Process optimization is crucial in industries with high energy demands like industrial material processing. Neural networks are commonly used for this purpose, but their performance deteriorates over time due to material wear and changes in data distribution. This study addresses these challenges by enhancing modeling, prediction, and optimization through continual learning techniques that detect and adapt to concept drift. This approach maintains model performance despite changing conditions and represents the first contribution to adaptive optimization in industrial material processing. The methodology’s effectiveness is validated with real-world data.

The evolution of Industry 4.0 standards and trends introduces new challenges for traditional Manufacturing Execution Systems, necessitating their transformation into modular, interconnected components with standardized communication interfaces. Although these trends provide a conceptual framework, a distributed architecture for monitoring and aggregating production data remains underdeveloped. This paper addresses this gap by proposing a system of distributed monitoring components that interact hierarchically and heterarchically using Industry 4.0 technologies. The approach integrates Asset Administration Shells for passive data representation with Multi-Agent Systems and investigates its applicability for different control system architectures in different case studies.

Intralogistics 4.0, a key component of Industry 4.0, requires driverless transport systems such as autonomous mobile robots to operate reliably and flexibly in increasingly complex production environments. However, the today’s manual setup of the Robot Operating System 2 and Software stacks is still very time-consuming, error-prone and dependent on skilled workers. We present an agent-based approach that automates both the installation and configuration processes, as well as the system integration into the factory architecture. Our novel Resource Initializer Agent is able to initialize an blank autonomous mobile robot from a minimal operating system to a fully operational robot. By referencing the required software parts within the Asset Administration Shell Submodels, we can automatically install the runtime system (ROS2), configure its software (launch, parameter files, maps, etc.) and also open interfaces to integrate it into the overall multi-agent system without manual operations. Initial results successfully demonstrate that our approach significantly reduces commissioning time and integration risk compared to traditional manual methods. It also lays the foundation for further enhancements such as dynamic reconfiguration of autonomous mobile robot fleets in modular production setups.

Skill-based robotic reconfigurable cyber-physical production modules (RCPPMs) have an aggregated structure, and their functionality depends on the submodules and components that define their current topology. This application-oriented paper proposes a method for integrating the OPC UA Relative Spatial Location (RSL) Companion Specification models and the ROS2 tf2-library functionality to automatically update the SRL models of the RCPPMs when their structure changes. This enables the proper composition and sharing of the necessary information regarding the kinematic structure of each RCPPM’s submodule. On an example, we illustrate how this approach can be employed to enhance the coupling functionality of the skill-based robotic RCPPMs.

Modern Digital Twin standards and technologies, designed to enable smart factories in line with the Industry 4.0 vision, have not yet addressed all the challenges associated with contemporary digital twins. This paper shows how to leverage the Asset Administration Shell as an Industry 4.0 compliant Digital Twin technology to provide all necessary data for Artificial Intelligence Planning, represented via the Planning Domain Definition Language. For this purpose, a model for preconditions and effects of planning actions is defined. This model is applied to three different use cases showing different problems of fetching data from different partially standardized shell templates including three defined actions of transport, assembly, and storage.

While the benefits of reconfigurable manufacturing systems (RMS) are well-known, there are still challenges to their development, including, among others, a modular software architecture that enables rapid reconfiguration without much reprogramming effort. Skill-based engineering improves software modularity and increases the reconfiguration potential of RMS. Nevertheless, a skills' composition framework with a focus on frequent and rapid software changes is still missing. The behavior trees (BTs) framework is a novel approach that enables the intuitive design of modular hierarchical control structures. BTs have been mostly explored from the AI and robotics perspectives, and little work has been done in investigating their potential for composing skills in the manufacturing domain. This paper proposes a framework for skills' composition and execution in skill-based reconfigurable cyber-physical production modules (RCPPMs). It is based on distributed BTs and provides good integration between low-level devices' specific code and AI-based task-oriented frameworks. We have implemented the provided models for the IEC 61499-based distributed automation controllers. This shows the instantiation of the proposed framework with specific industrial technology to enable its evaluation by the automation community.

Automated diagnosis of factories becomes increasingly important to avoid production stops and down times. Additionally, market trends such as small batch size induced by individual customer requirements lead to modular production systems elevating flexibility as well as complexity of smart factories. That leads to more difficulties using conventional diagnosis methods. The ever changing constellations of production modules requires to include context information of the manufacturing domain to enable a suitable fault diagnosis.This paper shows the identification and classification of relevant context information for fault diagnosis in skill-based production systems. The concept of faults is embedded into the Capability-Skill-Service Model used to describe the aforementioned context. Suitable entities and their relations in the production system are outlined and the Knowledge Graph technology is used to make the the context information accessible. An implementation in a real-world demonstrator of SmartFactory-KL is presented and an outlook for further research is given.

We present a comprehensive comparison of hybrid Data-Driven Control (DDC) applied on a hydraulic excavator. DDC offers a state-of-the-art, high performance control based on data and expert knowledge. On the one hand, expert knowledge is complex to adapt to each unique excavator requiring substantial engineering efforts. On the other hand, purely data based control overcomes this drawback by adapting a Neural Network (NN) inverse control directly on measured input/output data. Yet, coverage of the entire phase space and extrapolation to unknown situations is challenging for solely data-driven approaches. On a real demonstrator, we analyze expert white box methods, solely data-driven black box approaches and a hybrid grey box approach which combines a data-driven and simplified expert model. We examine trajectory tracking performance, engineering effort and safe exploration as goal criteria. Besides various experiments for testing safety, we apply a Support-Vector-Machine (SVM) to analyze the extrapolation fitness of the data-driven components to unknown data.

Experimental results for a handheld six degrees of freedom parallel robot with realtime position control and disturbance attenuation are presented. The robot's base is freely movable in space by a human operator, while the tool is position controlled and stabilized relatively to a reference coordinate system. The challenge is to decouple the tool as far as possible from the user involved arm tremor and unintended motions. The robot is equipped with a camera-based position and orientation tracking system, linear actuators, and a realtime control system. Kinematics, co-ordinate transforms and workspace limits are presented with its specifications. Experiments with sinusoidal stimuli on a test stand and during the handheld operation are discussed with respect of stabilizing performance and work-space violation. The resulting error bounds of less than 0.5 mm in the Cartesian position demonstrate that such a robot device has the potential to improve the classical manual surgical interventions. Further, the paper demonstrates the compliance of the user motion to the workspace provided by the robot.

This paper represents the Lyapunov stability study for the wheeled mobile robot C3P based on a non-linear and complex mathematical model. The candidate Lyapunov function used in this work is extracted from the robot kinetic energy equation, which includes the robot dynamics and kinematics models. Such function contains uncontrolable and unpredicticable variables that yeild to very complicated and long analitical analysis. Therefore the analysis is done -part- analiticaly and numericaly in this work, which shows some stability uncertainty within the stable regions. The uncertain stability regions are considered as unstable regions generated by specified operating points. The effect of the control parameters values on overcoming such non-stable regions is presented for different operating points. Forced disturbances are applied on the robot in two different practical experiments. The practical results illustrate the stability study.

Hierarchical localization frameworks provide efficient means for localizing a robot topologically and geometrically. The metric localization performance is enhanced since the searchable space is confined to a previously identified topological place. This adapts well to large-scale environments. In this paper, a two-level hierarchical localization using a compact hybrid map is presented. The map preserves information set at two different abstractions and resolutions, and possesses both geometric and non-geometric properties. The coarse-resolution information enables global topological localization through place matching. The higher-resolution enables local metric localization through triangulation. The presented hierarchical localization is totally independent on the robot's motion model. For effectiveness in both mapping and localization, the map is constructed based on information-theoretic evaluation that selects only highly qualitative information. The approach is demonstrated using a vision sensor and the scale invariant feature transform.

Topological localization is a qualitative solution approach that can assist obtaining a faster quantitative metric solution by limiting the searchable space. Consequently, its efficiency is an essential requirement in hierarchical localization frameworks. This paper presents a topological map generation method with a localization scheme. Good compromise of performance measures - accuracy, memory and processing time - indicates the method's efficiency. The suggested implementations rely on information-theoretic selection of local features for node distinctive representation, and a visual codebook for compression. Testing the proposed approach on the COLD database, a recent specific benchmarking database for robotic topological mapping and localization, reveals its customization according to the vision sensor and environment characteristics. The approach guarantees over 90% localization accuracy with more than 50% overhead reduction, and is suitable for application in highly unstructured cluttered environments that are influenced by dynamics and illumination variations.

In this paper, the collision avoidance behaviour in a recursive nested behaviour control structure for Unmanned Aerial Vehicles will be discussed. The architecture is an extension of the behaviour-based recursive control structure, which has been applied successfully to mobile robot applications. The system structure, as an abstraction of multiple cascaded control loops with feedback mechanisms, is robust against disturbances and model uncertainties. The obstacle awareness and collision avoidance are considered to be the most important issues in the field of Unmanned Aerial Vehicles (UAV), thus more attention will be paid to this topic. The collision avoidance utilizes a potential field method tailored for UAV application domain. The formulation of a repulsion forces based on radar sensor measurement is presented. The implementation of the collision avoidance system is also described.

Hierarchical localization provides both topological and quantitative metric solutions, with faster performance for the latter since the searchable space is minimized. The initial topological localization step is crucial in those frameworks and should be highly accurate. In this paper, a hierarchical localization approach that primarily focuses on the efficiency of the topological module is presented. The approach relies on a minimal set of qualitative entropy-based local features, which achieves both speed and localization accuracy. The abundant features are triangulated in a next step using a photogrammetric projective model to obtain a metric solution. The metric localization selects only the correct matches by regarding a simple yet efficient distance measure to overcome problems of data association and environment dynamics.

A generic system architecture for dependable interactive systems is proposed. The architecture is an extension of the behavior-based recursive control structure, which has already been applied successfully to robotic applications. The interaction between multiple robots or between a human user and robotic systems takes place on multiple behavioral levels providing fixed interfaces. A complex system is decomposed into levels according to the dynamics of behaviors and bandwidth of the interfaces. Signals from and to multimodal interface devices are processed according to the behavior-related information contents and fused within the behavior levels. The system structure, as an abstraction of multiple cascaded control loops with feedback mechanisms, is robust against disturbances and model uncertainties. The usage of the generic system architecture is demonstrated on the example of a semi-autonomous flying robot.

The presented work is related to previous development of the holonomic wheeled mobile robot C3P. This paper focuses on the platform implementation and the kinematics/ dynamics solutions used for its position control structure. The platform prototype is proposed in detailed description concerning its construction and configuration. A controller based on feed forwarding the inverse dynamics torques with the inverse kinematics to overcome the platform singularities is proposed. The based controllers practical experiments results illustrate the position controller performance and its efficiency.

In this paper, an appearance-based modeling of the environment is presented for the sake of mobile robot localization. The model allows perception and recognition within a topological context. Highly descriptive SIFT is used to extract local features from visual data acquired from an indoor environment. A method is developed to select those features, which are best for localization using a probabilistic modeling and an entropy measure. The impact of feature selection on the localization performance is more than 60% reduction in the storage and recognition time overhead. The methodology insures the recognition of different places with 96% precision, in spite of perceptual aliasing and image variability.

A dynamic model for the holonomic mobile robot C3P is proposed. The C3P has three caster wheels with angular velocities actuation. The model consists of two main dynamic equations which have been derived symbolically using a Lagrange approach. The first equation is the forward dynamics which is used to calculate the wheels angular velocities corresponding to the applied torques on the wheels. The second equation is the steering dynamic estimator for recursive calculation of the steering angles and their derivatives with respect to the wheel angular velocities and acceleration. The velocity control of the C3P using the dynamic model is simulated and compared to the kinematics based controller which have been proposed earlier. The simulation and experimental results clearly show the advantages of the dynamic model in relation to the kinematic one