Refine
Department, Institute
Document Type
- Conference Object (29)
- Article (7)
- Report (2)
- Book (1)
- Part of a Book (1)
- Diploma Thesis (1)
- Doctoral Thesis (1)
- Preprint (1)
Year of publication
Keywords
- Aerodynamics (2)
- Heart Rate Prediction (2)
- MAP-Elites (2)
- Quality Diversity (2)
- Surrogate Modeling (2)
- UAV (2)
- aerodynamics (2)
- dynamic vector fields (2)
- flight zone (2)
- geofence (2)
Theoretische Informatik
(2002)
Eine anschauliche Einführung in die klassischen Themenbereiche der Theoretischen Informatik für Studierende der Informatik im Haupt- und Nebenfach. Die Autoren wählen einen Ansatz, der durch zahlreiche ausgearbeitete Beispiele auch LeserInnen mit nur elementaren Mathematikkenntnissen den Zugang zu Berechenbarkeit, Komplexitätstheorie und formalen Sprachen ermöglicht. Die mathematischen Konzepte werden sowohl formal eingeführt als auch informell erläutert und durch grafische Darstellungen veranschaulicht. Das Buch umfasst den Lehrstoff einführender Vorlesungen in die Theoretische Informatik und bietet zahlreiche Übungsaufgaben zu jedem Kapitel an.
An iterative computer-aided ideation procedure is introduced, building on recent quality-diversity algorithms, which search for diverse as well as high-performing solutions. Dimensionality reduction is used to define a similarity space, in which solutions are clustered into classes. These classes are represented by prototypes, which are presented to the user for selection. In the next iteration, quality-diversity focuses on searching within the selected class. A quantitative analysis is performed on a 2D airfoil, and a more complex 3D side view mirror domain shows how computer-aided ideation can help to enhance engineers' intuition while allowing their design decisions to influence the design process.
Evolutionary computation and genetic algorithms (GAs) in particular have been applied very successfully to many real world application problems. However, the success or failure of applying Genetic Algorithms is highly dependent on how a problem is represented. Additionally, the number of free parameters makes applying these methods a science of its own, presenting a huge barrier to entry for beginners. This tutorial will give a summary on various representational aspects, discuss parametrization and their influence on the dynamics of GAs.
A comprehensive analysis of cardiovascular control (CVC) patterns with multiple subjects is presented. It became feasible by recent methodological advances. Simple computer models were generated automatically, reproducing only factors of the true model that are relevant to the focus if investigation. These models?named aspect-models?could in turn be used in model individualization, thus reducing the necessary computational amount. The achieved speedup by a factor of more than three thousand and the high numerical stability of the resulting method allows the unsupervised identification of a large body of experimental data. The analysis of tilt table experiments of 18 subjects revealed a remarkable variety of reaction patterns. Closer examination yielded different classes of subjects. Two main groups corresponding to basic types of CVC were observed. Three outliers could be assigned to the specific situation of some subjects.
A concept called model individualization is presented. It is used to modify computer based models to reproduce observed individual behavior. During D2-, MIR97- and Neurolab-missions tilt-table and LBNP-experiments were carried out. Physiological data describing the cardiovascular reactions of the astronauts were recorded. The appropriateness of the rheoretical principles is demonstrated with MIR97 tilt-table experiments. Finally the resulting individualized model is investigated to propose hypotheses on probable alterations in the cardiovascular system induced by microgravity.
Surrogate-assistance approaches have long been used in computationally expensive domains to improve the data-efficiency of optimization algorithms. Neuroevolution, however, has so far resisted the application of these techniques because it requires the surrogate model to make fitness predictions based on variable topologies, instead of a vector of parameters. Our main insight is that we can sidestep this problem by using kernel-based surrogate models, which require only the definition of a distance measure between individuals. Our second insight is that the well-established Neuroevolution of Augmenting Topologies (NEAT) algorithm provides a computationally efficient distance measure between dissimilar networks in the form of "compatibility distance", initially designed to maintain topological diversity. Combining these two ideas, we introduce a surrogate-assisted neuroevolution algorithm that combines NEAT and a surrogate model built using a compatibility distance kernel. We demonstrate the data-efficiency of this new algorithm on the low dimensional cart-pole swing-up problem, as well as the higher dimensional half-cheetah running task. In both tasks the surrogate-assisted variant achieves the same or better results with several times fewer function evaluations as the original NEAT.
Neuroevolution methods evolve the weights of a neural network, and in some cases the topology, but little work has been done to analyze the effect of evolving the activation functions of individual nodes on network size, an important factor when training networks with a small number of samples. In this work we extend the neuroevolution algorithm NEAT to evolve the activation function of neurons in addition to the topology and weights of the network. The size and performance of networks produced using NEAT with uniform activation in all nodes, or homogenous networks, is compared to networks which contain a mixture of activation functions, or heterogenous networks. For a number of regression and classification benchmarks it is shown that, (1) qualitatively different activation functions lead to different results in homogeneous networks, (2) the heterogeneous version of NEAT is able to select well performing activation functions, (3) the produced heterogeneous networks are significantly smaller than homogeneous networks.
Behaviour-based robotics (cf. Brooks [2]) has mainly been applied to the domain of autonomous systems and mobile robots. In this paper we show how this approach to robot programming can be used to design a flexible and robust controller for a five degrees of freedom (DOF) robot arm. The implementation of the robot controller to be presented features the sensor and motor patterns necessary to tackle a problem we consider to be hard to solve for traditional controllers. These sensor and motor patterns are linked together forming various behaviours. The global control structure based on Brooks' subsumption architecture will be outlined. It coordinates the individual behaviours into goal-directed behaviour of the robot without the necessity to program this emerging global behaviour explicitly and in advance. To conclude, some shortcomings of the current implementation are discussed and future work, especially in the field of reinforcement learning of individual behaviours, is sketched.
This paper describes the development of a Pedelec controller whose performance level (PL) conforms to European standard on safety of machinery [9] and whose soft- ware is verified to conform to EPAC standard [6] by means of a software verification technique called model checking. In compliance with the standard [9] the hardware needs to implement the required properties corresponding to categories “C” and “D”. The latter is used if the breaks are not able to bring the velomobile with a broken motor controller to a full stop. Therefore the controller needs to implement a test unit, which verifies the functionality of the components and, in case of an emergency, shuts the whole hardware down to prevent injuries of the cyclist. The MTTFd can be measured through a failure graph, which is the result of a FMEA analysis, and can be used to proof that the Pedelec controller meets the regulations of the system specification. The analysis of the system in compliance with [9] usually treats the software as a black box thus ignoring its inner workings and validating its correctness by means of testing. In this paper we present a temporal logic specification according to [6], based on which the software for the Pedelec controller is implemented, and verify instead of only testing its functionality. By means of model checking [1] we proof that the software fulfills all requirements which are regulated by its specification.
A detailed analysis of autonomic cardiovascular control (ACVC) may provide a key to a better understanding of the mechanisms underlying postflight orthostatic hypotension. The central substrate of human ACVC is not directly accessible to measurements and observation in space research. Modelling--supporting inference and physiological reasoning--is a valuable tool to disclose its involvement We are currently determining the suitability of artificial neural networks (ANN's) as a model of the central substrate of ACVC. Having conducted a number of experiments with simulated tilt test data to clarify the choice of input coding and of architectural biases in network training we will now report on the approximation of data obtained from human subjects during preparation of the German MIR'97 and D-2 missions.
We present a model checking algorithm for ∀CTL (and full CTL) which uses an iterative abstraction refinement strategy.
It terminates at least for all transition systems M that have a finite simulation or bisimulation quotient. In contrast to other abstraction refinement algorithms, we always work with abstract models whose sizes depend only on the length of the formula θ (but not on the size of the system, which might be infinite).
In der vorliegenden Arbeit werden Verfahren vorgestellt, die geeignet sind, Modelle des menschlichen kardiovaskulären Systems an individuelle Kreislaufreaktionen anzupassen. Allgemeine Kreislaufmodelle des menschlichen kardiovaskulären Systems sind in der Regel nichtlineare Differentialgleichungssysteme, für die es keine effizienten Optimierungsverfahren gibt. Durch die Einschränkung auf relevante Aspekte (bzgl. der Individualisierungsaufgabe) wird ein solches Modell auf Modelle einfacherer Struktur projiziert, die eine Approximation durch Funktionsapproximatoren erlauben, für welche wiederum effiziente Optimierungsalgorithmen existieren. In Abhängigkeit von der Struktur der Individualisierungsaufgabe kommt zusätzlich ein modifiziertes BFGS-Verfahren zum Einsatz, das Approximationen solcher Modellaspekte verwendet um die Konvergenz der Modellindividualisierung zu verbessern.
An evolutionary algorithm is presented to solve the optimal control problem for energy optimal driving. Results show that the algorithm computes equivalent strategies as traditional graph searching approaches like dynamic programming or A*. The algorithm proves to be time efficient while saving multiple orders of magnitude in memory compared to graph searching techniques. Thereby making it applicable in embedded applications such as eco-driving assistants or intelligent route planning.
Design optimization techniques are often used at the beginning of the design process to explore the space of possible designs. In these domains illumination algorithms, such as MAP-Elites, are promising alternatives to classic optimization algorithms because they produce diverse, high-quality solutions in a single run, instead of only a single near-optimal solution. Unfortunately, these algorithms currently require a large number of function evaluations, limiting their applicability. In this article we introduce a new illumination algorithm, Surrogate-Assisted Illumination (SAIL), that leverages surrogate modeling techniques to create a map of the design space according to user-defined features while minimizing the number of fitness evaluations. On a two-dimensional airfoil optimization problem SAIL produces hundreds of diverse but high-performing designs with several orders of magnitude fewer evaluations than MAP-Elites or CMA-ES. We demonstrate that SAIL is also capable of producing maps of high-performing designs in realistic three-dimensional aerodynamic tasks with an accurate flow simulation. Data-efficient design exploration with SAIL can help designers understand what is possible, beyond what is optimal, by considering more than pure objective-based optimization.
An evolved neural network controller is presented to solve the optimal control problem for energy optimal driving. A controller is produced which computes equivalent control commands to traditional graph searching approaches, while able to adapt to varied constraints and conditions. Furthermore, after training, trivial amounts of computation time and memory are required, making the approach applicable for embedded systems and path planning applications.
The MAP-Elites algorithm produces a set of high-performing solutions that vary according to features defined by the user. This technique has the potential to be a powerful tool for design space exploration, but is limited by the need for numerous evaluations. The Surrogate-Assisted Illumination algorithm (SAIL), introduced here, integrates approximative models and intelligent sampling of the objective function to minimize the number of evaluations required by MAP-Elites.
The ability of SAIL to efficiently produce both accurate models and diverse high performing solutions is illustrated on a 2D airfoil design problem. The search space is divided into bins, each holding a design with a different combination of features. In each bin SAIL produces a better performing solution than MAP-Elites, and requires several orders of magnitude fewer evaluations. The CMA-ES algorithm was used to produce an optimal design in each bin: with the same number of evaluations required by CMA-ES to find a near-optimal solution in a single bin, SAIL finds solutions of similar quality in every bin.
Maximal covering location problems have efficiently been solved using evolutionary computation. The multi-stage placement of charging stations for electric cars is an instance of this problem which is addressed in this study. It is particularly challenging, because a final solution is constructed in multiple steps, stations cannot be relocated easily and intermediate solutions should be optimal with respect to certain objectives.
