• Two scenarios where smart production planning assistance improves human workflows • A questionnaire to assess levels of explainability needed to implement this system • Survey results by a group of active production planners (n=11) • Light levels of XAI are a strict requirement for an AI-based smart production system • Higher explainability is requested but should not harm performance In recent years, intelligent systems have increased their capabilities greatly increasing their practical applicability. However, for the foreseeable future, such AI-powered agents will not act autonomously but assist a human that will ultimately responsible. Here, the explainability of agents’ suggestions becomes paramount to provide trust and acceptance by their human co-workers. For the field of stochastic/evolutionary optimization, it has not yet been investigated what levels of explainability real human stakeholders without deep technical knowledge of these systems actually request. In this article, we report an exploratory case study where we questioned a group of production planners ( n = 11 ) about their needs for AI assistance and what types of explanations they would require to integrate AI into their day-to-day work-flow and still feel comfortable with the cooperation. While five participants expect their individual position to be threatened by these systems in the mid-term, all participants agree that AI is beneficial for safeguarding the location against competitors or migration. We find that AI-based assistance is requested to a large degree across all age groups and that stakeholders greatly request explainability of the agent’s recommendations. From this real-world empirical evidence it becomes evident that implementing explainable optimization in production planning is a crucial next step towards Industry 5.0.

Global Navigation Satellite Systems (GNSS) are used on a daily basis, providing Positioning, Navigation and Timing (PNT) services for various applications ranging from smartphones over the financial sector up to areas such as aviation and space. Classical GNSS constellations positioned in Medium Earth Orbit (MEO) often experience reduced performance in areas of low visibility like forests and cities. To rectify this, augmentation constellations are deployed, improving the provided positioning accuracy. Recent proposals for augmentation systems have often been based in Low Earth Orbit (LEO), which, for global coverage, require a large number of satellites and are complex to design due to dependencies, coverage requirements and the large search space. This makes the constellation design problem well-suited for applying Genetic Algorithms (GA) to find an optimal solution. However, previous research has only addressed highly constrained versions of the problem. This paper presents an approach for applying GAs to constellation designs with a large search space. In particular, the focus is on the description of the multi-objective fitness function and the simulation necessary for its evaluation, options for the solution encoding, and a discussion of algorithmic features applicable in this scenario.

Deterministic Wireless Sensor Networks over IEEE 802.15.4 can provide latency-bounded transmission of flows, which is an important enabler for current and future Internet of Things (IoT) use cases. To realise such networks, a viable routing and scheduling solution must be found that can accept all given flows and maintain their latency requirements. The joint routing and scheduling (JRaS) problem promises optimal routing and scheduling decisions-however, at the expense of very high computation times. To overcome this issue, we propose efficient modifications to the separate routing and scheduling problems, such that we can obtain a success rate similar to the optimal JRaS approach. However, our solutions can be found in much less time. We conduct extensive performance evaluations for different problem complexities and found a speedup in the range of 3.03× up to 6.09× compared to JRaS while having almost no statistical difference in success rates.

Rule-based machine learning (RBML) models are often presumed to be very beneficial for tasks where explainabil-ity of machine learning models is considered essential. However, their models are only really explainable as long as their rule sets are compact. This leads to the need for an optimizer to take prediction error and rule count as objectives. Given the highly complex fitness landscape of rule set learning tasks, good hyperparameters of the optimizer as well as their robustness against local minima is detrimental. In this paper, we explore the use of four self-adaptive genetic algorithms (SAGAs) for the optimization of a recent evolutionary RBML system to reduce the number of hyperparameters to tune and hopefully find better minima. To evaluate the advantages, we benchmark against a non-adaptive genetic algorithm (GA) on five real-world data sets. We find-with the support of a rigorous statistical analysis-that some of the SAGAs deliver a suitable alternative, which is easier to handle for non-experts in GA configurations. This is crucial for a wider application of this RBML method.

Modern embedded systems comprise more and more software. This yields novel challenges in development and quality assurance. Complex software interactions may lead to serious performance issues that can have a crucial economic impact if they are not resolved during development. Henceforth, we decided to develop and evaluate a machine learning-based approach to identify performance issues. Our experiments using real-world data show the applicability of our methodology and outline the value of an integration into modern software processes such as continuous integration.

Practitioners frequently encounter the challenge of selecting the best optimization algorithm from a pool of options. However, why not, rather than selecting a single algorithm, let evolution determine the optimal combination of all algorithms? In this paper, we present an approach to algorithm design inspired by a well-known traditional method for coarse-grained hybridization: the heterogeneous island model. Our hyper-heuristic framework represents island models as graphs and identifies optimal island topologies and parameters for specific sets of problem instances. Since the framework operates at the level of metaheuristic algorithms rather than components and incorporates a configuration mechanism directly into the search, it combines concepts from algorithm design, selection, and configuration. The proposed framework is investigated on 24 training sets of varying difficulty and demonstrates its ability to discover complex hybrids. A post-evaluation on real-world constrained optimization problems shows a significant improvement over the algorithms on their own. These results suggest that it is a promising way to design hybrid metaheuristics with minimal manual intervention, given representative training instances, a set of optimization algorithms, and sufficient computational resources.

The Machine Health (MH) sector—which includes, for example, Predictive Maintenance, Prognostics and Health Management, and Condition Monitoring—has the potential to improve efficiency and reduce costs for maintenance and machine operation. This is achieved by data-driven analytics applications, utilising the vast amount of data collected by sensors during machine runtime. While there are numerous possible fields of application, the overall complexity of machines and applications in scientific publications is still low, preventing MH technologies from being implemented in many real-world scenarios. This may be the result of a diffuse understanding of the term complexity in the publications of this field, which results in a lack of focus towards the core problems of real-world MH applications. This article introduces a new way of discerning complexity in data-driven MH applications, enabling an effective discussion and analysis of present and future MH applications. This is achieved by creating a new taxonomy based on observations from relevant literature and substantial domain knowledge. Using this newly introduced taxonomy, we categorise recent applications of MH to demonstrate the usefulness of our approach and illustrate a still-prevalent research gap based on our findings.

Real world datasets might contain duplicate or redundant attributes—or even pure noise—which may not be filtered out by data preprocessing algorithms. This might be problematic, as it decreases the performance of learning algorithms. Cartesian Genetic Programming (CGP) is able to choose its own input attributes by design. Thus, we hypothesize that CGP should be able to ignore redundant or noise attributes. In this work, we empirically show that CGP is indeed able to handle such problematic datasets. For this task, six different datasets are extended with different kinds of redundancies: Duplicated-, duplicated and noised-, and pure noise attributes. Different numbers of unwanted attributes are examined, and we present our results which indicate that CGP is robust against additional redundant or noisy attributes in a dataset. We show that there is no decrease in performance as well as no change in CGP’s convergence behaviour.

Maintenance of complex machinery is time and resource intensive. Therefore, decreasing maintenance cycles by employing Predictive Maintenance (PdM) is sought after by many manufacturers of machines and can be a valuable selling point. However, currently PdM is a hard to solve problem getting increasingly harder with the complexity of the maintained system. One challenge is to adequately prepare data for model training and analysis. In this paper, we propose the use of expert knowledge–based preprocessing techniques to extend the standard data science–workflow. We define complex multi-purpose machinery as an application domain and test our proposed techniques on real-world data generated by numerous machines deployed in the wild. We find that our techniques enable and enhance model training.

This work presents and evaluates a novel modification to existing mutation operators for Cartesian Genetic Programming (CGP). We discuss and highlight a so far unresearched limitation of how CGP explores its search space which is caused by certain nodes being inactive for long periods of time. Our new mutation operator is intended to avoid this by associating each node with a dynamically changing weight. When mutating a connection between nodes, those weights are then used to bias the probability distribution in favour of inactive nodes. This way, inactive nodes have a higher probability of becoming active again. We include our mutation operator into two variants of CGP and benchmark both versions on four Boolean learning tasks. We analyse the average numbers of iterations a node is inactive and show that our modification has the intended effect on node activity. The influence of our modification on the number of iterations until a solution is reached is ambiguous if the same number of nodes is used as in the baseline without our modification. However, our results show that our new mutation operator leads to fewer nodes being required for the same performance; this saves CPU time in each iteration.

Electrical motors need specific parametrizations to run in highly specialized use cases. However, finding such parametrizations may need a lot of time and expert knowledge. Furthermore, the task gets more complex as multiple optimization goals interplay. Thus, we propose a novel approach using Bayesian Optimization to find optimal configuration parameters for an electric motor. In addition, a multi-objective problem is present as two different and competing objectives must be optimized. At first, the motor must reach a desired revolution per minute as fast as possible. Afterwards, it must be able to continue running without fluctuating currents. For this task, we utilize Bayesian Optimization to optimize parameters. In addition, the evolutionary algorithm NSGA-II is used for the multi-objective setting, as NSGA-II is able to find an optimal pareto front. Our approach is evaluated using three different motors mounted to a test bench. Depending on the motor, we are able to find good pa rameters in about 60-100%.

64

308

300

Analysing the behaviour of metaheuristics comprehensively and thereby enhancing explainability requires large empirical studies. However, the amount of data gathered in such experiments is often too large to be examined and evaluated visually. This necessitates establishing more efficient analysis procedures, but care has to be taken so that these do not obscure important information. This paper examines the suitability of clustering methods to assist in the characterisation of the behaviour of metaheuristics. The convergence behaviour is used as an example as its empirical analysis often requires looking at convergence curve plots, which is extremely tedious for large algorithmic datasets. We used the well-known K-Means clustering method and examined the results for different cluster sizes. Furthermore, we evaluated the clusters with respect to the characteristics they utilise and compared those with characteristics applied when a researcher inspects convergence curve plots. We found that clustering is a suitable technique to assist in the analysis of convergence behaviour, as the clusters strongly correspond to the grouping that would be done by a researcher, though the procedure still requires background knowledge to determine an adequate number of clusters. Overall, this enables us to inspect only few curves per cluster instead of all individual curves.

Deep-Q-Networks are built in a way that, given a state, they predict the Q-values for the entire action-space. However, given an experience, the training update only incorporates the loss value for a single action-the one that has actually been executed. This is due to the rewards and follow-up states (required for computing the loss via the temporal-difference error) associated with the other actions being unknown. With these missing values at hand, or at least estimates of them, an update over the entire action-space would be possible. We present the Full-Update-DQN which is able to do just that. Sub-losses are weighted to compensate for uncertainty and noise and we are able to show in four different experiments in sparse reward settings, that our approach is able to solve these problems more consistently and even faster than the original approach.

176

In sociotechnical settings, human operators are increasingly assisted by decision support systems. By employing such systems, important properties of sociotechnical systems, such as self-adaptation and self-optimization, are expected to improve further. To be accepted by and engage efficiently with operators, decision support systems need to be able to provide explanations regarding the reasoning behind specific decisions. In this article, we propose the use of learning classifier systems (LCSs), a family of rule-based machine learning methods, to facilitate and highlight techniques to improve transparent decision-making. Furthermore, we present a novel approach to assessing application-specific explainability needs for the design of LCS models. For this, we propose an application-independent template of seven questions. We demonstrate the approach's use in an interview-based case study for a manufacturing scenario. We find that the answers received do yield useful insights for a well-designed LCS model and requirements for stakeholders to engage actively with an intelligent agent.

Learning Classifier Systems (LCSs) are a family of versatile rule-based machine learning algorithms. Despite their long research history, the foundations of most LCSs are still informal due to them having been developed in an ad-hoc manner. An exception to this is the fully Bayesian LCS described by Drugowitsch in his 2008 book. In the present paper we shortly reiterate the central points of that system and then showcase our Python implementation of it, reporting on our attempt at replicating Drugowitsch's empirical results as well as on the results of a preliminary comparison study with the well-known LCS XCSF on the same learning tasks. We are able to replicate parts of Drugowitsch's results and explain the remaining differences. The comparison results make us conclude that the system may be competitive and exhibits unique features. Finally, we identify its current greatest shortcomings and based on those the next steps towards making it more universally usable.

To fill the increasing demand for explanations of decisions made by automated prediction systems, machine learning (ML) techniques that produce inherently transparent models are directly suited. Learning Classifier Systems (LCSs), a family of rule-based learners, produce transparent models by design. However, the usefulness of such models, both for predictions and analyses, heavily depends on the placement and selection of rules (combined constituting the ML task of model selection). In this paper, we investigate a variety of techniques to efficiently place good rules within the search space based on their local prediction errors as well as their generality. This investigation is done within a specific LCS, named SupRB, where the placement of rules and the selection of good subsets of rules are strictly separated in contrast to other LCSs where these tasks sometimes blend. We compare a Random Search, (1,λ)-ES and three Novelty Search variants. We find that there is a definitive need to guide the search based on some sensible criteria, i.e. error and generality, rather than just placing rules randomly and selecting better performing ones but also find that Novelty Search variants do not beat the easier to understand (1,λ)-ES.

The concept of Experience Replay is a crucial element in Deep Reinforcement Learning algorithms of the DQN family. The basic approach reuses stored experiences to, amongst other reasons, overcome the problem of catastrophic forgetting and as a result stabilize learning. However, only experiences that the learner observed in the past are used for updates. We anticipate that these experiences posses additional valuable information about the underlying problem that just needs to be extracted in the right way. To achieve this, we present the Interpolated Experience Replay technique that leverages stored experiences to create new, synthetic ones by means of interpolation. A previous proposed concept for discrete-state environments is extended to work in continuous problem spaces. We evaluate our approach on the MountainCar benchmark environment and demonstrate its promising potential.

Abstract Metaheuristics provide the means to approximately solve complex optimisation problems when exact optimisers cannot be utilised. This led to an explosion in the number of novel metaheuristics, most of them metaphor-based, using nature as a source of inspiration. Thus, keeping track of their capabilities and innovative components is an increasingly difficult task. This can be resolved by an exhaustive classification system. Trying to classify metaheuristics is common in research, but no consensus on a classification system and the necessary criteria has been established so far. Furthermore, a proposed classification system can not be deemed complete if inherently different metaheuristics are assigned to the same class by the system. In this paper we provide the basis for a new comprehensive classification system for metaheuristics. We first summarise and discuss previous classification attempts and the utilised criteria. Then we present a multi-level architecture and suitable criteria for the task of classifying metaheuristics. A classification system of this kind can solve three main problems when applied to metaheuristics: organise the huge set of existing metaheuristics, clarify the innovation in novel metaheuristics and identify metaheuristics suitable to solve specific optimisation tasks.

Touch interfaces are human machine interface (HMI) that can be found in a wide range of products ranging from mobile phones over cars to home appliances.Many of these HMIs measure digital signals which are used to detect touch events.These signals are processed using filters in order to decide whether there is a touch event or not.The filterchain must be functional even if the signal contains heavy noise.Thus a precise calibration of the individual filters is necessary.We employ a genetic algorithm (GA) to choose the filter parameters automatically.We evaluate our approach in a series of experiments which includes simulated as well as real data.We additionally compare our GA with manually calibrated parameters and thereby show the superiority of our method in terms of the accuracy of the calibration provided.A cost-intensive manual calibration can thus be avoided.

Human machine interfaces (HMI) have become a part of our daily lives.They are an essential part of a variety of products ranging from computers over smart phones to home appliances.Customer's requirements for HMIs are rising and so does the complexity of the devices.Several years ago, many products had a rather simple HMI such as mere buttons.Nowadays lots of devices have screens that display complex text messages and a variety of objects such as icons.This leads to new challenges in testing, the goal of which it is to ensure quality and to find errors.We combine a genetic algorithm with computer vision techniques in order to solve two testing use cases located in the automated verification of displays.Our method has a low runtime and can be used on low budget equipment such as Raspberry Pi which reduces the operational cost in practice.

Planning the right grasp pose and motion into it has been a problem in the robotic community for more than 20 years.This paper presents a model-based approach for a Pick action of a robot that increases the automation of FDM based additive manufacturing by removing a produced object from the build plate.We treat grasp pose planning, motion planning and simulation-based verification as separate components to allow a high exchangeability.When testing a variety of different object geometries, feasible grasps and motions were obtained for all objects.We also found that the computation time is highly dependent on the random seed, leading us to employ a system of budgeted runs for which we report the estimated success probability and expected running time.Within the budget, some objects never found feasible picks.Thus, we rotated these objects by 90 • which lead to a substantial improvement in success probabilities.

Multipath Tunnelling (MT) is an effective approach to increase the reliability and throughput of network connections. In contrast to Multipath TCP (MPTCP) or SCTP, it also works for UDP traffic and its deployment is easy. It requires neither full access to all uplink endpoints nor a multipath protocol implementation in every host. Despite its potential, research on MT is sparse. Most known prototypes use either round robin or MPTCP schedulers that are not able to fully utilize the potential of MT. This paper proposes and evaluates LLMT, a novel specific MT packet scheduling algorithm that achieves very low latency and high throughput for reliable and unreliable traffic. In multiple testbed experiments, we compare LLMT with several schedulers including LowRTT, OTIAS and AFMT for different network configurations. For TCP traffic, LLMT gains low latency and high throughput. For fixed rate UDP traffic, jitter and latency are the lowest among all competitors.

Multipath Tunnelling (MT) is an effective approach to increase the reliability and throughput of network connections. In contrast to Multipath TCP (MPTCP) or SCTP, it also works for UDP traffic and its deployment is easy. It requires neither full access to all uplink endpoints nor a multipath protocol implementation in every host. Despite its potential, research on MT is sparse. In this paper we provide a comprehensive Evaluation of several state-of-the-art MT approaches and compare them to Tunnelling-over-MPTCP. In several experiments we compare throughput for reliable traffic as well as jitter for unreliable throughput. We find that HTMT (High Throughput Multipath Tunnelling, a specific MT approach) delivers comparable jitter with a considerably higher throughput for different path configurations.

Multipath Tunnelling (MT) is an effective approach to increase the reliability and performance of network connections. In contrast to Multipath TCP (MPTCP) or SCTP, it also works for UDP traffic and its deployment is easy. It requires neither full access to all uplink endpoints nor a multipath protocol implementation in every involved host. Despite its potential, research on MT is sparse. Most known prototypes use either round robin or MPTCP schedulers that are not able to fully utilise the potential of MT. This paper proposes HTMT, a novel packet scheduling algorithm designated for MT that achieves high throughput for reliable traffic on asymmetric paths. Using the tunnel transport protocol's path estimation, it is able to dynamically adapt to changing subtunnel characteristics. Additionally flow awareness allows HTMT to associate packets to a flow and only send it on subtunnels where it won't overtake any of its predecessors. This avoids packet reordering and its detrimental effect on TCP throughput, without the latency and performance costs of a re-reordering buffer at the tunnel exit. In multiple testbed experiments, we compare HTMT with several schedulers for MT and MPTCP, including LowRTT (current MPTCP default), OTIAS and AFMT. In terms of throughput, HTMT offers similar performance for symmetric paths but outperforms all competitors on asymmetric paths.

318

XCS constitutes the most deeply investigated classifier system today. It offers strong potentials and comes with inherent capabilities for mastering a variety of different learning tasks. Besides outstanding successes in various classification and regression tasks, XCS also proved very effective in certain multi-step environments from the domain of reinforcement learning. Especially in the latter domain, recent advances have been mainly driven by algorithms which model their policies based on deep neural networks, among which the Deep-Q-Network (DQN) being a prominent representative. Experience Replay (ER) constitutes one of the crucial factors for the DQN's successes, since it facilitates stabilized training of the neural network-based Q-function approximators. Surprisingly, XCS barely takes advantage of similar mechanisms that leverage remembered raw experiences. To bridge this gap, this paper investigates the benefits of extending XCS with ER. We demonstrate that for single-step tasks ER yields strong improvements in terms of sample efficiency. On the downside, however, we reveal that ER might further aggravate well-studied issues not yet solved for XCS when applied to sequential decision problems demanding for long-action-chains.

An important component of many Deep Reinforcement Learning algorithms is the Experience Replay which serves as a storage mechanism or memory of made experiences. These experiences are used for training and help the agent to stably find the perfect trajectory through the problem space. The classic Experience Replay however makes only use of the experiences it actually made, but the stored samples bear great potential in form of knowledge about the problem that can be extracted. We present an algorithm that creates synthetic experiences in a nondeterministic discrete environment to assist the learner. The Interpolated Experience Replay is evaluated on the FrozenLake environment and we show that it can support the agent to learn faster and even better than the classic version.

Testing is a crucial part in the development of a new product.Due to the change from manual testing to automated testing, companies can rely on a higher number of tests.There are certain cases such as smoke tests where the execution of all tests is not feasible and a smaller test suite of critical test cases is necessary.This prioritization problem has just gotten into the focus of reinforcement learning.A neural network and an XCS classifier system have been applied to this task.Another evolutionary machine learning approach is the XCSF which produces, unlike XCS, continuous outputs.In this work we show that XCSF is superior to both the neural network and XCS for this problem.

The minimum set cover problem (MSCP) is one of the first NP-hard optimization problems discovered.Theoretically it has a bad worst case approximation ratio.As the MSCP turns out to appear in several real world problems, various approaches exist where evolutionary algorithms and metaheuristics are utilized in order to achieve good average case results.This work is intended to revisit and compare current results regarding the application of metaheuristics for the MSCP.Therefore, a recapitulation of the MSCP and its classification into the class of NP-hard optimization problems are provided first.After an overview of notable approximation methods, the focus is shifted towards a brief review of existing metaheuristics which were adapted for the MSCP.In order to allow for a targeted comparison of the existing algorithms, the theoretical worst case complexities in terms of the big O-notation are derived first.This is followed by an empirical study where the identified metaheuristics are examined.Here we use Steiner triple systems, Beasley's OR library, and introduce a new class of instances.Several of the considered approaches achieve close to optimal results.However, our analysis reveals significant differences in terms of runtime and shows that some approaches may even have exponential runtime.

Commissioning of machines takes up a considerable share of time and money of the total cost of developing a machine.Our project aims at developing an approach to decrease the time needed to commission machines by automating parameter optimisation with the help of formalised expert knowledge.The approach will be developed on the Fused Deposition Modelling (FDM) process, which is an additive manufacturing technique.We pay particular attention to keeping the approach sufficiently abstract to be applied to machines from other domains to benefit its industrial application.

This paper proposes AFMT, a packet scheduling algorithm to achieve adaptive flow-aware multipath tunnelling. AFMT has two unique properties. Firstly, it implements robust adaptive traffic splitting for the subtunnels. Secondly, it detects and schedules bursts of packets cohesively, a scheme that already enabled traffic splitting for load balancing with little to no packet reordering. Several NS-3 experiments over different network topologies show that AFMT successfully deals with changing path characteristics due to background traffic while increasing throughput and reliability.

Self-adaptation has been proposed as a mechanism to counter complexity in control problems of technical systems. A major driver behind self-adaptation is the idea to transfer traditional design-time decisions to runtime and into the responsibility of systems themselves. To deal with unforeseen events and conditions, systems need creativity—typically realized by means of machine learning capabilities. Such learning mechanisms are based on different sources of knowledge. Feedback from the environment used for reinforcement purposes is probably the most prominent one within the self-adapting and self-organizing (SASO) systems community. However, the impact of other (sub-)systems on the success of the individual system’s learning performance has mostly been neglected in this context. In this article, we propose a novel methodology to identify effects of actions performed by other systems in a shared environment on the utility achievement of an autonomous system. Consider smart cameras (SC) as illustrating example: For goals such as 3D reconstruction of objects, the most promising configuration of one SC in terms of pan/tilt/zoom parameters depends largely on the configuration of other SCs in the vicinity. Since such mutual influences cannot be pre-defined for dynamic systems, they have to be learned at runtime. Furthermore, they have to be taken into consideration when self-improving their own configuration decisions based on a feedback loop concept, e.g., known from the SASO domain or the Autonomic and Organic Computing initiatives. We define a methodology to detect such influences at runtime, present an approach to consider this information in a reinforcement learning technique, and analyze the behavior in artificial as well as real-world SASO system settings.

Learning Classifier Systems (LCS) have been strongly investigated in the context of regression tasks and great successes have been achieved by applying the function approximating Extended Classifier System (XCSF) endowed with sophisticated prediction models. In this paper, a novel approach to model a classifier's payoff prediction is proposed. Radial Basis Function (RBF) interpolation is utilized as a new means to capture the underlying function surface complexity. We pose the hypothesis that by the use of a more flexible RBF-based classifier prediction, that alleviates the a priori bias injected via choosing the degree of a polynomial approximation, the classifiers can evolve toward a higher generality by maintaining at least a competitive level of performance compared to the current and probably mostly used state of the art approach - polynomial approximation in combination with the Recursive Least Squares (RLS) technique for incremental coefficient optimization. The presented experimental results underpin our assumptions by revealing that the RBF-based classifier prediction outperforms the n-th order polynomial approximation on several test functions of varying complexity. Additionally, results of experiments with various degrees of noise will be reported to touch upon the proposed approach's applicability in real world situations.

Carbon fiber reinforced plastics (CFRP) play a key role for the production of leightweight structures. Simultaneously, online quality inspection of CFRP becomes more important, especially for environments with high safety standards. In this context, vision systems aim to find defects of different shape, size, contour and orientation. Little effort, however, has been made in detecting defect areas in images taken from the surface of carbon fibers. A common approach for segmenting filament defects are edge detection and thresholding. With every change of material and process adjustments, the filter parameters have to be adapted. In this paper, we propose a cartesian genetic programming (CGP) approach to semi-automatically select the best parameters. This strategy saves time for parameter identification while at the same time increases precision. A test run on randomly selected samples shows how the approach can substantially improve detection reliability.

In this paper, we show how an evolutionary rule-based machine learning technique can be applied to tackle the task of self-configuration of smart camera networks.More precisely, the Extended Classifier System (XCS) is utilized to learn a configuration strategy for the pan, tilt, and zoom of smart cameras.Thereby, we extend our previous approach, which is based on Q-Learning, by harnessing the generalization capability of Learning Classifier Systems (LCS), i.e. avoiding to separately approximate the quality of each possible (re-)configuration (action) in reaction to a certain situation (state).Instead, situations in which the same reconfiguration is adequate are grouped to one single rule.We demonstrate that our XCS-based approach outperforms the Q-learning method on the basis of empirical evaluations on scenarios of different severity.

The increase in mobility leads to a higher number of kilometres driven per vehicle and more delay due to congestion which poses a recent and future problem.Congestion generates growing environmental pollution and more car accidents.We apply machine learning concepts to the task of congestion detection in road traffic.We focus on the extended classifier system XCSR, an evolutionary rule-based on-line learning classifier system.Experiments with real-world detector data demonstrate high accuracy of XCSR for congestion detection on interstates.

Intersection management, routing, and congestion avoidance are key factors for improved mobility and better road network utilisation.Organic Traffic Control (OTC) is a self-organising traffic management system for urban road networks.Its main features are the self-adaptive traffic-responsive signalisation of intersections, the coordination of traffic light controllers, and dynamic route guidance of traffic streams.This paper aims at presenting how the automatic and fully distributed incident detection within OTC works and how OTC makes use of these incident alarms for the automated adaptation of signalisation.

In this paper, we present a realtime fluid simulation based on cellular automata (CAs). The main goal is to demonstrate the performance and extensibility of this approach. To show this, we created a fluid simulation and extended it by simulating different kinds of fluids at the same time with effects like oil foating on water and a focus on short computation time. This makes the simulation interesting for games in VR. In our simulation, we had a high framerate of 100 FPS for a CA running on the CPU with 1763 cells due to parallelization and optimization.

XCSF is a rule-based on-line learning system that makes use of local learning concepts in conjunction with gradient-based approximation techniques. It is mainly used to learn functions, or rather regression problems, by means of dividing the problem space into smaller subspaces and approximate the function values linearly therein. In this paper, we show how local interpolation can be incorporated to improve the approximation speed and thus to decrease the system error. We describe how a novel interpolation component integrates into the algorithmic structure of XCSF and thereby augments the well-established separation into the performance, discovery and reinforcement component. To underpin the validity of our approach, we present and discuss results from experiments on three test functions of different complexity, i.e. we show that by means of the proposed strategies for integrating the locally interpolated values, the overall performance of XCSF can be improved.

Open, distributed systems face the challenge to maintain an appropriate operation performance even in the presence of bad behaving or malicious agents.A promising mechanism to counter the resulting negative impact of such agents is to establish technical trust.In this paper, we investigate strategies to improve the efficiency of trust mechanisms regarding the isolation of undesired participants by means of reputation and accusation techniques.We demonstrate the potential benefit of the developed techniques within simulations of a Trusted Desktop Computing Grid. TRUSTED DESKTOP GRIDWe use an open, distributed Trusted Desktop Grid (TDG) as application scenario to show and prove the effective application of distributed algorithms as well as Organic Computing (Müller-Schloer et al., 2011) methods.In this scenario, we use an open and heterogeneous Multi-Agent System (MAS) and we do not assume benevolence.The agents in the system cooperate to gain an advantage.The mechanism determining this cooperation is Trust.Because of the openness of the system, different agents may try to exploit it.They may be uncooperative, malfunctioning or even malicious.An agent, which acts on behalf of the user, is submitting jobs it wants to have calculated (Klejnowski, 132

Robust and optimized agent behavior can be achieved by allowing for learning mechanisms within the underlying adaptive control strategies.Therefore, a classic feedback loop concept is used that chooses the best action for an observed situation -and learns the success by analyzing the achieved performance.This typically reflects only the local scope of an agent and neglects the existence of other agents with impact on the reward calculation.However, there are significant mutual influences among agents population.For instance, the success of a Smart Camera's control strategy depends (in terms of person detection or 3D-reconstruction) largely on the current strategy performed by its spatially neighbors.In this paper, we compare two concepts to consider such influences within the adaptive control strategy: Distributed W-Learning and Q-Learning in combination with mutual influence detection.We demonstrate that the performance can be improved significantly, if taking detected influences into account.

Mobile communication and data services face a rapid growth and result in an ever increasing demand of bandwidth. As alternative to investing in the infrastructure by splitting cells down to nano or pico scale, this paper introduces and evaluates a concept for traffic data offloading to ad-hoc communication in terms of device-to-device communication. The idea is to combine infrastructure-aided route discovery with local mobile ad-hoc network communication to handle traffic with a local focus, i.e. intra-cell or neighbouring-cell traffic. We analyse the concept in terms of latencies and packet delivery ratio by using Omnet++ simulations. We demonstrate that the approach comes with low overhead and allows for an efficient distribution of traffic depending on the expected communication range.

Increasing mobility and raising traffic demands lead to serious congestion problems. Intelligent traffic management systems try to alleviate this problem with optimised signalisation of traffic lights and dynamic route guidance (DRG). One solution for the former aspect is Organic Traffic Control (OTC), offering a self-organised, decentralised traffic control system. Based on OTC, this paper presents two proactive routing protocols, resembling techniques known from the Internet domain, applied to the traffic routing problem: Distance Vector Routing and Link State Routing. These protocols were adapted to utilise forecasts of traffic flows to offer anticipatory and time-dependant DRG for road users. The efficiency of these protocols is demonstrated with simulations of two Manhattan-type road networks under disturbed and undisturbed conditions. The results indicate their benefit in terms of lower travel times and emissions, even under low compliance rates.

Self-organized systems typically consist of distributed autonomous entities. An increasing part of such systems is characterized by openness and heterogeneity of participants. For instance, open desktop computing grids provide a framework for unrestrictedly joining in. However, openness and heterogeneity present severe challenges to the overall system’s stability and efficiency since uncooperative and even malicious participants are free to join. A promising solution for this problem is to introduce technical trust as a basis; however, in turn, the utilization of trust opens space for negative emergent behavior. This article introduces a system-wide observation and control loop that influences the self-organized behavior to provide a performant and robust platform for benevolent participants. Thereby, the observation part is responsible for gathering information and deriving a system description. We introduce a graph-based approach to identify groups of suspicious or malicious agents and demonstrate that this clustering process is highly successful for the considered stereotype agent behaviors. In addition, the controller part guides the system behavior by issuing norms that make use of incentives and sanctions. We further present a concept for closing the control loop and show experimental results that highlight the potential benefit of establishing such a control loop.

Desktop Computing Grids provide a framework for joining in and sharing resources with others. The result is a self-organised system that typically consists of numerous distributed autonomous entities. Openness and heterogeneity postulate severe challenges to the overall system’s stability and efficiency since uncooperative and even malicious participants are free to join. In this paper, we present a concept for identifying agents with exploitation strategies that works on a system-wide analysis of trust and work relationships. Afterwards, we introduce a system-wide control loop to isolate these malicious elements using a norm-based approach – due to the agents’ autonomy, we have to build on indirect control actions. Within simulations of a Desktop Computing Grid scenario, we show that the intelligent control loop works highly successful: these malicious elements are identified and isolated with a low error rate. We further demonstrate that the approach results in a significant increa se of utility for all participating benevolent agents.

Network protocols are deployed in highly dynamic environments, but typically configured with a static setup of configurations. The Organic Network Control system (ONC) has been developed to alter protocol configurations at runtime. ONC is equipped with online learning capabilities and safety considerations. This paper presents a first TCP-based study on how this approach can be applied to end-to-end protocols and simultaneously alleviating the drawbacks of a simulation-based optimisation procedure. The paper explains the developed algorithm and demonstrates the benefit of the solution in an Omnet++ scenario.

Grid Computing Systems are examples for open systems with heterogeneous and potentially malicious entities. Such systems can be controlled by system-wide intelligent control mechanisms working on trust relationships between these entities. Trust relationships are based on ratings among individual entities and represent system-wide information. In this paper, we propose to utilise a normative approach for the system-level control loop working on basis of these trust values. Thereby, a normative approach does not interfere with the entities’ autonomy and handles each system as black box. Implicit rules already existing in the system are turned into explicit norms – which in turn are becoming mandatory for all entities. This allows the distributed systems to derive the desired behaviour and cooperate in reaction to disturbed situations such as attacks.

The utilisation of cell phone networks increases continuously, especially driven by the introduction of new mobile services and smart phones. Network operators can follow two directions to deal with the problem: either install new hardware or increase the efficiency of the existing infrastructure. This paper presents a novel algorithm to improve the efficiency of current networks by allowing for a self-organised load-dependent reconfiguration of antennas. The algorithm is capable of identifying hotspot traffic, assigning this to a neighbouring cell, and learning the best strategy at runtime. This leads to a self-improving intelligent control mechanism. The simulation-based evaluation results demonstrate the potential benefit, while simultaneously keeping the hardware’s deterioration at a comparable level.

In this paper we present an approach to designing swarms of autonomous, adaptive robots. An observer/controller framework that has been developed as part of the Organic Computing initiative provides the architectural foundation for the individuals' adaptivity. Relying on an extended Learning Classifier System (XCS) in combination with adequate simulation techniques, it empowers the individuals to improve their collaborative performance and to adapt to changing goals and changing conditions. We elaborate on the conceptual details, and we provide first results addressing different aspects of our multi-layered approach. Not only for the sake of generalisability, but also because of its enormous transformative potential, we stage our research design in the domain of quad-copter swarms that organise to collaboratively fulfil spatial tasks such as maintenance of building facades. Our elaborations detail the architectural concept, provide examples of individual self-optimisation as well as of the optimisation of collaborative efforts, and we show how the user can control the swarm at multiple levels of abstraction. We conclude with a summary of our approach and an outlook on possible future steps.

System engineers are facing demanding challenges in terms of complexity and interconnectedness. Current research initiatives like Organic or Autonomic Computing propose to increase the freedom of the system to be developed using concepts like adaptivity and self-organisation. Adaptivity means that for such systems we defer a part of the design process from design time to runtime. Therefore, we need a runtime infrastructure which takes care of runtime modifications. This paper presents a meta-design process to develop adaptive systems and parametrise the runtime infrastructure in a unified way. To demonstrate the proposed design process, we applied it to a communication scenario and evaluate the resulting system in a realistic setting.

Today, high performance and feature rich surveillance systems are very costly as they require an expensive set of infrastructure components. As a consequence, such systems including, e.g., complex automatic video content analysis, are restricted to large scale applications, such as airports or train stations. In smaller settings, e.g. in shop surveillance, mostly low-cost display or record-only systems are in use. In this position paper we propose to combine two well-known approaches in order to make Intelligent Video Surveillance applicable and affordable in small to medium-scale scenarios. The proposal includes to combine the concept of Smart Cameras, i.e. cameras equipped with local processing resources, with the ideas of Cloud Computing, i.e. the on-demand provisioning of computing and storage services for complex calculations, and the management of large amounts of data, i.e. video storage. The former allows for the cost effective pre-processing of video data close to the sensor , while using the latter concept does not require large initial investments into expensive infrastructure components such as powerful compute servers. The paper presents research issues of the necessary system design, including precise system goal and system model aspects. Based on this, we discuss several research issues required to be addressed for solving the overall goals.

In this paper we present an open multi-agent based Desktop Grid System that improves the performance of cooperative clients while decreasing the performance of non-cooperative clients as an incentive for good conduct. This is achieved by trust-based job client and worker algorithms that take into account the local and global history of the credibility and reliability experiences between agents, as well as their current state. We show with simulation results that the algorithms provide performance increases by applying standard Desktop Grid performance metrics and evaluating the system in several scenarios with a varying number of disturbances to the system. In this paper, we focus on the task of minimising wasteful computation and thus increasing the benefit of grid participation for the agents.

Systems capable of adapting to changing conditions have gained increasing attention in the last decade. Typically, vast situation and configuration spaces do not allow for using a predefined set of adaptation policies. Based on the principles of Organic Computing, a 3-layered learning architecture has been developed which is capable of coping with the problem by enabling self-adaptation and self-improvement. A major focus has been set on developing safety-based and efficient machine learning concepts founding on evolutionary search heuristics and rule-based learning. The general design has been successfully applied to safety critical real-world applications like urban traffic control and data communication protocols. This paper investigates the question for which class of technical systems the design is applicable. Thus, a generalised model based on mathematical functions is introduced and evaluated. The evaluation demonstrates that the approach works well for systems where the configuration spaces are steadily representable by functions of the situation space. This statement holds even in the presence of noise.

The Organic Computing (OC) Initiative deals with technical systems, that consist of a large number of distributed and highly interconnected subsystems. In such systems, it is impossible for a designer to foresee all possible system configurations and to plan an appropriate system behaviour completely at design time. The aim is to endow such technical systems with the so-called self-X properties, such as self-organisation, self-configuration or self-healing. In such dynamic systems, trust is an important prerequisite to enable the usage of Organic Computing systems and algorithms in market-ready products in the future. The OC-Trust project aims at introducing trust mechanisms to improve and assure the interoperability of subsystems. In this paper, we deal with aspects of organic systems regarding trustworthiness on the subsystem level (agents) in a desktop grid system. We develop an agent-based simulation of a desktop grid to show, that the introduction of trust concepts improves the system's performance, in such that they speed up the processes on the agent level. Specifically, we investigate a bottom-up self-organised development of trust structures that create coalition groups of agents that work more efficiently than standard algorithms. Here, an agent can determine individually to what extent it belongs to a Trusted Community.

Due to steadily increasing mobility and the resulting rising traffic demands, serious congestion problems can be observed in many cities. One promising approach to alleviate the congestion effects is the coordination of the network's traffic signals in response to the traffic flow. The recently introduced Decentralised Progressive Signal Systems approach is an adaptive coordination mechanism for traffic signals in urban road networks that relies on local traffic data only. Since the decentralised process cannot lead to optimal results in some special cases, it is extended with an optional hierarchical component introduced in this paper. Based on a broader view on the current network traffic, this Regional Manager is responsible for determining which intersections are coordinated. The efficiency of the coordination determined by the Regional Manager is demonstrated in a simulation-based evaluation that considers the decentralised mechanism and an uncoordinated system for comparison.

Massively multiplayer online games have become increasingly popular. However, their operation is costly, as game servers must be maintained. To reduce these costs, we aim at providing a communication engine to develop massively multiplayer online games based on a peer-to-peer system. In this paper we analyze the requirements of such a system and present an overview of our current work.

The performance of distributed algorithms in mobile ad hoc networks is strongly influenced by the connectivity of the network. In cases where the connectivity is low, network partitioning occurs. The mobility and the density of network nodes as well as the communication technology are fundamental properties that have a large impact on partitioning. A detailed characterization of this behavior helps to improve the performance of distributed algorithms. In this paper we introduce a set of metrics that characterize partitioning in mobile ad hoc networks. Based on an extensive simulation study we show the impact of node mobility, density and transmission range on the proposed metrics for a wide range of network scenarios.

MANETs are used in situations where networks need to be deployed immediately but no network infrastructure is available. If MANET nodes have sensing capabilities, they can capture and communicate the state of their surroundings, including environmental conditions or objects in their proximity. If the sensed state information is propagated to a database to build a consistent model of the real world, a variety of promising context-aware applications becomes possible. We introduce a novel consistency concept that preserves the chronological ordering of sensed state transition events. Based on this concept, we propose a data replication algorithm for MANETs that guarantees the consistency concept without relying on synchronized clocks and show its correctness. Our simulation experiments show that replicated copies are updated regularly even if the network load in the system is high.

No abstract available.

Distribution of data in mobile ad hoc networks is challenged when the mobility of nodes leads to frequent topology changes. Existing approaches so far address either the network partitioning problem or are capable of handling large amounts of data, but not both at the same time. In this paper, a novel approach is presented which is based on a negotiation scheme enhanced by an adaptive repetition strategy. Different strategies for the selection of repeated data are presented and evaluated. Simulation results show a reduction of data transfer volume compared to hyper-flooding by 30% to 40% even in the presence of frequent network partitions.