Research Topics

Today, electronic systems form an integral part of most technical devices. They are used in cars, planes, medical equipment or satellites. With the growing complexity of such systems, the design and implementation is becoming more and more challenging.

Goals:

1. Correct design of technical systems including hardware and software - an integrated view over different levels of abstraction

2. Methods and tools for system design - from natural language specification to complex realizations in hardware

3. Modelling of systems - considering non-functional properties like power, robustness and testability

4. Technical implementation of systems with a special focus on correctness, reliability and control paradigms - applications in robotics, space systems and mobility

Ongoing Research Projects:

Please note that the following abstracts describe example projects of our PhD students. They are listed here to give an impression of the research topics in SyDe. These projects are not open for new applicants.

• Fault-Tolerance Techniques for FPGAs
  Gökçe Aydos, ZESY / DLR, advised by Prof. Dr. Görschwin Fey

  Reliability is an important issue in safety-critical systems. This also applies to space systems like satellites. In contrast to many other safety-critical systems, they can hardly be repaired in case of a failure, once they are in space. Additionally, these systems are exposed to the ionizing radiation from the sun, which can induce current impulses on the electronic components. This can lead to transient and permanent changes in the digital circuits which can eventually lead to unwanted behavior of the system. Field-programmable Gate Arrays (FPGAs) are integrated circuits (ICs) composed of generic logic blocks, which can be programmed to perform a specific data-processing task. FPGAs are popular in avionics because they allow efficient data processing, and can be adapted to a different task or mission by reconfiguration. In order to safely use these ICs in space, fault-tolerance techniques must be implemented. In this project, fault-tolerance techniques for FPGAs on different algorithmic and technological levels are investigated which can be implemented on commercially-available FPGA architectures and do not require an expensive custom IC manufacturing process.

• Agile System Design
  Melanie Diepenbeck, AGRA, advised by Prof. Dr. Rolf Drechsler

  Nowadays, the design of hardware systems is a challenging and erroneous task, making it more and more difficult to ensure correctness while designing a system. Testing and verification are usually applied as a post-process to the implementation. Hence, errors are found late the design process, leading to high costs to fix them. In agile development techniques the design of systems is driven by tests in short development cycles. In this work new development techniques (such as Test Driven Development (TDD)) are considered which can reduce the number of errors in systems and - in some cases - may even ensure the correctness of a system. Based on Behavior Driven Development (BDD), an extension of TDD, a new design flow is proposed which handles the specifities that arise in hardware design. This design flow is extended to ensure correctness by automatically generalizing properties from these test cases and generating missing test cases for uncovered code.

• Semantic Object Recognition in 3D Point Clouds
  Malgorzata Goldhoorn, AG Robotics, advised by Prof. Dr. Frank Kirchner

  The extraction of relevant knowledge from the natural human environment is one of the most important capabilities which a mobile robot should be equipped with to act autonomously. Without knowledge about the meaning of the objects in a given task, the robot cannot perform it correctly. Dealing with entities in natural human environments is not a trivial problem, but rather a challenging task. The robot has to extract the objects from the raw sensor data and give them a meaningful description, while dealing with uncertainties and ambiguities which may occur. The goal of this thesis is to develop a new algorithm for robust semantic object recognition from 3D point clouds. Using this approach, the robot should be able to extract objects from sensor data and give them a correct semantic description.

• Consistent System Design
  Judith Peters, AGRA, advised by Prof. Dr. Rolf Drechsler

  The development of electronic systems is a process which relies on translating abstract descriptions of a system to more concrete ones. A first description, as provided by the customer, describes the desired system using natural language. In the following, a developer has to translate this description into a formal model of the system, which will then be refined (possibly by other developers) and finally translated to program code. This process is error prone due to the different languages and persons involved, causing misunderstandings and translation errors. To avoid these problems, system design shall be done in future mostly automatically from the first description in natural language on. In this problem area, this thesis investigate strategies to translate models automatically to code and to verify this translation.

• Reversible Logic
  Eleonora Schönborn, AGRA, advised by Prof. Dr. Rolf Drechsler

  With the breakthrough of the mobile web on smartphones and the integration of sensors and computational components into everyday objects, low energy consumption of integrated circuits has become a crucial design goal. Established power management techniques like voltage scaling are reaching their limits. Incorporating the energy consumption in the design flow is not a simple task either. Reversible logic is a promising alternative to conventional hardware. It can be applied e.g. in the area of low-power design and quantum computing. Since the rules of conventional logic do not necessarily apply to reversible logic, design methods have to be reconsidered or developed from scratch. The focus of this work is on the design of reversible circuits using hardware description languages.

• Dynamic Bat Control and Audience Interaction of a Redundant Ball Playing Robot
  Dennis Schüthe, AGEBV, advised by Prof. Dr. Udo Frese

  Dynamic processes place high demands on the control systems of robots. Especially the hitting of balls is a well known application, where bats like table-tennis rackets or baseball bats are used. For the hitting process in this project, a sphere is placed at the end of a rod. The hitting motion shows how the robot deals with fast movements and how precisely it works. The recognition of the thrown ball and the entry point of the ball in the workspace of the robot must be performed in real time. If position and time of the ball has been calculated at the impact of the hitting point, only a few milliseconds are available to move the robot's end-effector with given end time, speed and a certain point in order to hit the ball actively back to the thrower or another person. This should be done very dynamically and will drive the motors during the movement of the end-effoctor for a short time at their load limits. Therefore a linear quadratic regulator ( LQR ) should be used for non-linear Systems by optimizing the feedback parameters, such that it can be used to handle the nonlinearities and reaches a nearly optimal perfomance.

• Verification of System Models
  Julia Seiter, AGRA, advised by Prof. Dr. Rolf Drechsler

  The application of modeling languages such as UML and SysML for system design creates a formal description of the system at an early stage. This description can be used, amongst others, for a correctness proof before the system is built or implemented. Here, many promising approaches exist already. However, these approaches allow only the verification of single aspects of a model, e.g. if an operation of a class diagram is executable. They do not consider the system as a whole so that it is not possible to make a statement about the completeness of the verification. Consequently, it remains unclear whether the model still allows erroneous behaviour. Thus, verification methods and techniques which consider the complete system are to be explored. Further aspects are the proof of equivalence of two systems and the generation of properties which cover the complete behaviour of the system. On the one hand, existing approaches are to be applied and optimized, but also new technologies and methods are to be considered.

• Equivalence Checking of Hardware/Software Systems
  Niels Thole, ZESY / DLR, advised by Prof. Dr. Görschwin Fey

  When developing hardware systems, it is common practice to start with an abstract „golden model“. This model is usually written in a system level programming language like C++ or SystemC. If the model fulfills its specification, it is then refined in an iterative process and finally transformed into a hardware description language. The system needs to behave equivalently before and after each transformation step. Formal methods are used to prove or disprove the equivalence between the old and the new model. This method is also relevant when exchanging a subsystem for a robust and more complex one. In this project, I investigate methods for the equivalence check between two C++ programs. This serves as a basis to enable the checking between two hardware models or between a hardware model and a software model.

Projects

How can I turn natural language into correct source code?
Model Driven System Design
» Source: Universität Bremen, AG Rechnerarchitektur

How can I design systems that consume less energy?
The Green Revolution on a Chip
» Source: DFKI GmbH, Cyber-Physical Systems