Asynchronously operating systems, such as tapped delay lines, are the designer?s favorite, if high resolution and precision in time are required. Their drawback, however, is that they require extensive calibration, which prohibits, among other things, sporadic recalibration during the mode of operation. Recent research has shown that the tight coupling of two selective high-precision systems inside a single FPGA substantially reduces the required calibration time: it was reduced from several hours to about 30 minutes. But even this method has not solved the problem that human intervention is required for selecting suitable calibration points. The research presented in this poster suggests that a hybrid approach is able to solve this problem: rather than tightly coupling two systems, the present approach employs hybrid elements, called X-BOUNCE, that seamlessly incorporate an X-ORCA element into a BOUNCE element. In the practical experiments, X-BOUNCE has reduced the required calibration time from 30 minutes to one second and has abandoned any human intervention. Furthermore, the proposed X-BOUNCE element can be realized by just one FPGA-LUT, which allows for easy scalability. The results were produced on a Cyclone II FPGA that has implemented 200 X-BOUNCE elements. Unfortunately, some elements exhibit a calibration inaccuracy that can be as large as 300 ps.

Among other things, field-programmable gate arrays (FPGAs) available today contain numerous bit-serial transceivers for communication purposes. Unlike analog modulation schemes, such as quadrature amplitude modulation, bit-serial communication is relatively easy to implement in digital hardware, and is thus usually used for inter FPGA communication. In this view, only the data rate and frequency limit the bandwidth of the circuit. In order to overcome the bandwidth limit, this research proposes a pulse-width modulation (PWM) scheme for data transmission. The information is coded by modulating the length of the high and low voltage parts of the pulse. Although this approach is not new, existing PWM modulators have unsatisfactorial data rates due to their synchronous implementation nature. Therefore, this research implements both the modulator and demodulator by using asynchronous logic. The result is a proof-of-concept comprising two Terasic DE2-70 development boards and a 1 m coaxial cable. Both the PWM modulator and demodulator run at 333 MHz, and pulses are transmitted every 3 ns. Each pulse carries 3 to 4 bits of data. The experimental results indicate an achievable data rate of one gigabit per second, which is about 50 % larger than the FPGA's handbook states.

Non-trivial hardware architectures consist of a significant number of fine-grained modules that communication with each other via dedicated signal lines. In field-programmable gate arrays (FPGAs), these communication lines are provided in forms of global vertical and horizontal routing channels, and are subject to the routing process. Since the effects of physical properties on the signal skew along these lines is well understood, this paper investigates the observable effects on a signal's duty cycle. Practical experiments show that the distortion on the duty cycle progressively increases along such wires (connections) and that in the extreme case, a signal may entirely vanish.

Cochlear implants require a first calibration, which is usually done during surgery. This calibration is normally based on the inspection of the stapedius reflex, which is the contraction of a very tiny muscle in the inner ear. Even though being state-of-the-art, this visual inspection is limited and error prone due to various reasons. This paper proposes a small embedded system that automatically processes the action potentials of the stapedius muscle. Among other things, this process also signals the appearance of the stapedius reflex. Since particular emphasis is to keep computational demands as low as possible, the developed system may be integrated into future version of the implant's speech processor.

Fuzzy Cognitive Maps (FCMs) are well-suited techniques to model the behavior of complex systems. But, since they depend on the quality of human knowledge they are prone to subjective false estimations. Moreover, even domain experts can fully describe only less complex systems which, in turn, limits the effectiveness of FCMs in real world applications. Thus, a technique is needed that, on the one hand, eases the use and, on the other hand, optimizes the formal structure of an FCM. Against that background, this article elucidates the use of different bio-inspired algorithms to optimize the application of the FCMs: Evolutionary Strategies (ES), Particle Swarm Optimization (PSO), and Big Bang-Big Crunch (BB-BC). The proposed approaches are applied to a novel valuation technique using FCMs. Besides principal feasibility, we demonstrate its potential to facilitate the use of vague and incomplete expert knowledge in the valuation of complex investment opportunities such as, but not limited to, young high-technology ventures.

Recent research has developed a new, entirely digital architecture, called X-ORCA, that determines the phase shift of two periodic signals with a resolution as good as about 20 ps. This paper incorporates the X-ORCA system into a wireless experimental setup to form a localization system. The practical experiments utilize a 2.484 GHz transmitter and run the X-ORCA core on a Cyclone II FPGA. The results indicate that this simple localization system easily yields a spatial resolution in the sub-millimeter range.

Previous research has proposed several platforms for the direct hardware implementation of neural networks. Most of these platforms employ a small number of special-purpose processors that emulate a given number of neurons. By contrast, this paper proposes a light-weight hardware model, called time coded output (TiCO) neuron, which codes its internal activation as the duration of its output pulses. Since the implementation of such a neuron requires only about 200 logical elements, standard state-of-the-art field-programmable gate arrays may host several hundreds of neurons on a single chip. Due to its output coding and its low footprint, TiCO neurons are particularly suited for the usage in embedded systems.

Normally, research on evolutionary computation applies its algorithms to the solution or optimization of some technical or mathematical problems. But for some technical tasks, such as localization, nature seem to provides optimal solutions. This paper discusses how the barn owl auditory system can be conceptually realized on a digital system, such as a field programmable gate array. This adapted system yields a time resolution as small as 20 ps, even though it is clocked at only 85 MHz, which corresponds to a duty cycle of about 12 ns. The system achieves this result by copying the natural role model's core principles, i.e., employing a large number of simple, slowly operating processing elements, which are all connected to two passive wires, which induce only a very small additional time delay; these properties are the result of a natural evolutionary process that has taken millions of years.

Venture valuation is an essential technique in economics. Depending on the reason for a valuation, there are several methods applicable. It becomes even more complex when it comes to the valuation of companies that possess mainly intangible assets and whose success lies in the future. In this article a further step towards a Fuzzy Method is presented and discussed that allows the determination of a relevant parameter for a widely used valuation method. We focus on the application of a novel methodological framework to convert Fuzzy Cognitive Maps into Fuzzy rule-based Inference Systems. We demonstrate the principal feasibility and elucidate a method to optimize the functionality.

The nucleus laminaris of the barn owl auditory system is quite impressive, since its underlying time estimation is much better than the processing speed of the involved neurons. Since precise localization is also very important in many technical applications, this paper explores to what extent the main principles of the nucleus laminaris can be implemented in digital hardware. The first prototypical implementation yields a time resolution of about 20 ps, even though the chosen standard, low-cost device is clocked at only 85 MHz, which leads to an internal duty cycle of approximately 12 ns.

Even though evolutionary algorithms have been applied to a vast number of problems, they are still not always applicable. As an example, this paper briefly discusses smart appliances ensembles, which are characterized by an unknown, dynamically changing number of actuators and sensors. To even cope with such problems, research has previously presented AGE-P, an entirely distributed evolutionary algorithm that allows actuators as well as sensors to join and leave an ensemble at any time without any notice. This paper presents an extension that makes AGE-P suitable for environments that cannot provide reliable data communications, as is often the case in real-world scenarios. The real-world results indicate that the proposed self-synchronization scheme resolves the reliability issue while retaining AGE-P's open system character.

In the context of building automation, one of the many tasks is to localize and identify thousands of sensors and actuators, such as smoke detectors, switches, light bulbs, and window blinds. Since performing this task by hand is too costly and error-prone, this paper proposes a simple, semi-automatic system, called WiiLoc, for this task. WiiLoc employs two infrared cameras, a three-dimensional acceleration sensor, and an electronic compass, and yields a localization error of as low as 2% with respect to the distance to the object. SinceWiiLoc employs only simple standard components, such as the Wii remote control, the costs will be below 150 Euro.

Many everyday life activities require precise localization information. Navigation by means of the global positioning system (GPS) is a well-known example. Despite its broad usage, this localization method cannot be utilized in all application areas. Logistics, factory and laboratory automation, as well as warehouse management, for example, require the low-cost localization of various objects with a precision of a few centimeters, particularly in indoor environments. For this application domain, this paper proposes a new procedure to measures the relatively distance difference between two points (e.g., transmitter) to a third point (e.g., receiver) with a low resource complexity. The procedure based on the superposition of particularly parametrized beats. In addition, this paper also presents a preliminary theoretical analysis as well as some laboratory experiments.

Field-programmable gate arrays provide a very flexible implementation platform that supports the development of tamper-proof networked embedded systems. In order to achieve a secure mode of operation, many of these systems employ keyed-hash message authentication code, to provide security against intentional tampering. Since this algorithm is of high computational complexity, this paper utilizes various hardware-software co-design techniques for its implementation, and compares the results to standard software. The considered techniques vary in the degree of required design expertise and of how (software) functionalities are implemented in hardware logic.

Field-programmable gate arrays provide a flexible and easy-to-configure implementation platform that supports the development of tamper-proof networked embedded systems. Many of these systems employ the Advanced Encryption Standard algorithm in order to achieve a secure mode of operation. Since this algorithm is of a high computational complexity, this paper utilizes various hardware-software co-design techniques for its implementation. These techniques vary in the degree of required design expertise, the degree of how (software) functionalities are implemented in hardware logic, and the achievable speedup, which is about 12 to approximately four thousand.

Tapped delay lines are chain-like structures, which are able to measure short time intervals. Due to their homogeneous structure, they are particularly suited for being implemented on field-programmable gate arrays. But unfortunately, the attainable resolution in time is inherently limited to the average processing speed of the chainpsilas elements. As an alternative, this paper proposes a new architecture, called BOUNCE, in which all processing elements run in parallel. With its inherent parallelism, BOUNCE yields a resolution that depends on the variation, i.e., the differences, among the elementspsila processing speeds. The first prototype was implemented on an ALTERA StratixII 2S60 board Even though this board is clocked at only 85 MHz, the prototype yields a resolution of about 100ps.

Many everyday life activities require precise localization information. Navigation by means of the Global Positioning System (GPS) is a well-known example. Despite its broad usage, this localization method cannot be utilized in all application areas. Logistics, factory and laboratory automation, as well as warehouse management, for example, require the low-cost localization of various objects with a precision of a few centimeters, particularly in indoor environments. For this type of applications, this paper proposes a new localization procedure that is based on the superposition of particularly parameterized beats. In addition, this paper also presents a preliminary theoretical analysis as well as some laboratory experiments.

Tapped delay lines are chain-like structures, which are able to measure short time intervals. Due to their homogeneous structure, they are particularly suited for being implemented on field-programmable gate arrays. But unfortunately, the attainable resolution in time is inherently limited to the average processing speed of the chain elements. As an alternative, this paper proposes a new architecture, called BOUNCE in which all processing elements run in parallel. With its inherent parallelism, BOUNCE yields a resolution that depends on the variation, i.e., the differences, among the elementspsila processing speeds. The first prototype was implemented on an ALTERA StratixII 2S60 board. Even though this board is clocked at only 85 MHz, the prototype yields a resolution of about 150 ps.

Currently the developing process for enterprise applications is improved by the Service Oriented Architecture (SOA) paradigms. With SOAs the creation of modular and clearly defined software architectures at a high grade of interoperability and reusability is possible. For resource constraint networked devices the Devices Profile for Web Services (DPWS) specification adapt the SOA paradigms to create a framework for interoperable and standardized communication between embedded devices. This paper explains how special parts of DPWS can be adapted to provide real-time capabilities. So this paper shows how developers of real-time software can use the advantages of Web services and which adaptations are necessary to create Web services with real-time capabilities.

Print media are still very important for everyone's daily life. Current efforts are concerned with the application of the well-established offset-printing technology to other media, particularly cardboards, which require some substantial adaptations. To this end, this paper proposed a new specific pre-processing stage. This pre-processing stage can be configured by several parameters. This paper optimizes these parameter settings by using evolution strategies. It turns out that this optimization reduces the required energy and the number of wrongly generated pixels by about 15%, respectively.

A raytracer calculates how a camera would observe a potentially complex scene consisting of numerous objects and light sources. If all the objects are modeled by n primitives, e.g., triangles, the runtime of a software raytracer scales at least logarithmically. This scaling behavior effectively imposes limitations on the scene's complexity, its size, and the raytracer's real-time capabilities. As an alternative, this paper proposes a parallel hardware raytracing machine. A prototypical implementation on a field-programmable gate array, as offered by markets today, validates that this machine achieves rendering in constant time, regardless of both the scene's size and its complexity.

In object-oriented real-time computing scenarios, particularly where Corba is used in embedded systems with resource constraints, developers and system architects often utilize Corba object references as authorization tokens. This paper investigates the security of this method in principle, and it presents the results of the analysis of the work and computing effort necessary for a potential attacker to fabricate Corba object references to existing objects for the purpose of illegitimately gaining access to these objects at the instance of seven widely used Corba products

Evolutionary gradient search (EGS) is an approach to optimization that combines features of gradient strategies with ideas from evolutionary computation. Recently, several modifications to the algorithm have been proposed with the goal of improving its robustness in the presence of noise and its suitability for implementation on parallel computers. In this paper, the value of the proposed modifications is studied analytically. A scaling law is derived that describes the performance of the algorithm on the noisy sphere model and allows comparing it with competing strategies. The comparisons yield insights into the interplay of mutation, multire combination, and selection. Then, the covariance matrix adaptation mechanism originally formulated for evolution strategies is adapted for use with EGS in order to make the algorithm competitive on objective functions with large condition numbers of their Hessians. The resulting strategy is evaluated experimentally on a number of convex quadratic test functions.

This paper presents the implementation of a genetic algorithm based path planning on RoboCup's small-size league robots. The algorithm is adjusted to the resource constraints of micro controllers that are used in embedded environments. Because path planning on mobile robots is a continuous process, the path planning runs until the robot arrives its destination. Hereby, the path is updated to environmental changes, such as moving obstacles.

Communication networks today are facing an ever increasing network traffic as well as raising quality-of-service agreements, which together demand for high performance network routers. Since a router has to search a large set or routing rules for every incoming packet, it normally utilizes efficient search mechanisms, such as trees or hash tables. This paper evolves hash functions directly in hardware and also discusses an improved initialization process. On a benchmark test consisting of 65,536 routing rules, the final hash functions consume an average of about 1.3 memory accesses for rule searching for every incoming data packet.

State-of-the-art technologies in very large scale integration (VLSI) aim at the realization of fast energy-efficient circuits. Recent technological achievements offer a design parameter with which both the processing speed and power consumption of every single gate can be fine tuned. With respect to this design parameter, a VLSI design constitutes a multi-dimensional multi-modal optimization problem. Since existing algorithms yield only suboptimal designs, this paper investigates how genetic algorithms perform in this application domain. It turns out that genetic algorithms are able to reduce the power consumption by about 10-40%.

While being very successful in everyday life, GPS-based localization systems exhibit limited performance under trees, behind walls, and in closed rooms, and sometimes induce costs that are way too high to be acceptable for some applications. This paper addresses these problems by exploring an alternative localization system, which is based on image processing and does most of the processing in analog. Because of its low-cost and low-energy properties, this system might also be interesting for other application areas, such as sensor networks.

In the RoboCup small-size league, the positions of the robots are commonly calculated via a camera mounted above the field and different kinds of artificial intelligence methods hosted on a PC outside the field. In this case, the robot's position must be predicted because of the various time delays until the position data arrived at the robot. This paper focuses on the use of local sensors attached to the robot, and a neural network to estimate the actual robot position. This paper shows how local sensors can compensate for the effect of latency times and how the actual robot position can be estimated. Slip and friction effects that cannot be measured with local sensors can be adjusted by a neutral network, for which the implementation is adapted to the low resources available on the robot.

In many application areas, robots most suitably employ classical PID controllers and the like. In the field of autonomous mobile robots, however, further adaptation features are required in order to adapt to dynamically changing environmental conditions. In recent contests, the particular research area of soccer-playing robots, called RoboCup, has observed the emergence of omnidirectional driven robots. Such drives consist of three independently controllable motors with which a robot can simultaneously perform both translational movements and rotations, which yield a significant advantage in soccer games. This paper describes how a Kohonen-feature-map-based neural network is able to learn the required capabilities and how to adapt to changing environmental conditions

Obstacles are but pleasing for many aspects of large real-world sensor networks. Among other things, the presence of obstacles distort the sensor node localization process and might lead to costly routing because of unnecessary detours and/or dead ends. In order to relieve these problems, this paper proposes a distributed obstacle localization algorithm, called DOLfor short, in which the sensor nodes interact with each other mostly locally. The proposed algorithm is very resource and communication efficient in that all sensor nodes send only a small number of additional messages. Finally, the sensor network fine tunes the employed routing algorithm.

The precise and cost-efficient localization is of particular interest in large wireless sensor networks. Ap- proximative algorithms, such as coarse-grained localization with centroid determination, require only very little com- putational resources (in terms of hardware components) but lack high precision. This paper discusses a few learning schemes, which improve the localization errors by several orders of magnitude at negligible extra costs. Furthermore, these (local) learning schemes makes it possible to apply these localization algorithms to a new class of applications.

It might come to a surprise, if some properly designed evolutionary algorithms might not be able to sufficiently optimize simple, continuous, unimodal objective functions. Keeping in mind the evolutionary algorithms' high performance especially in the field of global function optimization, getting stuck (also known as premature convergence) at suboptimal function values is attributed mostly to the presence of distracting local optima. This paper describes some examples in which such a behavior occurs. It also gives some explanations of the underlying reasons. These explanations indicate, as a conclusion, that premature convergence might happen very well at continuous, unimodal functions (and consequently in real-world applications).

Most experiments in research on autonomous agents and mobile robots are performed either in simulation or on robots with static physical properties; evolvable hardware is hardly ever used. One of the very rare exceptions is the eyebot on which Lichtensteiger and Eggenberger (1999) have evolved simplified insect eyes. Even though substantially improved, the evolutionary models currently applied still lack both scalability and noise resistance. To tackle these problems, this paper proposes a biologically-inspired force model for this class of real-world applications. The simulation results clearly indicate that this model provides a significant improvement over existing limitations. Furthermore, this paper argues that the force model is of more general utility.

This paper reviews some attempts that exploit a phenomenon, also known as motion parallax, to estimate the distance of closest approach of a moving object. Despite their success, the existing evolutionary methods lack some desirable properties, such as reasonable scalability and online learning. To overcome these practically-relevant limitations, this paper proposes a new model that is based on Hebbian learning. Due to its scalability and online learning capabilities, this model is especially suited to mobile robots.

Biological evidence strongly suggests that insects utilize visual cues for their navigation tasks. This paper discusses the evolution of a simple controller for visual homing by means of evolutionary algorithms. The application is representative for a class of (real world) problems, for which the choice of the fitness function is non-trivial, since the data are not known in advance. For this class of problems, recombination has a much higher influence on the convergence than previously assumed. We show how convergence rates comparable to those of neural network learning algorithms can be achieved.

Most experiments in research on autonomous agents and mobile robots are performed either in simulation or on robots with static physical properties; evolvable hardware is hardly ever used. One of the very rare exceptions is the eyebot on which Lichtensteiger and Eggenberger have evolved simplified insect eyes. Even though substantially improved, the evolutionary models currently applied still lack both scalability and noise-resistance. To tackle these problems, this paper proposes a biologically-inspired force model for this class of real-world applications. The simulation results clearly indicate that this model provides a significant improvement over existing limitations. Furthermore, this paper argues that the force model is of more general utility.

Object avoidance is a fundamental task of autonomous, mobile robots. For this task, the pertinent literature proposes various architectures, which vary from simple Braitenberg vehicles to camera-lens systems inspired by the compound eyes of insects. Due to certain hardware limitations, existing research resorts to prespecified sensor systems that remain fixed during all experiments and does modifications only in the software components of the controllers. By contrast, this paper is about the direct evolution of an artificial compound eye in hardware. The hardware consists of a particular robot that is able to autonomously modify the angular positions of 16 light sensors. Even though first experiments have been successful in evolving some solutions by means of evolutionary algorithms, they have also indicated that systematic comparisons between different evolutionary algorithms and codings schemes are required in order to speed up the evolutionary process. This paper summarizes some comparative simulation studies and validates their achievements on a physical robot. It turns out that these simulation studies can help to drastically improve the evolution of the eye's morphology with respect to both convergence speed and robustness if certain critical simulation parameters (e.g., noise level) are adopted from the physical robot.

Recent literature on genetic algorithms provides a controversial discussion on the efficiency of this particular class of randomized optimization procedures; despite several encouraging empirical results, recent theoretical analyses have argued that in most cases, the runtime behavior of genetic algorithms is increased by at least a factor of ln(n) with n denoting the number of parameters to be optimized. It has been argued that these inefficiencies are due to intrinsic resampling effects. As a result of these theoretical considerations, a deterministic genetic algorithm has been suggested as a theoretical concept. Since its proposition, informal discussions have been raised concerning some implementation details as well as efficacy issues. Since some implementation details are a bit tricky, this paper discusses some of them in a pseudo programming language similar to Pascal or C. In addition, this paper presents two possible variants in detail and compares their runtime behavior with another fairly established procedure, the breeder genetic algorithm. It turns out that on widely-used test functions, the deterministic variants scale strictly better. Furthermore, this paper discusses some specific fitness functions on which random algorithms yield better worst-ease expectations than deterministic algorithms; but both types require constant time on average, i.e., one function evaluation.

For the evaluation of evolutionary algorithms, the research community has devised numerous (artificial) test functions. In so doing, increasing emphasis has been devoted to "hard" problems, especially multimodal functions. This paper revisits some of the real-world applications originally used by Rechenberg (1997) and Schwefel (1997), and subsequently designs some similar artificial functions that can be conveniently used in computer experiments. Since "standard" evolutionary algorithms exhibit some performance difficulties on these test functions, this paper explores a specially-adapted algorithm, the optimize-and-refine procedure, which is an evolution strategy variant and inspired by common engineering methods. In the experiments, it turns out that on real-world-inspired test functions, the optimize-and-refine procedure yields a considerable speedup by exploiting certain constraints.

Previous research has proposed the distributed adaptive control architecture as a self-organizing controller for autonomous mobile robots. Despite its successes, practical experience has indicated several options for improvement. The paper shows how proprioceptive sensors can be used to develop controllers for real-world robots, such as Khepera, that behave very robust with respect to dynamically changing environments and sensor failures. Proprioceptive patterns refer to stimuli arising within an organism, and its usage is largely motivated by their omnipresence in nature.

Explores the use of hand gestures as a means of human-computer interactions for virtual reality applications. For the application, specific hand gestures, such as "fist", "index finger" and "victory sign", have been defined. Most existing approaches use various camera-based recognition systems, which are rather costly and very sensitive to environmental changes. In contrast, this paper explores a data glove as the input device, which provides 18 measurement values for the angles of different finger joints. The paper compares the performance of different neural network models, such as backpropagation and radial-basis functions, which are used by the recognition system to recognize the actual gesture. Some network models achieve a recognition rate (training as well a generalization) of up to 100% over a number of test subjects. Due to its good performance, this recognition system is the first step towards virtual reality applications in which program execution is controlled by a sign language.

Classical gradient methods and evolutionary algorithms represent two very different classes of optimization techniques that seem to have very different properties. This paper discusses some aspects of some "obvious" differences and explores to what extent a hybrid method, the evolutionary-gradient-search procedure, can be used beneficially in the field of continuous parameter optimization. Simulation experiments show that on some test functions, the hybrid method yields faster convergence than pure evolution strategies, but that on other test functions, the procedure exhibits the same deficiencies as steepest-descent methods.

The development of a sound theory that predicts and verifies existing evolutionary algorithms (EA) is one of the most important research issues in the field today. In mathematical proofs, the assumption of spherical symmetry is probably one of the most widely used simplifications. This paper discusses the extent to which spherical symmetry is appropriate for certain EAs. It turns out that spherical symmetry leads to simplifications in (self-adaptive) EAs but seems inappropriate for certain genetic algorithm variants, since small mutation rates bias a search algorithm toward the coordinate axes. This paper also argues that current test suites are weak in that they do not provide problems with significant epistasis that describes the interaction between different parameters. Consequently, when using an empirical test for pushing existing theory beyond its limits, benchmark functions should include more epistatic interaction or at least should use coordinate rotations.