report

Giorgio Buttazzo — 2010-12-15T22:07:55Z

1. Introduction

The “Graduate Course on Embedded Real-Time Control Systems: Theory and Practice” (called the Eurolab10 in the following) was organized within the ArtistDesign NoE (cluster on Operating Systems and Networks). The purpose of this course was twofold:

Introducing the most relevant concepts and methodologies used to develop a real-time embedded system, including fundamentals of real-time scheduling, control and distributed systems;
Showing how to apply these concepts in practice, using an embedded platform and a real-time operating system to develop simple control applications and make experience with wireless sensor networks.

The course was hosted by the Laboratory on Real-Time and Embedded Systems (RETIS Lab) of the Scuola Superiore Sant'Anna, located in Pisa (Italy), from June 14 to June 18, 2010.

Organizing Committee
Giorgio Buttazzo (Scuola Superiore Sant'Anna, Italy)
Ettore Ricciardi (ISTI-CNR, Pisa)

Teachers
Giorgio Buttazzo (Scuola Superiore Sant'Anna, Pisa, Italy)
Paolo Gai (Evidence, Italy)
Pau Marti (University of Catalonia, Spain) - Manel Velasco (University of Catalonia, Spain) - Mauro Marinoni (Scuola Superiore Sant'Anna, Pisa, Italy) - Gianluca Franchino (Scuola Superiore Sant'Anna, Pisa, Italy)

Technical Assistants
Mauro Marinoni (Scuola Superiore Sant'Anna, Pisa, Italy) - Gianluca Franchino (Scuola Superiore Sant'Anna, Pisa, Italy) - Francesco Prosperi (Scuola Superiore Sant'Anna, Pisa, Italy)

2. Course structure

The school was structured as a five day intensive course with lectures and laboratory exercises. For the first four days, mornings were dedicated to theory and afternoons to practical experience. The last day was entirely dedicated to practice. The goal of the lectures was to introduce the fundamentals on real-time systems in scheduling, control, and networks, and to provide a detailed presentation of the Erika operating system, used for the practical laboratory part. Each day was dedicated to a specific topic:

The first day introduced the basic principles of real-time computing and illustrated the most significant results on real-time scheduling and resource management.
The second day was entirely devoted to the Erika kernel, to enable participants to quickly write a simple real-time demo, using the methodologies they learnt in the previous day. During the second day, 15 possible projects were presented by the teachers to the participants, who divided in groups of 2-3 people and selected a project to develop in the next days.
The third day was focused on real-time control and explained how to design control applications taking timing constraints into account, and how to use control techniques to make real-time systems more adaptive to dynamic changes.
he forth day was dedicated to real-time networks and addressed the problems of synchronization and medium access control that are encountered in distributed embedded systems, together with an analysis of end-to-end latencies.
The fifth day was entirely dedicated to practical experience and implementation. The participants interacted with the teachers and lab assistants to propose a real-time control application, preferably distributed, either in a simulated environment or for controlling one of the platforms that will be provided by the teachers.

3. Practical laboratory experience

The practical part of the course required the participants to develop a real-time control application and exchange relevant data using a wireless communication channel, using the embedded platform provided by the school.

Each control application had to be developed under the Erika operating system, using a number of concurrent threads, with periodic and aperiodic nature, interacting through shared buffers or message passing communication channels. The low level code for accessing the hardware was already provided as a number of library functions, so the participants had to develop only the code for sensory data processing, control and wireless communication.

The participants had the unique opportunity to use and test different advanced kernel mechanisms and several types of communication mechanisms.

4. Participants

The course hosted 32 participants, from universities, research centers and industries, coming from 10 different countries:

5. Diploma

At the end of school, each participant received a diploma certifying his/her attendance to the course. A final exam was available for students that wanted to acquire 3 credits, according to the European Credit Transfer and Accumulation System (ECTS).

55. Participants

Giorgio Buttazzo — 2010-12-01T18:04:51Z

No	Name	Affiliation
1	Besem Abid	SSSA (ITA)
2	Khurram Ali	Polito (ITA)
3	Fredrik Asplund	KTH (SWE)
4	Fedor Babkin	Centric TSolve GmbH (DEU)
5	Laura Barros Bastante	U.Cantabria (ESP)
6	Sagar Behere	KTH (SWE)
7	Simone Brienza	Unipi (ITA)
8	Shahzad Butt	Polito (ITA)
9	Asma Charfi	CEA LIST (FRA)
10	Andrea Claudi	UP Marche (ITA)
11	Cesar Cuevas Cuesta	Univ. of Cantabria (ESP)
12	Angela del Barrio Fernandez	Univ. of Cantabria (ESP)
13	Sena Ergullu	Sabanci University (TUR)
14	Francesco Esposito	SSSA (ITA)
15	Egidio Falotico	SSSA (ITA)
16	Sergi Fernandez Fort	UP Catalunya (ESP)
17	Evangelos Georgiou	Kings College London (GBR)
18	Hashem Ghazzawi	University of York (GBR)
19	Yasir Javed	RTRACKP (SAU)
20	Sarmad Khan	Polito (ITA)
21	Tiong Hoo Lim	University of York (GBR)
22	Carlo Lombardi	Microchip (ITA)
23	Giulio Mancuso	SSSA (ITA)
24	Nicola Manica	UniTN (ITA)
25	John Mogo	KU Leuven (BEL)
26	Matteo Moretti	UP Marche (ITA)
27	Niels Penneman	Ghent Univ (BEL)
28	Antonio Romano	SSSA (ITA)
29	Stefano Sestini	UNIFI (ITA)
30	Martijn van den Heuvel	TU Eindhoven (NLD)
31	Gregorio Vettori	UNIFI (ITA)
32	Davide Zambrano	SSSA (ITA)

20. Programme

Giorgio Buttazzo — 2010-02-24T16:17:47Z

MONDAY (slides)

8:30	Welcome Giorgio Buttazzo, SSSA
9:30	Introduction to embedded real-time systems Giorgio Buttazzo, SSSA
10:30	coffee break
11:00	Overview of the Flex platform Paolo Gai, Evidence
12:00	The OSEK standard Paolo Gai, Evidence
13:00	lunch
14:00	The Erika kernel: tasks, resources Paolo Gai, Evidence
15:00	The Erika kernel: events, alarms Paolo Gai, Evidence
16:00	coffee break
16:30	Example on the Erika kernel Paolo Gai, Evidence
17:30	Example on the Erika kernel Paolo Gai, Evidence

TUESDAY (slides)

8:30	dsPic architecture Mauro Marinoni, SSSA
9:30	Details on the Flex board Mauro Marinoni, SSSA
10:30	coffee break
11:00	Lab. practice on Flex board Mauro Marinoni, SSSA
12:00	Lab. practice on Flex board Mauro Marinoni, SSSA
13:00	lunch
14:00	Introduction to wireless communication Gianluca Franchino, SSSA
15:00	Wireless communication protocols Gianluca Franchino, SSSA
16:00	coffee break
16:30	Lab. practice on wireless communication Gianluca Franchino, SSSA
17:30	Lab. practice on wireless communication Gianluca Franchino, SSSA

WEDNESDAY (slides)

8:30	Introduction to Control Pau Martì, Manel Velasco, UPC
9:30	Integrated control and scheduling Pau Martì, Manel Velasco, UPC
10:30	coffee break
11:00	Feedback scheduling of control systems Pau Martì, Manel Velasco, UPC
12:00	Event-driven control systems Pau Martì, Manel Velasco, UPC
13:00	lunch
14:30	Lab. practice on control (with Flex and Erika) Pau Martì, Manel Velasco, UPC
15:30	Lab. practice on control (with Flex and Erika) Pau Martì, Manel Velasco, UPC
16:30	Lab. practice on distributed control Gianluca Franchino, SSSA
17:30	Lab. practice on distributed control Gianluca Franchino, SSSA
19:00	buffet
21:00	Luminara (Candellight Feast) in Pisa

THURSDAY (slides)

11:00	Scilab/Scicos code generation for embedded systems
12:00	Lab. practice on code generation
13:00	lunch
14:00	Lab. practice
15:00	Lab. practice
16:00	coffee break
16:30	Lab. practice
17:30	Lab. practice

FRIDAY (slides)

8:30	Lab. practice (or final exam)
9:30	Lab. practice (or final exam)
10:30	coffee break
11:00	Project presentations
12:00	Evaluations
13:00	lunch

90. Participants

Giorgio Buttazzo — 2010-02-21T18:33:09Z

No	Name	Affiliation
1	Tesnim Abdellatif	Verimag
2	Simon Aittamaa	LTU
3	Gaetano Anastasi	SSSA
4	Claudia Barberis	PoliTo
5	Imene Ben Hafaiedh	Verimag
6	Irene Bicchierai	DSI
7	Torsten Bruns	University Paderborn
8	Laura Carnevali	DSI
9	Jacques Combaz	Verimag
10	Johan Eriksson	LTU
11	Francesco Esposito	SSSA
12	Dario Faggioli	SSSA
13	Cyril Faure	IFP
14	Stefano Fontanelli	SSSA
15	Nadereh Hatami	PoliTo
16	Benjamin Kuch	SSSA
17	Rom Langerak	University of Twente
18	Holger Nahrstaedt	TU Berlin
19	Christian Nastasi	SSSA
20	Moritz Neukirchner	TU Braunschweig
21	Simon Perathoner	ETH Zurich
22	Francesco Prosperi	SSSA
23	Sophie Quinton	Verimag
24	Lorenzo Ridi	DSI
25	Antonio Romano	SSSA
26	Andreas Schranzhofer	ETH Zurich
27	Vassiliki Sfyrla	Verimag
28	Colin Shaw	BCIT
29	Gang Yao	SSSA

A Common Infrastructure for Adaptive Real-Time Systems

Giorgio Buttazzo — 2006-10-10T13:25:26Z

Activity Leader: Giorgio Buttazzo (Scuola Superiore S. Anna)

Artist2 Clusters:
Adaptive Real-Time

Contents
Baseline
Problem Tackled in Year2
Previous Work
Current Results
Keynotes, Workshops, Tutorials
Related Publications
Participants

Abstract
The objective of this activity is to show how current operating systems and network protocols have to be extended to support emerging real-time applications that exhibit a high degree of complexity and operate in dynamic environments.

The impact on operating system standards (like RT-POSIX and OSEK) as well as network protocols will also be taken into account.

Baseline

At the beginning of the activity, most of the partners were not familiar with the Shark operating system and could not exploit the potentiality of the kernel in developing software applications using advanced real-time mechanisms, such as dynamic scheduling, proirity inheritance, resource reservation, aperiodic servers, and overload management algorithms. Moreover, they could not practically evaluate the impact of a specific kernel mechanism on the performance of a dynamic real-time application, simply because the existing kernels usually provide fixed policies that cannot be easily replaced or modified by the user.

However, each partner had a solid theoretical expertize in real-time systems and also some practical experience in one or more application domains. These conditions were exploited to evaluate the suitability of several kernel mechanisms to support the development of different types of real-time applications, including monitoring systems, robotics, control systems, and multimedia processing.

In particular, each team working in this activity gave a complementary contribution to define a common infrastructure for the development of advanced real-time systems:

Scuola Superiore S. Anna, together with its affiliates University of Pavia and Evidence, has a strong expertize in kernel development, so it ensured the proper support to the other teams for maintaining the Shark kernel and teaching how to use it.
University of Aveiro, Polytechnic Institute of Porto and University of Catania (affiliated to Pisa) have a strong expertize in real-time networks, which they used to evaluate the impact of kernel mechanisms on distributed real-time applications.
Mälardalen University (now replaced by Technical University of Kaiserslauten) had already worked on video streaming and multimedia applications, therefore it committed itself to test the kernel on media processing.
Universitat Politècnica de Catalunya, expert in control systems, gave the availability for testing the kernel on critical control applications.
University of Cantabria and University of York, actively involved in standards for operating systems interfaces, gave support for implementing a POSIX interface.
University of York, expert in feasibility analysis of fixed priority systems, gave support for implementing and analyzing new scheduling schemes.
UP Madrid and its affiliate Univ. Carlos III of Madrid, expert in quality of service management, were the ideal candidates for evaluating the impact of adaptive kernel strategies in multimedia systems.

Previous Work

It focused in building a common kernel infrastructure for testing novel scheduling algorithms and resource management policies in different application environments. To do that, we had to work with a real-time kernel with a modular structure, in which the implementation of a mechanism were as much as possible independent on the implementation of the other parts. In additions, the kernel had to be open source to allow extentions, well documented, and availble for PC-based platforms, supporting a sufficient number of devices to allow partners to develop control applications. Finally, to compare the performance of different algorithms, we wanted a kernel able to measure the execution of application tasks. Operating systems like RT-Linux and RTAI were excluded due to their quite complex structure and the dependency of the scheduler with the other kernel mechanisms. The only kernel we found suitable for this project was Shark, for the reasons explained in Deliverable 2-2 JPIA-a-ART-Y1.

Once the kernel was selected, we developed a web site to create a forum for the various Shark users, to exchange messages, search for questions, etc.

The main action was to organize a workshop to introduce the Shark kernel to all the partners of the ART cluster, enabling the participants to quickly use the kernel, develop simple real-time applications, and implement novel scheduling algorithms. The workshop was held in Pontedera, Pisa (Italy), at the Scuola Superiore Sant'Anna, from February 28 to March 4, 2005.

A collaboration was started with the University of Siena at Arezzo, to enable the users at the development of real-time control applications on robot devices that are available in their control laboratory. The Scuola Universitaria Professionalizzante della Svizzera Italiana (SUPSI) was also contacted for porting a code generator for Scilab/Scicos tools toward the Shark platform.

The kernel was maintained by writing and updating the existing documentation, removing bugs, and modifying the internal structure of the scheduler. The task_endcycle() system call was modified to make it sensitive to system termination.

In addition, a graduate thesis was started at the ReTiS Lab for porting the low-level layer of Shark (OsLib) to the L4 microkernel.

The University of Aveiro, in collaboration with the University of Pavia, started investigating a kernel methodology for supporting the hard real-time communication over a dedicated Ethernet network. Finally, new device drivers were implemented based on Linux technology.

Problem Tackled in Year2

The objective of this activity in the second year was to deploy a working platform for experimenting RTOS mechanisms in control and distributed systems. Each partner selected a specific application with the aim of evaluating the impact of kernel policies on system performance. Sample real-time applications included mobile robotics, distributed systems, multimedia processing, and process control systems. The specific kernel mechanisms that have been considered for evaluation include:

Periodic task scheduling algorithms. Periodic task scheduling is extremely important since most software control systems are implemented through a set of periodic activities performing data sampling, sensory processing, control, action planning, and actuation. Although not strictly necessary, periodic execution simplifies the design of control algorithms and allows using standard control theory to guarantee system stability and performance requirements. When several of such activities execute concurrently in the same processor, however, each task may experience a variable delay and jitter, mainly due to the interference created by the other tasks. If not properly taken into account, delays and jitter may degrade the performance of the system and even jeopardize its stability. Fixed priority schemes and deadline-based algorithms (like EDF), both available into Shark, were evaluated in terms of overhead, complexity, jitter and robustness in handling overload conditions.
Aperiodic service mechanisms. In addition to periodic tasks, most of real-time applications include a number of event-driven activities generated by asynchronous interrupts coming from peripheral devices. If not properly handled, aperiodic activities may create transient overload conditions and introduce unbounded delays in periodic control tasks that could degrade system performance in an unpredictable fashion. Several aperiodic service policies, available into Shark, have been tested and evaluated, including Polling Server, Sporadic Server, Deferrable Server, Total Bandwidth Server and Constant Bandwidth Server.
Resource access protocols. When tasks interact through shared memory buffers, additional blocking times can be introduced in task execution due to the access of mutually exclusive resources. Several solutions have been proposed in the literature to bound the maximum delay caused by resource conflicts (e.g. Non-Preemptive Protocol, Higher Locker Priority, Priority Inheritance Protocol, Priority Ceiling Protocol, Stack Resource Policy, Immediate Priority Ceiling, etc.). All these protocols are available in Shark as a specific resource modules and can be used in combination with different scheduling algorithms. In addition, Shark also includes a non blocking communication mechanism (namely, the Cyclic Asynchronous Buffer), based on memory duplication, which represent an interesting alternative for the communication among periodic tasks with different periods. Testing the effectiveness of such inter-task communication mechanisms was one of the objective of this research activity.
Overload management techniques. Dynamic real-time systems are subject to unpredictable load variations which may cause transient or permanent overload conditions. If the system is not designed to handle such a situations, the system performance may degrade in an unpredictable fashion. Transient overload conditions due to execution overruns can be efficiently handled by Resource Reservations techniques, which basically isolate the temporal behavior of a task (or subset of tasks) protecting the rest of the systems from potential overruns. On the other hand, Elastic Scheduling provides an effective solution to cope with permanent overload conditions. According to this method, task utilizations are treated as flexible springs that can be compressed (by enlarging periods) to reduce the load up to a desired value. Such novel techniques (not yet available in commercial operating systems), have been implemented into Shark as basic scheduling modules to be tested and evaluated in real-world control applications.

Current Results

A New Kernel Release: Shark 1.5.1 A new kernel release, Shark 1.5.1, was issued on July 25, 2006, introducing several new features with respect to the previous version of the kernel (Shark 1.5). The work has been carried out at the University of Pavia, in collaboration with the Scuola Superiore Sant'Anna and Evidence S.r.l. The new kernel includes the following additions:
Event filters have been added to the Tracer to allow the user to select the events to be monitored at runtime. Moreover, the tracer output can now be saved on local disks, in addition to remote servers.
The support for the USB has been introduced. The support for Host and Hub devices is fully working; USB mouse, keyboards, joystick and joypad are already fully supported, while PWC chipset-based webcams are working with negligible problems on some specific machines.
The IntDrive interrupt server for Linux-imported drivers has been updated in order to correctly implement the ideas proposed in the related scientific paper. There is now a new interface for initialization, and some delicate internal mechanisms has been fixed.
A support for Dynamic Voltage Scaling has been provided to allow S.Ha.R.K. to exploit the power management techniques available in the most recent processors. AMD PowerNow and Intel Centrino SpeedStep are currently fully supported. The feature has been implemented as a kernel module and it is fully compliant with the Linux CPUFreq driver.
A S.Ha.R.K. Quick Guide has been released to simplify the work to the new users. It covers the basic topics for getting familiar with S.Ha.R.K., such as installation, basic application development, system configuration, and remote execution of S.Ha.R.K. applications to improve productivity.
Other new documents include the manual for the new S.Ha.R.K. Tracer (which covers the initialization, usage, and log saving procedures) and the HTML and TXT versions of the kernel manual. In particular, the TXT version is useful while developing under DOS, where both PDF and HTML documents cannot be handled.
Several bugs have been fixed. The most relevant concern:
the dependency of the tracer library from the network library;
a bandwidth leakage in the EDF scheduling module;
dependencies of ".h" files into makefiles;
a mistake in the group_activate call, which tried to start tasks rejected from the guarantee test;
a wrong behavior of the IntDrive in particular cases.

A Repository of Real-time Applications A repository for all real-time applications developed using the Shark kernel was created to facilitate the users in the development of new real-time software. The repository includes a a folder of supported applications and a folder of all unsupported software. The supported applications are tested and maintained by the developers to be consistent with the current kernel version, while the unsupported folder includes all programs, demos, and advanced applications developed under Shark and made available by the maintenance team.

A Repository for Scheduling Modules and Resource Management A repository of all kernel modules developed for the Shark operating system was created to facilitate the users in the development of new kernel mechanisms. The modules include scheduling modules (implementing periodic schedulers, aperiodic servers, or overload management policies) and resource modules (implementing concurrency control protocols for accessing shared resources). Each modules contains the C source code and required headers compliant with the Shark module specifications.
See it online!

Shark at University of York This work has been done by Alex Zerzelidis at the Computer Science Department of University of York. The Preemption Level Protocol (PLP) allows EDF to be used on top of priority queues and was introduced in Burns, A., Wellings, A.J., and Taft, T. S., "Supporting Deadlines and EDF Scheduling in Ada", Lecture Notes in Computer Science, Springer-Verlag, Volume 3063 / 2004. Alex Zerzelidis extended the PLP to include multi-unit resources. To achieve that, two amendments were introduced to the original protocol: Amendment 1 (applies to point 2 of Section 3) Resource ceiling preemption levels are dynamic and any resource R has a current ceiling defined as. where π(τ) is the preemption level of task τ, νR denotes the number of units of R that are currently available and μR is the maximum requirement of task τ for R. That is, the current ceiling of resource R is the maximum of zero and the preemption levels of all the tasks that may be blocked directly when there are vR units of R available. Amendment 2 (applies to point 4 of Section 3) When a task accesses a resource, the current ceiling level of the resource is recalculated and the task's active priority is raised to the new current ceiling level of the resource. The active priority of all other tasks already accessing the resource remains unchanged. When the task releases the resource the ceiling is recalculated. The PLP and its multi-unit extension was implemented on the SHARK RTOS as a new scheduling module in the kernel, called MPLP (Multi-unit PLP), which is based on the EDF and POSIX modules. The module can be registered as usual in the function TIME __kernel_register_levels__(void *arg) if we include a call to MPLP_register_level(int flags, int prioritylevels). The module implements PLP on a set of priority levels (their number specified by the prioritylevels parameter) and allows tasks to share multi-unit resources. Tasks can be created at runtime and declare a list of resources to use.

Shark at the Technical University of Kaiserlautern Shark has been used at the Technical University of Kaiserlautern (TUKL) as a platform for video processing applications and, in particular, for experimenting adaptive resource management policies for user quality control. The methods estimate resource demands for frame decoding and use Shark scheduling algorithms for guarantees. At TUKL, the Shark kernel was also adopted for education to teach real-time scheduling in an undergraduate course. Two labs are being developed: in one lab, students are required to develop a scheduling algorithm, analyse it and test it on the Shark kernel; in the other lab, students have to implement a simple video processing algorithm running on Shark, learning implementation and overhead issues.
See it online!

Shark for Control Applications A contact has been established with Dr. Sjur Vestli, from Logobject AG – Switzerland (http://www.logobject.ch/), for a possible use of Shark in robotic applications, and in particular for the control of autonomous vehicles. The group is currently using a fixed priority kernel for Power PC platforms, but they are moving to PC architectures and are looking for a kernel supporting dynamic scheduling and advanced resource management techniques for an efficient exploitation of the onboard resources.
See it online!

Shark for Education in a Master Course for IIT Graduate Students Shark was used as a sample real-time kernel in a Master course on Real-Time Systems organized in Pisa by the Scuola Superiore Sant'Anna, from March 2006 to May 2006, for 20 Indian graduate students selected from the India Institute of Technology (IIT). Students worked in groups of two or three people to develop a number of real-time concurrent applications and make experience in using advanced scheduling techniques.

Shark for education at University of Catania Shark was used as a sample real-time kernel in a graduate course on Real-Time Systems given by Lucia Lo Bello at the University of Catania, from March 2006 to June 2006. Students worked in groups of two or three people to develop a number of real-time concurrent applications and make experience in using advanced scheduling techniques. A description of the projects is available here.

Shark at the University of Illinois – Urbana Champaign A testbed was developed at the Real-Time Systems Lab of the Computer Science Department of University of Illinois at Urbana-Champaign to show the applicability of real-time OS Shark and real-time MAC protocol RI-EDF for distributed control applications, where sensors/actuators are connected through a wireless channel. A wireless distributed control system for an inverted pendulum was built as a testbed environment. In the application, camera sensors are used in complement with potentiometer sensors on the cart to balance the pendulum pole. Real-time sensory acquisition was performed on a PC using the S.Ha.R.K. operating system , whereas Berkeley Mica2 motes running TinyOS were used for wireless communication with the RI-EDF protocol. We were able to successfully control the pendulum and utilize the extra network bandwidth for other real-time communications on the same shared channel using RI-EDF protocol. More details can be found here.

Shark for robot control at University of Dresden A contact has been established during ECRTS 2006 with the robotic group at the University of Dresden, for a possible use of Shark in robot control applications. The group is currently using a simple priority-based kernel developed in their lab for Motorola 68020 architectures, and then adapted to run on Power PC platforms.

Keynotes, Workshops, Tutorials

First European Laboratory on Real-Time and Control for Embedded Systems RETIS Lab, Scuola Superiore Sant'Anna, Pisa, Italy, July 10-14, 2006. A graduate course on real-time and control for embedded systems was held in Pisa (Italy), at the RETIS Lab of the Scuola Superiore Sant'Anna, on July 10-14, 2006. It was organized in collaboration with the Control cluster lead by Karl-Erik Arzen. The purpose of the course was twofold. The first objective was to provide the most important concepts and methodologies used in developing real-time embedded systems, including fundamentals of real-time scheduling, operating systems, distributed systems, and control theory. The second and more challenging goal was to show how to apply theory into practice, teaching students how to develop simple real-time distributed control applications using the state-of-the-art technologies available into the Shark kernel. Outcomes: The course was very successful and actracted 30 participants from all over the world (10 different countries). Among them, some came from industry and research centers, indicating a great interest on such types of activities. Each participant had the possibility to closely interact with all the teachers, for the entire duration of the course, to ask specific questions and clarify obscure points. In additions, 5 lab assistants helped the participants to quickly get familiar with the Shark kernel and followed them during the software development process. Thanks to the joint cooperation of the involved ARTIST2 teams, 10 robot control devices and 6 distributed applications were prepared for the course, and each participant had the opportunity to design, develop, and test a control application of his choice, working alone or together another participant. See it online! All the projects developed in the course are available here.

Course: Improving your research skills: a mini workshop for graduate students RETIS Lab, Scuola Superiore Sant'Anna, Pisa, Italy – March 14-17, 2006 Speaker: Lui Sha (University of Illinois at Urbana Champagne, USA) This course was aimed at teaching students how to do research. In particular, it addressed the problems of how to learn, formulate and solve research problems, and how to communicate. Research means re-search: searching again and again in the product space of problem formulations and solutions until potentially high impact technologies is found. The efficiency of any search depends greatly on the methods we use, no matter the search is for oil under the ground or for new knowledge. In this mini-workshop, Professor Lui Sha shared his experience on research and education with the students, who have been guided to discover their skills and better organize their research plans. Students have learnt how to organize teams to create and improve research plans. See it online!

Keynote: Real-Time Issues in Mobile Wireless Networks Speaker: Giorgio Buttazzo Conference: 9th Int. Conference on Principles of Distributed Systems (OPODIS 2005) Pisa, Italy – December 12-14, 2005 This talk presented some of the most challenging problems to be solved in order to support the development of mobile wireless networks of cooperating robots. Some of these problems include the real-time execution of acquisition and control processes, the efficient management of computational resources, the software control of energy consumption, the real-time communication protocols on wireless networks, and the development of distributed agreement algorithms for reaching a consensus in collective decisions. See it online!

Keynote: Towards Component-Based Operating Systems Speaker: Giorgio Buttazzo Conference: Workshop on Operating Systems Platforms for Embedded Real-Time applications (OSPERT 2006) Dresden, Germany – July 4, 2006 This talk presented the characteristics and the advantages of having a component based operating system, also discussing the difficulties to be solved and possible solutions that can be adopted at the kernel level. See it online!

Keynote: Real-Time, Distributed Control Systems: Performance, Resource Planning, Applications Speaker: Josep Fuertes RETIS Lab, Scuola Superiore Sant'Anna, Pisa, Italy – May 3, 2006 In real-time control systems, the interplay of real-time constraints and control specifications can be somewhat subtle. In order to successfully analyze the behavior of a distributed control system, the type and location of jitter and delay must be characterized, and their effect on the performance of the control system understood. This talk presented how timing affects control and real-time performance and how the group at the University of Catalonia is facing the analysis and design methods for distributed control systems that cannot guarantee equidistant sampling and actuation.

The talk gave an introduction of control theory centred in the time related properties, explaining the relation between continuous and discrete sampled control systems and clarifying the interaction between performance specifications of control and real-time systems.

Publications Resulting from these Achievements

[1] Mauro Marinoni and Giorgio Buttazzo, "Adaptive DVS Management through Elastic Scheduling", Proceedings of the 10th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA 2005), Catania, Italy, September 2005.
[2] Gianluca Franchino, Giorgio Buttazzo, and Tullio Facchinetti, "A Distributed Architecture for Mobile Robot Coordination", Proceedings of the 10th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA 2005), Catania, Italy, September 2005.
[3] Paulo Pedreiras, Paolo Gai, Luis Almeida, and Giorgio Buttazzo, "FTT-Ethernet: A Flexible Real-Time Communication Protocol that Supports Dynamic QoS Management on Ethernet-based Systems", IEEE Transactions on Industrial Informatics, Vol. 1, No.3, August 2005.
[4] M. Caccamo, T. Baker, A. Burns, G. Buttazzo, and L. Sha, "Real-Time Scheduling for Embedded Systems", in Handbook of Networked and Embedded Control Systems, D. Hristu-Varsakelis and W. S. Levine Editors, Birkhauser, Boston, 2005.
[5] Giorgio Buttazzo, "Real-Time Operating Systems: The Scheduling Aspects", in The Embedded Systems Handbook, Edited by Richard Zurawski, CRC Press, 2005.
[6] Giorgio Buttazzo et al. "Adaptive Real-Time Systems for Quality of Service Management", in Embedded Systems Design, The ARTIST Roadmap for Research and Development, Lecture Notes in Computer Science, Vol. 3436, Edited by Bruno Bouyssounouse and Joseph Sifakis, Springer, 2005.
[7] Tullio Facchinetti, Giorgio Buttazzo, Mauro Marinoni, and Giacomo Guidi, "Non-Preemptive Interrupt Scheduling for Safe Reuse of Legacy Drivers in Real-Time Systems", IEEE Proceedings of the 17th Euromicro Conference on Real-Time Systems, Palma de Mallorca, Spain, July 2005.
[8] Enrico Bini, Giorgio Buttazzo, and Giuseppe Lipari, "Speed Modulation in Energy-Aware Real-Time Systems", IEEE Proceedings of the 17th Euromicro Conference on Real-Time Systems, Palma de Mallorca, Spain, July 2005.
[9] Mauro Marinoni, Giorgio Buttazzo, Tullio Facchinetti, and Gianluca Franchino, "Kernel Support for Energy Management in Wireless Mobile Ad-Hoc Networks", Proc. of the Workshop on Operating Systems Platforms for Embedded Real-Time applications (OSPERT 2005), Palma de Mallorca, Spain, July 5, 2005.
[10] Mauro Marinoni and Giorgio Buttazzo, "Balancing Energy vs. Performance in Processors with Discrete Voltage/Frequency Modes", Proceedings of the 12th IEEE International Conference on Embedded and Real-Time Computing Systems and Applications, Sydney, Australia, August 2006.
[11] Hoai Hoang, Giorgio Buttazzo, Magnus Jonsson, and Stefan Karlsson, "Computing the Minimum EDF Feasible Deadline in Periodic Systems", Proceedings of the 12th IEEE International Conference on Embedded and Real-Time Computing Systems and Applications, Sydney, Australia, August 2006.
[12] Giorgio Buttazzo and Paolo Gai, "Efficient Implementation Of An EDF Scheduler For Small Embedded Systems", Proceedings of the 2nd Workshop on Operating Systems Platforms for Embedded Real-Time applications (OSPERT 2006), Dresden, Germany, July 2006.
[13] Giorgio Buttazzo, "Real-Time Operating System Support for Energy-Aware Computing", Automazione e Strumentazione, 14(1), pp. 88-95, January 2006.

ARTIST2 Participants: Expertise and Roles

Cluster Leader: Giorgio Buttazzo – Scuola Superiore S. Anna (Italy)
Activity coordinator, kernel maintenance, development of robotic applications.
Team Leader: Luis Almeida – University of Aveiro (Portugal)
networking platform, development of distributed applications.
Team Leader: Gerhard Fohler – Technical University of Kaiserslauten (Germany)
video streaming applications, scheduling.
Team Leader: Michael Gonzalez Harbour – University of Cantabria (Spain)
definition of the POSIX operating system interface.
Team Leader: Alan Burns – University of York (UK)
feasibility analysis of fixed priority real-time systems.
Team Leader: Eduardo Tovar – Polytechnic Institute of Porto (Portugal)
distributed applications and QoS over heterogeneous networks.
Affiliated Participants: Expertise and Roles
Team Leader: Ivo De Lotto – University of Pavia (Italy)
kernel maintenance.
Team Leader: Paolo Gai – Evidence s.r.l. (Italy)
real-time kernels and operating systems standards.
Team Leader: Lucia Lo Bello – University of Catania (Italy)
Distributed real-time applications.
Team Leader: Pau Marti – Universitat Politècnica de Catalunya (Italy)
control applications.

70. Industrial Applications

Gerhard Fohler, Giorgio Buttazzo — 2006-09-06T15:46:36Z

The most important industrial fields that can benefit from adaptive real-time technology include Consumer Electronics, Industrial Automation, and Telecommunications. Consumer Electronics (CE) products range from miniature cameras and MP3 players to advanced media servers and large displays. Mainly driven by Moore's law, the evolution in the CE industry is very fast. The software content, measured in ROM size, grows one order of magnitude every 6 to 7 years. To keep up with this speed, the industry moves to families of products based on retargetable platforms: systems on chip that allow media streaming and featurization.

In the area of Industrial Automation there is a trend to search distributed solutions and to prepare hardware and software for connecting the general plant actuators, sensors and the controllers. Distributed solutions give a natural automation condition to common industrial needs as usually such plants are physically and topologically distributed. At the same time, there is an increase of demands for new options and improvements in the automation results, fetching more control of plant secondary data. This imposes a continuous increment in processing power and memory capacity that requires adaptivity at different levels of system operation.

Embedded systems for telecommunications applications are mainly targeted to the interfaces between communication technologies and to coding/decoding operations. They may be considered real-time as they have timeliness requirements for some of the critical operations they must perform. The referred systems are microprocessor based platforms, often integrating a second processor (e.g., a DSP) devoted to specific functions, like MPEG coding. Depending on the system, a proprietary kernel can be used or a real-time operating system, (e.g. iRMX). Currently Linux starts to become the adopted operating system. The verification of real-time behavior is made by the experimental inspection of the timeliness requirements at the development phase and at the operation phase by in-situ monitoring.

30. Main Research Trends

Gerhard Fohler, Giorgio Buttazzo — 2006-09-06T15:43:34Z

Most of today's embedded systems are required to work in dynamic environments, where the characteristics of the computational load cannot always be predicted in advance. Still timely responses to events have to be provided within precise timing constrains in order to guarantee a desired level of performance. The combination of real-time features in dynamic environments, together with cost and resource constraints, creates new problems to be addressed in the design of such systems, at different architecture levels.

To cope with dynamic environments, a system must be adaptive; that is, it must be able to adjust its internal strategies in response to a change in the environment, to keep the system performance at a desired level. Implementing adaptive embedded systems requires specific support at different levels of the software architecture. The most important component affecting adaptiveness is the kernel; however, flexibility can also be introduced above the operating system, in a software layer denoted as a middleware, and also in the programming language used to develop the application. Some embedded systems are large and distributed among several computing nodes. In these cases, special network methodologies are needed to achieve adaptive behavior and predictable response. Often such a support cannot be found in today's commercial software.

IST-FRESCOR (FPVI)

Giorgio Buttazzo — 2006-08-25T13:02:09Z

In the area of development methods for embedded applications with complex timing requirements we have identified a large gap between the research state of the art and the applicability. The project will be addressed at closing this gap, and providing embedded systems developers the engineering solutions to manage timing requirements at a high level of abstraction, thus lowering the design and development costs, and speeding up the time to market.

The project will address the OS primitives needed to support the contract-based scheduling, the development of middleware to support the contracts themselves and adaptively manage the quality of service, the integration of this infrastructure into a container framework for component-based development methodologies, the development of simulation and analysis tools, and the evaluation and exploitation of the results.

See it online!

OSEK

Giorgio Buttazzo — 2006-08-25T12:57:05Z

OSEK is an industry standard for an open-ended architecture for distributed control units in vehicles.

See it online!

What is OSEK/VDX ?

In May 1993 OSEK has been founded as a joint project in the German automotive industry aiming at an industry standard for an open-ended architecture for distributed control units in vehicles. OSEK is an abbreviation for the German term "Offene Systeme und deren Schnittstellen für die Elektronik im Kraftfahrzeug" (English: Open Systems and the Corresponding Interfaces for Automotive Electronics). Initial project partners were BMW, Bosch, DaimlerChrysler, Opel, Siemens, VW and the IIIT of the University of Karlsruhe as co-ordinator. The French car manufacturers PSA and Renault joined OSEK in 1994 introducing their VDX-approach (Vehicle Distributed eXecutive) which is a similar project within the French automotive industry. At the first workshop on October 1995 the OSEK/VDX group presented the results of the harmonised specification between OSEK and VDX. After the 2nd international OSEK/VDX Workshop in October 1997 the 2nd versions of the specifications were published.

The open architecture introduced by OSEK/VDX comprises the three areas:
Communication (data exchange within and between control units)
Operating System (real-time execution of ECU software and base for the other OSEK/VDX modules)
Network Management (Configuration determination and monitoring)

For Communication and Operating System, there are alternative standards for normal requirements (standard operating system/communication specifications), or special requirements covering globally synchronised fault-tolerant architectures (OSEKtime specifications).

Motivation

high, recurring expenses in the development and variant management of non-application related aspects of control unit software
incompatibility of control units made by different manufactures due to different interfaces and protocols

Goal

Support of the portability and reusability of the application software by:

specification of interfaces which are abstract and as application-independent as possible
specification of an user interface independent of hardware and network
efficient design of architecture: The functionalities shall be configurable and scalable, to enable optimal adjustment of the architecture to the application in question
verification of functionality and implementation of prototypes in selected pilot projects

Advantages

clear savings in costs and development time
enhanced quality of the software of control units of various companies
standardized interfacing features for control units with different architectural designs
sequenced utilization of the intelligence (existing resources) distributed in the vehicle, to enhance the performance of the overall system without requiring additional hardware
provides absolute independence with regards to individual implementation, as the specification does not prescribe implementation aspects

A structured and modular software implementation based on standardised interfaces and protocols as proposed by OSEKVDX is a necessary condition for the portability and extendibility, and thus the reusability of existing software. Functional extendibility means the integration of new application functions into a single control unit together with other application functions.

Application porting is the transfer of application functions from one hardware platform to another one with only minor modifications, e.g. porting of existing application software to the next product generation of an ECU. Moreover, extendibility and portability should be independent of the supplier of application functions, i.e. a "co-habitation" of software from different suppliers must be possible. It shall be remarked that OSEKVDX does not prescribe the implementation of OSEKVDX modules, i.e. different ECUs may have the same OSEKVDX interfaces, but different implementations, depending on the hardware architecture and the performance required.

POSIX

Giorgio Buttazzo — 2006-08-25T12:49:54Z

See it online!

ArtistDesign NoE

45. Report

1. Introduction

2. Course structure

3. Practical laboratory experience

4. Participants

5. Diploma

55. Participants

20. Programme

MONDAY (slides)

TUESDAY (slides)

WEDNESDAY (slides)

THURSDAY (slides)

FRIDAY (slides)

90. Participants

A Common Infrastructure for Adaptive Real-Time Systems

Baseline

Previous Work

Problem Tackled in Year2

Current Results

Keynotes, Workshops, Tutorials

Publications Resulting from these Achievements

ARTIST2 Participants: Expertise and Roles

Affiliated Participants: Expertise and Roles

70. Industrial Applications

30. Main Research Trends

IST-FRESCOR (FPVI)

OSEK

What is OSEK/VDX ?

Motivation

Goal

Advantages

POSIX