This paper presents an approach for autocompletion of architectural building designs in the form of graph-based floor plans during the early design phase. We utilize established case-based reasoning methods, such as subgraph matching and transformational adaptation, further we employ supervised and unsupervised machine learning techniques, such as graph clustering and graph neural networks. Combining those methods into a single approach, the goal is to predict possibly missing spaces in architectural designs of housing buildings, supporting the acceleration of the early design process of architects to make it more sustainable, while enriching it with the recent developments of artificial intelligence. The approach was validated by a performance evaluation and a user study with participation of representatives of the architecture domain.

Recent advances in technology established artificial intelligence (AI) as a crucial domain of computer science for industry, research and everyday life. Even though computer-aided architectural design (CAAD) and digital semantic building models (BIM) are essential aspects of the contemporary architectural design process, the acquisition of proper data proves challenging and AI methods are absent in established design software. An option to acquire rich data are design protocol studies sequenced through meaningful relations. However, this data requires a framework for pre-processing and training artificial neural networks (ANN). In this paper, we present our research on BIM and AI for autocompletion through suggesting further design steps to improve the design process of the early design stages, based on the methods of the 'metis' projects. We propose a recurrent neural network (RNN) model to predict future design phases through sequential learning of cognitive sequences, utilising enriched sketch protocol data.

Sketching is a craft supporting the development of ideas and design intentions, as well as an effective tool for communication during the early architectural design stages by making them tangible.Even though sketch-based interaction is a promising approach for Computer-Aided Architectural Design (CAAD) systems, it remains a challenge for computers to recognise information in a sketch.Design protocol studies conducted to deconstruct the sketch and sketching process collect solely qualitative data so far.However, the 'metis' projects aim to create an intelligent design assistant, using an artificial neural network (ANN), in the manner of Negroponte's Architecture Machine.By assimilating to the user's idiosyncrasies, the system suggests further design steps to the architect to improve the design decision making process for economic growth, qualitative selfeducation through the dialogue and reducing stress.For training such ANN quantitative data is needed.In order to produce quantifiable results from such a study, we propose our open-source web-tool 'Sketch Protocol Analyser'.By correlating different parameters (i.e.video, transcript and sketch built) through the same labels and their timestamps, we create quantitative data for further use.

The main aim of this research is to harness deep learning techniques to support architectural design problems in early design phases, for example, to enable auto-completion of unfinished designs.For this purpose, we investigate the possibilities offered by established deep learning libraries such as TensorFlow.In this paper, we address a core challenge that arises, namely the transformation of semantic building information into a tensor format that can be processed by the libraries.Specifically, we address the representation of information about room types of a building and type of connection between the respective rooms.We develop and discuss five formats.Results of an initial evaluation based on a classification task show that all formats are suitable for training deep learning networks.However, a clear winner could be determined as well, for which a maximum value of 98% for validation accuracy could be achieved.

In order to recommend architects design options, a system was developed which uses artificial intelligence (AI) methods of case-based reasoning (CBR) and deep learning. Since the system uses deep learning, it requires a sufficient amount of data for training, but currently, not enough amount of semantic building data is available publically. In this paper, a Generative Adversarial Network (GAN) is considered to generate the semantic building data to train a Deep Neural Network (DNN) to recommend design options.

This paper presents a system for case-based retrieval of architectural designs in the form of graph-based room configurations by means of applying a case preselection process using a convolutional neural network and the subsequent graph and subgraph matching on the preselected cases. An integral part of the system is its specific user interface that visualizes the architectural concepts of the system in the way familiar for the target user group. The goal of the system is to support higher architectural education with digital assistance methods by providing a tool that can be used to enhance early design phases. The evaluation showed that the system outperforms its predecessor and is suitable for use in education. The approach was developed in context of a bigger framework, however, the research can be considered self-contained and the methods transferred to the domains other than architecture.

Modeling the individual set of skills a person possesses is a difficult task. However, most people seek to find a job that suits them best, despite being unaware of which skills she inherits and what defines herself as a person. Recruiters are knowledgeable in terms of extracting these information by an informal phone call or when carrying out an assessment center. However, we suggest to set a step beforehand and provide a case-based decision support system by using T-Shapes as a model for matching personal skills with job requirements. We provide a first overview on how to structure and model the four knowledge containers of case-based reasoning and how we use these knowledge containers to obtain the most similar T-Shape (situation). This T-Shape can then be reused by the recruiter, providing her with all necessary information at a first glance.

School buildings are currently one of the largest portions of planning and building projects in Germany. In order to reflect the continuous developments in school building construction with constantly changing spatial requirements, an approach to analyse, derive and combine patterns of schools is proposed to adapt school typologies accordingly. Therefore, the topology is analysed, concerning interconnection methods, such as adjacency, accessibility, depth, and flow. The geometric analysis of e.g. room sizes or spatial proportions is enhanced by including grouping of rooms, estimated room clusters, or room shapes. Furthermore, text-matching is used to determine e.g. room program fulfilment, or assigning functional room descriptions to predefined room types, revealing huge differences of terms throughout time and architects. First results of the analyses show a relevant correlation between spatial proportion and room types.

This chapter presents a case study of a prototype Knowledge Management system that supports the process of Manufacturing Problem Solving in a multinational company. The prototype system allows capturing and reusing knowledge generated during the resolution of Overall Equipment Effectiveness (OEE) problems in multiple locations at shop floor level. The developed system was implemented in Exide Technologies. The system integrates the 8D method, Case-Based Reasoning (CBR) and Product Lifecycle Management (PLM). The PLM system is used as the source of extended problem context information (i.e. Products, Processes and Resources) that will enrich the similarity calculation of the CBR application. Process Failure Mode and Effect Analysis (PFMEA) is used as the source of the initial set of cases to populate the case-base. From the development perspective, the system comprises a multi-agent architecture based on SEASALT (Shared Experience using an Agent-based System Architecture LayouT) and programmed in Java. The development infrastructure comprises: Eclipse, JADE (Java Agent DEvelopment framework) and AML (Adaptive Mark-up Language) studio. The selected software applications are myCBR and Aras. The prototype system was tested and validated in three main steps with an increasing level of complexity. The results demonstrated the feasibility of the adopted approach. An overall description of the system, results, lessons learned, and recommendations are provided.

This paper describes the learning of new similarity values for existing measures within the framework FEATURE-TAK. Maintenance of similarity measures is not easy, especially when having a semiautomated approach to relieve the knowledge engineer. Based on the extension of the vocabulary, the newly added values have to be integrated into the similarity measures with an initial similarity value to be useful. We describe the extension of the similarity measures with automated taxonomy extension and one-mode projections and present a comprehensive evaluation and comparison between the different approaches to highlight the advantages and short comings.

The artificial intelligence methods, such as case-based reasoning and artificial neural networks were already applied to the task of architectural design support in a multitude of specific approaches and tools. However, modern AI trends, such as Explainable AI (XAI), and additional features, such as providing contextual suggestions for the next step of the design process, were rarely considered an integral part of these approaches or simply not available. In this paper, we present an application of a distributed AI-based methodology FLEA (Find, Learn, Explain, Adapt) to the task of room configuration during the early conceptual phases of architectural design. The implementation of the methodology in the framework MetisCBR applies CBR-based methods for retrieval of similar floor plans to suggest possibly inspirational designs and to explain the returned results with specific explanation patterns. Furthermore, it makes use of a farm of recurrent neural networks to suggest contextually suitable next configuration steps and to present design variations that show how the designs may evolve in the future. The flexibility of FLEA allows for variational use of its components in order to activate the currently required modules only. The methodology was initialized during the basic research project Metis (funded by German Research Foundation) during which the architectural semantic search patterns and a family of corresponding floor plan representations were developed. FLEA uses these patterns and representations as the base for its semantic search, explanation, next step suggestion, and adaptation components. The methodology implementation was iteratively tested during quantitative evaluations and user studies with multiple floor plan datasets.

As the complexity of building tasks and requirements increases, designers often find themselves confronted with interdisciplinary problems that go beyond the specific challenges and methods of architecture. The iterative nature of the design process results in a continuous exchange between creative, analytical and evaluative activities, through which the designer explores and identifies promising design variants. The ability to compare and evaluate relevant reference examples of already built or designed buildings helps designers to assess their own design and informs the design process.

This paper introduces FLEA-CBR, an alternative approach for composition of case-based reasoning (CBR) processes. FLEA-CBR extends the original 4R (Retrieve, Reuse, Revise, Retain) CBR cycle with a flexible order of execution of its main steps. Additionally, a number of combinatorial features for a more comprehensive and enhanced composition can be used. FLEA is an acronym for Find, Learn, Explain, Adapt and was initially created to solve the restrictiveness issues of case-based design (CBD) where many existing approaches consist of the retrieval phase only. However, the methodology can be transferred to other CBR domains too, as its flexibility allows for convenient adaptation to the given requirements and constraints. The main advantages of FLEA-CBR over the classic 4R cycle are the ability to combine and activate the main steps in desired or arbitrary order and the use of the explainability feature together with each of the steps as well as a standalone component, providing a deep integration of Explainable AI (XAI) into the CBR cycle. Besides the CBR methods, the methodology was also conceptualized to make use of the currently popular machine learning methods, such as recurrent and convolutional neural networks (RNN, ConvNet) or general adversarial nets (GAN), for all of its steps. It is also compatible with different case representations, such as graph- or attribute-based. Being a template for a distributed software architecture, FLEA-CBR relies on the autonomy of implemented components, making the methodology more stable and suitable for use in modern container-based environments. Along with the detailed description of the methodology, this paper also provides two examples of its usage: for the domain of CBRbased creativity and library service optimization.

We present an approach for computer-aided generation of different variations of floor plans during the early phases of conceptual design in architecture. The early design phases are mostly characterized by the processes of inspiration gaining and search for contextual help in order to improve the building design at hand. The generation method described in this work uses the novel as well as established artificial intelligence methods, namely, generative adversarial nets and case-based reasoning, for creation of possible evolutions of the current design based on the most similar previous designs. The main goal of this approach is to provide the designer with information on how the current floor plan can evolve over time in order to influence the direction of the design process. The work described in this paper is part of the methodology FLEA (Find, Learn, Explain, Adapt) whose task is to provide a holistic structure for support of the early conceptual phases in architecture. The approach is implemented as the adaptation component of the framework MetisCBR that is based on FLEA.

: In this paper, we describe the general structure and evaluation of a multi-agent based system module that was conceptualized to explain, and therefore, enrich the search results of the retrieval process within a distributed case-based framework for support of early conceptual design phase in architecture. This explanation module is implemented as an essential part of the framework and uses case-based agents, explanation ontology, and explanation patterns as its underlying foundational components. The module’s main goal is to provide the user with additional information about the search results to make the framework’s behavior during the retrieval stage more transparent and traceable. System’s justification for displaying of results plays an important role as well, and is also included in the explanations. We evaluated the explanation generation process with a ground-truth set of explanations and a case-based validation process to ensure the suitability of the generated explanation expressions for displaying in user interfaces connected to the framework. The results of the evaluation confirmed our expectation and showed the general validity of the explanations.

Building a case base for a case-based reasoning (CBR) system is incomplete without similarity measures. For the attribute-value case structure similarity between values of an attribute should logically fit their relationship. Bipartite graphs have been shown to be a good representation of relationships between values of symbolic attributes and the diagnosis of the cases in a technical diagnosis CBR system, while using an asymmetrical weighted one-mode projection on the values to model their similarity. However, the weighted one-mode projection assumes that the set of symbols is static, which is contradictory to the dynamic nature of case bases as defined by the retain phase of the CBR cycle. In this work we present two methods to update the similarity measure whenever new information is available and compare them. We show that even though updating the similarity measure to exactly reflect the case base had the new information been available a-priori produces better results, an imperfect update is a feasible, less time consuming temporary solution.

This paper addresses the issues of explainability of case-based support systems, particularly structural CBR systems dominated by knowledge-rich comprehensive cases and domain models. We show how explanation patterns and contextually enriched explanations of retrieval results can provide human-understandable insights on the system behavior, justify the shown results, and recommend the best cases to be considered for further use. We applied and implemented our approach as an agent-based system module within a case-based assistance framework for support of the early conceptual phases in architectural design, taking a single floor plan as a case with a high number of attributes. For the retrieval phase, a semantic search pattern structure, Semantic Fingerprint, was applied, whereas the explanation generation phase is controlled by a number of explanation patterns adapted from already existing explanation goals. Rulesets, case bases, and natural language generation are used for construction and automatic revision of explanation expressions. A contextualization feature categorizes the results into different context classes and includes this information into the explanation. A user study we conducted after the implementation of the explanation algorithm resulted in good acceptance by the representatives of the architectural domain, a quantitative experiment revealed a high rate of valid generated explanations and a reasonable distribution of patterns and contexts.