Search

Making Automated Driving (AD) a reality is a challenging task and still subject of intensive research and development. The core component for realizing AD beside the actual vehicle platform and the sensors for perceiving the environment is the AD system, i.e., the software that enables a vehicle to perform distinct dynamic driving tasks in a distinct operational design domain. Briefly speaking, an operational design domain defines environmental and time-of-delay conditions and restrictions, respectively, the presence or absence of certain traffic, and roadway characteristics.

This paper presents a convolutional neural network (CNN) which can be used for forecasting electricity load profiles 36 hours into the future. In contrast to well established CNN architectures, the input data is one-dimensional. A parameter scanning of network parameters is conducted in order to gain information about the influence of the kernel size, number of filters and number of nodes. Furthermore, different dropout methods are applied to the CNN and are evaluated. The results show that a good forecast quality can already be achieved with basic CNN architectures, the dropout improves the forecast. The method works not only for smooth sum loads of many hundred consumers, but also for the load of single apartment buildings.

Terminological resources play a central role in the organization and retrieval of scientific texts. Both simple keyword lists and advanced modelings of relationships between terminological concepts can make a most valuable contribution to the analysis, classification, and finding of appropriate digital documents, either on the web or within local repositories. This seems especially true for long-established scientific fields with elusive theoretical and historical branches, where the use of terminology within documents from different origins is often far from being consistent. In this paper, we report on the progress of a linguistically motivated project on the onomasiological re-modeling of the terminological resources for the grammatical information system grammis. We present the design principles and the results of their application. In particular, we focus on new features for the authoring backend and discuss how these innovations help to evaluate existing, loosely structured terminological content, as well as to efficiently deal with automatic term extraction. Furthermore, we introduce a transformation to a future SKOS representation. We conclude with a positioning of our resources with regard to the Knowledge Organization discourse and discuss how a highly complex information environment like grammis benefits from the re-designed terminological KOS.

Grammis is a web-based information system on German grammar, hosted by the Institute for the German Language (IDS). It is human-oriented and features different theoretical perspectives on grammar. Currently, the terminology component of grammis is being redesigned for this theoretical diversity to play a more prominent role in the data model. This also opens opportunities for implementing some machine-oriented features. In this paper, we present the re-design of both data model and knowledge base. We explore how the addition of machine-oriented features to the data model impacts the knowledge base; in particular, how this addition shifts some of the textual complexity into the data model. We show that our resource can easily be ported to a SKOS-XL representation, which makes it available for data science, knowledge-based NLP applications, and LOD in the context of digital humanities.

The Operations Research EXperiment Framework for Java (OREX-J) is an object-oriented software framework that helps users to design, implement and conduct computational experiments for the analysis of optimization algorithms. As it was designed in a generic way using object-oriented programming and design patterns, it is not limited to a specific class of optimization problems and algorithms. The purpose of the framework is to reduce the amount of manual labor required for conducting and evaluating computational experiments: OREX-J provides a generic, extensible data model for storing detailed data on an experimental design and its results. Those data can include algorithm parameters, test instance generator settings, the instances themselves, run-times, algorithm logs, solution properties, etc. All data are automatically saved in a relational database (MySQL, http://www.mysql.com/ ) by means of the object-relational mapping library Hibernate ( http://www.hibernate.org/ ). This simplifies the task of analyzing computational results, as even complex analyses can be performed using comparatively simple Structured Query Language (SQL) queries. Also, OREX-J simplifies the comparison of algorithms developed by different researchers: Instead of integrating other researchers' algorithms into proprietary test beds, researchers could use OREX-J as a common experiment framework. This paper describes the architecture and features of OREX-J and exemplifies its usage in a case study. OREX-J has already been used for experiments in three different areas: Algorithms and reformulations for mixed-integer programming models for dynamic lot-sizing with substitutions, a simulation-based optimization approach for a stochastic multi-location inventory control model, and an optimization model for software supplier selection and product portfolio planning.

In this paper, we consider a capacitated single-level dynamic lot-sizing problem with sequence-dependent setup costs and times that includes product substitution options. The model is motivated from a real-world production planning problem of a manufacturer of plastic sheets used as an interlayer in car windshields. We develop a mixed-integer programming (MIP) formulation of the problem and devise MIP-based Relax&Fix and Fix&Optimize heuristics. Unlike existing literature, we combine Fix&Optimize with a time decomposition. Also, we develop a specialized substitute decomposition and devise a computation budget allocation scheme for ensuring a uniform, efficient usage of computation time by decompositions and their subproblems. Computational experiments were performed on generated instances whose structure follows that of the considered practical application and which have rather tight production capacities. We found that a Fix&Optimize algorithm with an overlapping time decomposition yielded the best solutions. It outperformed the state-of-the-art approach Relax&Fix and all other tested algorithm variants on the considered class of instances, and returned feasible solutions with neither overtime nor backlogging for all instances. It returned solutions that were on average only 5% worse than those returned by a standard MIP solver after 4 hours and 19% better than those of Relax&Fix.

This paper introduces the theory of bi-decomposition of Boolean functions. This approach optimally exploits functional properties of a Boolean function in order to find an associated multilevel circuit representation having a very short delay by using simple two input gates. The machine learning process is based on the Boolean Differential Calculus and is focused on the aim of detecting the profitable functional properties available for the Boolean function. For clear understanding the bi-decomposition of completely specified Boolean functions is introduced first. Significantly better chance of success are given for bi-decomposition of incompletely specified Boolean functions, discussed secondly. The inclusion of the weak bi-decomposition allows to prove the the completeness of the suggested decomposition method. The basic task for machine learning consists of determining the decomposition type and dedicated sets of variables. Lean on this knowledge a complete recursive design algorithm is suggested. Experimental results over MCNC benchmarks show that the bi-decomposition outperforms SIS and other BDD-based decomposition methods in terms of area and delay of the resulting circuits with comparable CPU time. By switching from the ON-set/OFF-set model of Boolean function lattices to their upper- and lower-bound model a new view to the bi-decomposition arises. This new form of the bi-decomposition theory makes a comprehensible generalization of the bi-decomposition to multivalued function possible.

This article presents a theory for the bi-decomposition of functions in multi-valued logic (MVL). MVL functions are applied in logic design of multi-valued circuits and machine learning applications. Bi-decomposition is a method to decompose a function into two decomposition functions that are connected by a two-input operator called gate. Each of the decomposition functions depends on fewer variables than the original function. Recursive bi-decomposition represents a function as a structure of interconnected gates. For logic synthesis, the type of the gate can be chosen so that it has an efficient hardware representation. For machine learning, gates are selected to represent simple and understandable classification rules. Algorithms are presented for non-disjoint bi-decomposition, where the decomposition functions may share variables with each other. Bi-decomposition is discussed for the min- and max-operators. To describe the MVL bi-decomposition theory, the notion of incompletely specified functions is generalized to function intervals. The application of MVL differential calculus leads to particular efficient algorithms. To ensure complete recursive decomposition, separation is introduced as a new concept to simplify non-decomposable functions. Multi-decomposition is presented as an example of separation. The decomposition algorithms are implemented in a decomposition system called YADE. MVL test functions from logic synthesis and machine learning applications are decomposed. The results are compared to other decomposers. It is verified that YADE finds decompositions of superior quality by bi-decomposition of MVL function sets.

Facial communicative signals (FCSs) such as head gestures, eye gaze, and facial expressions can provide useful feedback in conversations between people and also in human-robot interaction. This paper presents a pattern recognition approach for the interpretation of FCSs in terms of valence, based on the selection of discriminative subsequences in video data. These subsequences capture important temporal dynamics and are used as prototypical reference subsequences in a classification procedure based on dynamic time warping and feature extraction with active appearance models. Using this valence classification, the robot can discriminate positive from negative interaction situations and react accordingly. The approach is evaluated on a database containing videos of people interacting with a robot by teaching the names of several objects to it. The verbal answer of the robot is expected to elicit the display of spontaneous FCSs by the human tutor, which were classified in this work. The achieved classification accuracies are comparable to the average human recognition performance and outperformed our previous results on this task.

Facial expressions are one important nonverbal communication cue, as they can provide feedback in conversations between people and also in human-robot interaction. This paper presents an evaluation of three standard pattern recognition techniques (active appearance models, gabor energy filters, and raw images) for facial feedback interpretation in terms of valence (success and failure) and compares the results to the human performance. The used database contains videos of people interacting with a robot by teaching the names of several objects to it. After teaching, the robot should term the objects correctly. The subjects reacted to its answer while showing spontaneous facial expressions, which were classified in this work. One main result is that an automatic classification of facial expressions in terms of valence using simple standard pattern recognition techniques is possible with an accuracy comparable to the average human classification rate, but with a high variance between different subjects, likewise to the human performance.

In this chapter, we develop two lot-sizing models for multi-level production with flexible BOMs: The Multi-Level Capacitated Lot-Sizing Problem with Substitution (MLCLSP-S), which is an extension of the MLCLSP (see Sect. 2.1.6) and uses hypergraphs to model substitution General Lot-sizing and Scheduling Problem for Multiple production Stages with Flexible Production Sequences (GLSPMS-FPS), which is an extension of the GLSPMS (see Sect. 2.1.7) that uses STN (see Sect. 3.2.4) to model flexible production sequences and flexible BOMs

This chapter deals with the modeling of product substitution and flexible BOMs. In Sect. 3.1, we describe several real-world applications where product substitution occurs. Section 3.2 presents four approaches for modeling substitution: Blending models, substitution graphs, substitution hypergraphs, and task-oriented modeling. Complementary classification criteria for product substitution models are developed in Sect. 3.3. Finally, Sect. 3.4 focusses on conditions where substitution can be beneficial, requirements for organizationally implementing substitutions, and potential pitfalls that should be kept in mind.

In this section, we consider a single-level capacitated lot-sizing problem with substitutions and sequence-dependent setups. The model was designed to map the industrial optimization problem in windshield interlayer production planning described in Sect. 3.1. Rather than building a specific model for this application, we aimed at devising a model and appropriate solution approach for a more general model that can capture the characteristics of this application as well as those of similar production planning problems. Why should it make sense to consider substitutions and sequence-dependent changeovers in one model, rather than treating each of the two aspects in separate subproblems that decompose the overall problem? The idea is that substitutions affect the optimal production sequencing and vice versa, as it might be beneficial to save setup times by refraining from producing certain products and substituting them by others, at least in some settings. This reduces the time spent with “worthless” changeovers, and thereby increases the total capacity available for production.

This section considers extensions of two well-known single-level lot-sizing models, namely the Wagner–Whitin Problem (WWP) and the Capacitated Lot-Sizing Problem (CLSP), that incorporate product substitution options.

This chapter intends to give an overview of the literature on dynamic lot-sizing models and stochastic transshipment models. These two types of models are used as a basis for developing models with substitution in the following chapters. Section 2.1 contains a classification of models for dynamic lot-sizing / production planning, and selected models. In Sect. 2.2, we give a brief overview of available methods for solving deterministic dynamic lot-sizing problems modeled using mixed-integer linear programming (MILP). Section 2.3 introduces transshipment problems and presents a classification scheme for transshipment models. Section 2.4 reviews selected solution approaches that can be applied to stochastic inventory control models such as transshipment problems.

In everyday conversation besides speech people also communicate by means of nonverbal cues. Facial expressions are one important cue, as they can provide useful information about the conversation, for instance, whether the interlocutor seems to understand or appears to be puzzled. Similarly, in human-robot interaction facial expressions also give feedback about the interaction situation. We present a Wizard of Oz user study in an object-teaching scenario where subjects showed several objects to a robot and taught the objects' names. Afterward, the robot should term the objects correctly. In a first evaluation, we let other people watch short video sequences of this study. They decided by looking at the face of the human whether the answer of the robot was correct (unproblematic situation) or incorrect (problematic situation). We conducted the experiments under specific conditions by varying the amount of temporal and visual context information and compare the results with related experiments described in the literature.

In order to provide personalized services and to develop human-like interaction capabilities robots need to recognize their human partner. Face recognition has been studied in the past decade exhaustively in the context of security systems and with significant progress on huge datasets. However, these capabilities are not in focus when it comes to social interaction situations. Humans are able to remember people seen for a short moment in time and apply this knowledge directly in their engagement in conversation. In order to equip a robot with capabilities to recall human interlocutors and to provide user- aware services, we adopt human-human interaction schemes to propose a face memory on the basis of active appearance models integrated with the active memory architecture. This paper presents the concept of the interactive face memory, the applied recognition algorithms, and their embedding into the robot's system architecture. Performance measures are discussed for general face databases as well as scenario-specific datasets.

The human face plays an important role in communication as it allows to discern different interaction partners and provides nonverbal feedback. In this paper, we present a soft real-time vision system that enables an interactive robot to analyze faces of interaction partners not only to identify them, but also to recognize their respective facial expressions as a dialog-controlling non-verbal cue. In order to assure applicability in real world environments, a robust detection scheme is presented which detects faces and basic facial features such as the position of the mouth, nose, and eyes. Based on these detected features, facial parameters are extracted using active appearance models (AAMs) and conveyed to support vector machine (SVM) classifiers to identify both persons and facial expressions. This paper focuses on four different initialization methods for determining the initial shape for the AAM algorithm and their particular performance in two different classification tasks with respect to either the facial expression DaFEx database and to the real world data obtained from a robot's point of view.

This article presents a theory for the bi-decomposition of functions in multivalued logic (MVL). MVL functions are applied in logic design of multi-valued circuits and machine learning applications. Bi-decomposition is a method to decompose a function into two decomposition functions that are connected by a two-input operator called gate. Each of the decomposition functions depends on fewer variables than the original function. Recursive bi-decomposition represents a function as a structure of interconnected gates. For logic synthesis, the type of the gate can be chosen so that it has an efficient hardware representation. For machine learning, gates are selected to represent simple and understandable classification rules.Algorithms are presented for non-disjoint bi-decomposition, where the decomposition functions may share variables with each other. Bi-decomposition is discussed for the min-and max-operators. To describe the MVL bi-decomposition theory, the notion of incompletely specified functions is generalized to function intervals. The application of MVL differential calculus leads to particular efficient algorithms. To ensure complete recursive decomposition, separation is introduced as a new concept to simplify non-decomposable functions. Multi-decomposition is presented as an example of separation.The decomposition algorithms are implemented in a decomposition system called YADE. MVL test functions from logic synthesis and machine learning applications are decomposed. The results are compared to other decomposers. It is verified that YADE finds decompositions of superior quality by bi-decomposition of MVL function sets.

A new method of determining some characteristics of binary images is proposed based on a special linear filtering. This technique enables the estimation of the area fraction, the specific line length, and the specific integral of curvature. Furthermore, the specific length of the total projection is obtained, which gives detailed information about the texture of the image. The influence of lateral and directional resolution depending on the size of the applied filter mask is discussed in detail. The technique includes a method of increasing directional resolution for texture analysis while keeping lateral resolution as high as possible.