Lothar Hotz

Configuration Knowledge Representation and Reasoning

Configuration models specify the set of possible configurations (solutions). A configuration model together with a defined set of (customer) requirements are the major elements of a configuration task (problem). In this chapter, we discuss different knowledge representations that can be used for the definition of a configuration model. We provide examples that help to further develop the understanding of the underlying concepts and include a UML-based personal computer (PC) configuration model that is used as a reference example throughout this book.

High-Level Expectations for Low-Level Image Processing

Scene interpretation systems are often conceived as extensions of low-level image analysis with bottom-up processing for high-level interpretations. In this contribution we show how a generic high-level interpretation system can generate hypotheses and initiate feedback in terms of top-down controlled low-level image analysis. Experimental results are reported about the recognition of structures in building facades.

A Short History of Configuration Technologies

Abstract Nearly 40-years of configuration technologies motivated us to give this brief overview of the main configuration technological streams. We outline the technological developments in the field starting from the first expert systems of the late 1970s down to today’s configuration solutions. The vastness and intricacies of the configuration field makes it impossible to cover concisely all of its relevant aspects. For the purpose of this overview, we decided to focus on the following four different yet overlapping technological developments: (1) rule-based configurators, (2) early model-based configurators, (3) mainstream configuration environments, and (4) mass customization toolkits.

Using a structure-based configuration tool for product derivation

Because of the possibly large variability in families of software systems and the complex dependencies between individual software components, product derivation in the context of software-intensive systems is not a trivial task. In this demonstration, we show the domain-independent structure-based configuration tool KONWERK, extended with a knowledge base containing a configuration model representing the domain of car periphery supervision systems. Using this tool, reasoning methods known from structure-based configuration are applied in the area of software-intensive systems. Starting with i) a model describing the variability of already realized software components and ii) a concrete task specification for a specific product, during a knowledge-based product derivation process a description of the needed software components is derived. This description is to be used for realizing, i.e. compiling and linking the desired product. The applicability and usefulness of such an approach will be shown in the demonstration.

Beyond physical product configuration --Configuration in unusual domains

Configuration technologies are typically applied in domains with physical products. In this article, we determine characteristics of configuration technologies that are used to compose non-pure physical products. Starting from two case studies software-intensive systems and scene interpretation where we successfully applied configuration, we determine some characteristics of knowledge representation languages and configuration systems that enable to solve configuration tasks in domains beyond pure physical products. As such, the article provides thinking outside the box of physical product configuration.

Smarthome Configuration Model

In this chapter, we present a configuration model taken from a building automation domain. It provides complex aspects of configuration models such as separation of features of a system from realizing system components and domain-dependent workflows for the configuration process.

Evaluation across multiple views for variable automation systems

Automation systems in industry are often software-intensive systems consisting of software and hardware components. During their development several engineers of different disciplines are involved, such as mechanical, electrical and software engineering. Each engineer focuses on specific system aspects to be developed. To enable an efficient development, product lines especially with feature models for variability modeling are promising technologies. In order to reduce the complexity of both feature models and development process, views on feature models can be applied. The use of views for filtering purposes constitutes an established method. However, views also enable further options missing in current approaches, such as evaluations regarding requirements, including non-functional ones. This paper presents an approach for evaluation across multiple views to enable collaborative development for developers who focus on different system aspects. We validate our approach by applying it in an industrial project for the planning of flying saws.

Model-Based Configuration Support For Software Product Families

In this paper, we present main aspects of the ConIPF methodology which can be used to derive customer-specific software products. The methodology is based on software product families and model-based configuration. First results from using the methodology in an industrial context are presented.

MAD: a real world application of qualitative model-based decision tree generation for diagnosis

Computer diagnosis systems grounded on hand-crafted decision trees are wide-spread in industrial practice. Since the complexity of technical system increases and innovation cycles are shortened, the need for systematic decision tree generation and maintenance arises. In this paper, the MAD system is introduced which generates decision trees based on qualitative device models. Existing resources such as design data and expert design know-how as well as decision trees and diagnosis knowledge can easily be reused and integrated into decision tree generation. Since decision tree generation is based on device models, applying MAD reduces average fault identification cost and facilitates quality management of diagnosis equipment. Furthermore, cost of diagnosis system generation, modification and maintenance is reduced. We have successfully evaluated the MAD system in cooperation with the german forklift manufacturer STILL GmbH Hamburg

A Robot Waiter Learning from Experiences

In this contribution, we consider learning tasks of a robot simulating a waiter in a restaurant. The robot records experiences and creates or adapts concepts represented in the web ontology language OWL 2, extended by quantitative spatial and temporal information. As a typical task, the robot is instructed to perform a specific activity in a few concrete scenarios and then expected to autonomously apply the conceptualized experiences to a new scenario. Constructing concepts from examples in a formal knowledge representation framework is well understood in principle, but several aspects important for realistic applications in robotics have remained unattended and are addressed in this paper. First, we consider conceptual representations of activity concepts combined with relevant factual knowledge about the environment. Second, the instructions can be coarse, confined to essential steps of a task, hence the robot has to autonomously determine the relevant context. Third, we propose a ”Good Common Subsumer” as opposed to the formal ”Least Common Subsumer” for the conceptualization of examples in order to obtain cognitively plausible results. Experiments are based on work in Project RACE where a PR2 robot is employed for recording experiences, learning and applying the learnt concepts.