COMES School
In the frame of the Nano -Tera Educational Program, the ALaRI Institute is organizing a Doctoral School on:
Autumn School on
Managing Complexity in Tera-Scale Embedded Systems:
The Information and Computational levels
16 - 20 November, 2009
Lugano, Switzerland
All the information regarding the COMES School are reported in this page.
Please use links on the right side to navigate among contents or just scroll down till the end.
School Programme
This school will bring together some of the best lecturers from Europe, in a one week programme, and be a fantastic opportunity for interaction.
Description
Systems envisioned within the Nano-Tera framework are quite often intrinsically very complex: beyond that, they are in most instances (as made clear in the Nano-Tera Technical Scope) devised to interact with the physical world, facing challenges that go from the modelling aspects of the phenomena they should deal with to intrinsic non-determinism of such phenomena. In fact, if technologies make it possible to create systems of tera-complexity, the challenge then arises of designing, simulating, managing such systems: only if such challenges are overcome will the systems become actually viable.
The overall problem of complexity management for embedded systems is clearly of such extension as to deserve more than one Autumn School in order to cover its various aspects; this aims at preparing a strong basis, considering different viewpoints and presenting challenges and solutions of specific relevance to Nano-Tera: on such background one can envision creating in the future more specialized schools targeting, e.g., the challenges of software for complex distributed systems or of reliability.
Subjects discussed will include design complexity of systems on chip, management of very complex, possibly distributed, systems, where real-time constraints have to be met, computational complexity (with particular reference to modeling), the problem of dealing with uncertainty and the concept of probably approximately correct computation, control complexity, specific design aspects of complex software systems. Theoretical aspects will be discussed in depth, and the analysis of case studies will be targeted.
Lectures
- Randomisation in embedded systems design: dealing with computational
Complexity
Dealing with uncertainty. Introducing the concept of probably approximately correct computation in embedded systems design.
Prof. C. Alippi, Politecnico di Milano, Italy. See abstract.
1. A hierarchy for uncertainty: embedded systems in a noisy environment
2. The noise to signal ratio: exploiting a priori information for an accurate system design
3. Dealing with complexity: from a deterministic approach to a probabilistic one
3.1 Randomised algorithms
3.2 PACC
3.3 Analysis problem
3.4 Synthesis problem
4. Some design examples
- Intelligent Adaptive Systems For Control
How do we control very complex systems containing sensors and actuators.
Prof. M. M. Polycarpou, University of Cyprus, Cyprus. See abstract.1. Motivation for Adaptation and Learning: Dealing with Uncertainty and Complexity 2. Introduction to Intelligent Control 3. Simple Adaptive Schemes 4. Adaptive Parameter Estimation 5. Adaptive Approximation Based Control
- Tackling the complexity problem in controller design and implementation
Managing the design of very complex automatic systems.
Prof. S. Balemi, SUPSI, Switzerland. See abstract.
Controller design problems are increasingly solved using numerical methods. Unfortunately many of these problems, in particular in presence of nonlinear constraints such as saturations and rate limitations, are computationally difficult. In some cases, the reformulation of the problem can lead to a lower complexity of the computation. Most often this is not possible, but approximations or iterations of fast methods can greatly improve the convergence rate. The newly developed techniques can also be used to speed up the implementation of controllers which rely on on-line computation. For example, model predictive controllers become increasingly interesting also in fast applications which were out of reach only few years ago.
- Design of distributed embedded systems with real-time constraints
Managing very complex distributed systems with real-time constraints.
Prof. L. Thiele, ETH, Zurich. See abstract.Often, the constraints imposed by a particular application domain require a distributed implementation of embedded systems. In this case, a number of software or hardware components communicate via some interconnection network. We find this structure on various levels of granularity, starting from multiprocessor systems on a chip via distributed embedded control units (ECU) in automotive applications and ending at large-scale sensor networks. Components in such systems are often specialized as they need to match their local environment and their intended functionality. The same observations holds for interconnection networks that may be composed of several interconnected dedicated sub-networks, each one with its own communication protocol and topology. As has been mentioned already, the architectural concepts of heterogeneity, distributivity and parallelism can even be observed on single hardware components themselves, as they are often implemented as so-called systems-on-chip (SoC) or multiprocessor-systems-on-a-chip (MPSoC). In these components, a collection of memories and heterogeneous computing resources are implemented on a single device, and communicate using networks-on-chip (NoC) that can be regarded as dedicated interconnection networks involving adapted protocols, bridges or gateways. Embedded systems are typically reactive systems that are in continuous interaction with their physical environment to which they are connected through sensors an actuators. Examples are applications in multimedia processing, automatic control, automotive and avionics, and industrial automation. Therefore, many embedded systems must meet real-time constraints, i. e. they must react to stimuli within a time interval dictated by the environment. It becomes apparent that heterogeneous and distributed embedded real-time systems as described above are inherently difficult to design and to analyze because of the tight interaction between computation, communication and the available resources. Part of this difficulty is caused by the fact that the functional and extra-functional behavior of the system is influenced by interferences on shared resources such as processors, memory or communication devices. Packet streams or tasks may prevent each other from using these resources, even if they belong to independent parts of the application. As a result, resource sharing strategies influence the system behavior to a large extend. In addition, the system environment which continuously interacts with the embedded system may vary and cause an additional degree of non-determinism. During the system level design process of an embedded system, a designer is typically faced with questions such as whether the timing properties of a certain system design will meet the design requirements, what architectural element will act as a bottleneck, or what the on-chip memory requirements will be. Consequently it becomes one of the major challenges in the design process to analyze specific characteristics of a system design, such as end-to-end delays, buffer requirements, or throughput in an early design stage, to support making important design decisions before much time is invested in detailed implementations. This analysis is generally referred to as system level performance analysis. If the results of such an analysis is able to give guarantees on the overall system behavior, it can also be applied after the implementation phase in order to verify critical system properties. One of the major requirements for models, methods and tools is their support for a modular, component-based design. This aspect covers as well the composition of the underlying hardware platform as well as the software design. Because of the import role of resource interaction, these components not only need to talk about functional properties but also about resource interaction. In the presentation, we will cover the following aspects of system level performance analysis of distributed embedded systems: a. Approaches to system-level performance analysis (simulation-based vs. analytic methods, review of existing methods and tools).Requirements in terms of accuracy, scalability, composability and modularity. b. Modular Performance Analysis (MPA): basic principles, methods and tool support. c. Examples that show the applicability, the embedding into design space exploration, and a comparison to simulation-based approaches.
- Testing Complex Software System
Dealing with complexity in advanced distributed test software systems.
Prof. Pezzè, USI, Switzerland. See abstract.Software systems are rapidly evolving and becoming more and more dynamic. Complexity, dynamic evolution and emergent behaviors challenge classic testing approaches that are no more adequate to guarantee the quality of complex software systems. Dynamic analysis and self-healing techniques are emerging as appealing approaches to control the quality of complex and dynamically evolving systems. In this lecture we will see the classic and modern approaches to dynamic analysis and we will understand how they can support the design of self-healing systems, that are systems able to automatically detect failures, diagnose and fix faults without human intervention.
- Dealing with complexity when simulating the physical world
Designing very complex digital systems on chip.
Prof. R. Krause, USI, Switzerland. See abstract.1. Reducing complexity: from the physical world to mathematical models 2. Maintaining complexity: discretization techniques in mechanics and biomechanics 3. Fighting complexity: efficient solution of linear and nonlinear systems and adaptivity
- Reducing System-on-Chip Design Complexity by Electronic System Level Design
Designing very complex digital systems on chip.
Prof. R. Leupers, RWTH-Achen, Germany. See abstract.Electronic System-Level Design (ESL) provides novel high-level methodologies and tools to efficiently design embedded multicore systems-on-chip (SoC). In fact, ESL is considered the only way to manage the complexity of future billion transistor SoC designs. We will outline the evolution in embedded applications and architectures, followed by an introduction to the most important ESL trends, such as HW/SW exploration, synthesis, compilation, and simulation approaches. The lecture will be complemented by a lab session, where the participants gain hands-on experience with selected ESL tools and the SystemC modeling language.
- Case Studies
Leading experts from the industrial world may be invited, in particular for introducing and discussing case studies. See abstract.Abstract not available yet. We apologise for that.
Schedule
Monday 16.11.2009
Time | Description |
---|---|
09:00 - 09:30 | Registration |
09:30 - 10:00 | Opening Prof. M. Sami (USI) |
10:00 - 13:00 | Randomisation in embedded systems design: dealing with computational Complexity Prof. C. Alippi (PoLiMi) |
14:30 - 17:30 | Design of distributed embedded systems with real-time constraints Prof. L. Thiele (ETH) |
Tuesday 17.11.2009
Time | Description |
---|---|
09:00 - 12:30 | Intelligent Adaptive Systems For Control (Part I) Prof. Polycarpou (University of Cyprus) |
14:00 - 16:00 | Intelligent Adaptive Systems For Control (Part II) Prof. Polycarpou (University of Cyprus) |
Wednesday 18.11.2009
Time | Description |
---|---|
09:00 - 12:30 | Dealing with complexity when simulating the physical world Prof. R. Krause (USI) |
14:00 - 17:00 | Practical session 1 (Lab 157) Dr. M. Roveri (PoLiMi) |
Thursday 19.11.2009
Time | Description |
---|---|
09:00 - 12:30 | Tackling the complexity problem in controller design and implementation Prof. S. Balemi (SUPSI) |
13:30 - 16:30 | Testing Complex Software System Prof. M. Pezzè (USI) |
Friday 20.11.2009
Time | Description |
---|---|
09:00 - 12:30 | Reducing System-on-Chip Design Complexity by Electronic System Level Design Prof. R. Leupers (RWTH-Achen) |
13:30 - 16:00 | Practical session 2 (room: ALaRI Lab) Eng. J. Castrillon (RWTH-Achen) |
Assignments
For PhD students who need to obtain credits from the Autumn school, assignments after the practical sessions will be set up and graded (pass/fail) so that attendees can gain credit through their postgraduate schools.
Cert. & ECTS
Certificates of attendance to the School will be provided upon request. For School students who have gone through their Assignment, the certificate will provide also an indication of the evaluation passed and of the 5 Credits (ECTS) corresponding to the School (ECTS).
Participation
The attendance to the School is free. It includes participation in all lectures and as well course materials, welcome receptions and lunches and coffee breaks. The free attendance does not cover the accommodation agreement, dinner, and travel.
Registration
While the School is designed for PhD students, other participants are welcome as well, as long as there will be openings.
The attendance to the COMES Autumn School is free. It includes participation in all lectures, as well as the course materials, lunches and coffee breaks.
The free attendance does not cover the accommodation, dinner, and travel.