Industrial
AI
Artificial intelligence for innovative robotics and automation in production
Your Contact Person
Dr.-Ing. Muhammad Saeed
Research Coordinator ARENA2036. e.V.
Industrial AI
The Industrial AI research focus in ARENA2036 investigates the use of artificial intelligence (AI) to optimize robotics and automation systems. Intelligent algorithms and machine learning are used to develop new solutions for adaptive production, safe human-robot interaction and autonomous decision-making. The aim is to make AI technologies usable for industrial applications and to exploit innovation potential.
Thematic focus
Drones for surveillance, object identification, and precision agriculture
Our drone-based intelligent surveillance system combines advanced image processing, AI-powered object recognition, and autonomous navigation capabilities to efficiently monitor large industrial and outdoor areas. Equipped with high-resolution cameras and edge AI processing, the drones can identify, classify, and track objects in real time, enabling applications such as infrastructure inspections, security monitoring, inventory tracking, and environmental assessment. The system uses machine learning algorithms to recognize specific objects, detect anomalies, and deliver actionable insights with minimal human intervention. Beyond industrial monitoring, the technology can be extended to precision agriculture, where drones can assess plant health, monitor irrigation status, detect pest infestations, and map areas requiring pesticide treatment. By integrating geodata analysis and AI-based image processing, the system enables targeted pesticide application, thereby reducing chemical use while simultaneously improving crop yields and sustainability. This intelligent aerial monitoring solution demonstrates how autonomous systems can transform both industrial and agricultural operations through data-driven decision-making.
Contacts
Anum Iqbal, Prof. Aamir Ahmed
Autonomous Guided Vehicles (AGVs) with Reinforcement Learning for Intelligent Logistics Planning
The AGV research platform focuses on intelligent logistics and material handling using deep reinforcement learning (DRL) and reinforcement learning (RL). The autonomous vehicles learn optimal navigation strategies between specified locations while dynamically adapting to changing environments, obstacles, and operational constraints. Unlike conventional path-planning methods, RL-based AGVs continuously improve their performance through interaction with their environment, thereby enabling more efficient and flexible logistics processes. To ensure industrial interoperability, the system integrates the VDA 5050 communication standard and Asset Administration Shell (AAS) technologies. This integration enables seamless communication between heterogeneous AGVs, logistics systems, and Industry 4.0 platforms. The project demonstrates how intelligent transport systems can contribute to smart factories by optimizing material flow, reducing transport times, and enabling autonomous decision-making in networked production environments.
Contacts
Sarah Zain, Johannes Scheu, Florenz Graf
Development of digital twins using FARO point cloud capture and object recognition
The "Digital Twin" initiative focuses on creating a comprehensive virtual representation of the entire production facility using FARO laser scanning technology. High-resolution point cloud data is captured throughout the facility, resulting in precise three-dimensional models of machines, workstations, robots, and infrastructure components. These digital representations provide a detailed and continuously updatable overview of the physical environment. Advanced AI and point cloud analytics are then used to identify, classify, and monitor objects within the scanned environment. By combining digital twins with object recognition technologies, the platform supports factory planning, plant management, process optimization, and real-time monitoring. The resulting digital ecosystem forms the foundation for simulation-based decision-making and enables the transition to fully connected smart manufacturing environments.
Contacts
Maaz Ahmed, Muhammad Tayyab, Ananya Kulkarni, Steffen Kappes, Dennis Hoppe, Kerem Kayabay
Integration of libraries for quantum-based machine learning into the HPC infrastructure
This research project focuses on integrating frameworks for quantum machine learning (QML) into the high-performance computing (HPC) resources provided by HLRS. The goal is to investigate how classical supercomputing and new quantum computing technologies can be combined to solve complex industrial and scientific challenges more efficiently. By employing hybrid quantum-classical workflows, researchers can evaluate the performance, scalability, and practical applicability of quantum-enhanced algorithms.
The project utilizes leading quantum software libraries and simulators in conjunction with HPC clusters to accelerate experiments and benchmarking. Application areas include optimization, predictive maintenance, manufacturing intelligence, logistics planning, and scientific discovery. This initiative serves as a springboard for future quantum-HPC ecosystems capable of solving computational problems that cannot currently be tackled with classical computing alone.
Contacts
Maaz Ahmed, Kanak Pandit, Dennis Hoppe, Kerem Kayabay
KUKA robot arm for path planning based on reinforcement learning
Die KUKA-Roboterarmplattform wird zur Entwicklung und Validierung von Algorithmen des verstärkenden Lernens für die intelligente Bahnplanung genutzt. Anstatt sich ausschließlich auf vordefinierte Bewegungsbahnen zu stützen, lernt der Roboter durch Interaktion mit seiner Umgebung optimale Bewegungsstrategien. Das System verbessert kontinuierlich seine Fähigkeiten zur Bewegungsplanung und ermöglicht so eine effiziente Navigation durch komplexe und dynamische Arbeitsbereiche. Dieser Ansatz bietet erhebliche Vorteile in flexiblen Fertigungsszenarien, in denen sich Layouts, Objekte und Aufgaben häufig ändern. Das RL-basierte Framework verbessert die Anpassungsfähigkeit, reduziert den Programmieraufwand und ermöglicht die autonome Optimierung von Roboterabläufen. Das Projekt zeigt, wie KI-gesteuerte Robotik Fertigungsumgebungen der nächsten Generation unterstützen kann, die sich durch Agilität, Effizienz und Resilienz auszeichnen.
Contacts
Lukas Ott, Sumayyah Khan, Kanak Pandit
KUKA robot arm for context-aware collision avoidance
This project focuses on context-aware collision avoidance using KUKA robotic systems and advanced AI technologies. Conventional collision avoidance systems primarily react to obstacles; however, the proposed concept takes into account contextual information such as object type, task priority, workspace conditions, and the presence of people to make intelligent decisions in real time. Using sensor fusion, machine learning, and environmental perception, the robotic arm can distinguish between different object categories and adapt its behavior accordingly. This capability improves safety, increases productivity, and supports more natural collaboration between humans and robots. The technology represents an important step toward cognitive manufacturing systems that understand their operational context and respond to it intelligently.
Contacts
Lukas Ott, Matthias Marquart
NEURA LARA for intelligent bin-picking applications
The NEURA LARA robot platform is being tested for advanced bin-picking applications in industrial environments. Equipped with sophisticated perception systems and precise manipulation capabilities, the robot can identify, locate, and retrieve objects from unstructured bins with high accuracy and reliability. The system integrates image-processing-based object recognition and motion planning to automate repetitive material-handling tasks. Beyond traditional manufacturing applications, the technology holds great potential in the field of medical technology. The automated handling of medical components, laboratory supplies, and medical devices can increase operational efficiency while reducing the workload on staff. The platform demonstrates how intelligent robotic manipulation can contribute to both industrial automation and future solutions in the healthcare sector.
Contacts
Johannes Scheu, Riffaqat Hussain, Kanak Pandit, Ananya Kulkarni, Ons Laroussi, Heba Khamis
NEURA MAiRA for cognitive sorting, stacking, and segmentation
NEURA MAiRA represents a new generation of cognitive robotics capable of understanding and interacting with complex environments. The system combines AI-based perception, object segmentation, and adaptive manipulation to sort, stack, and organize a wide variety of industrial and household items. Unlike conventional automation systems, MAiRA continuously interprets its environment and adapts its actions to changing conditions. Through advanced machine learning and semantic understanding, the robot can distinguish between object categories, optimize stacking strategies, and perform intelligent sorting tasks with minimal supervision. Applications range from logistics and warehousing to recycling and assembly operations, as well as smart home environments. The project illustrates the evolution of robotics from programmable automation toward truly cognitive systems.
Contacts
Johannes Scheu, Ateeqa Siddiqui, Kanak Pandit, Ananya Kulkarni, Kazim Raza, Chaima Daldoul, Lateef Khan
Humanoid robots and reinforcement learning for human-robot collaboration
This research paper examines the application of reinforcement learning in humanoid robots designed to collaborate with humans in shared environments. Humanoid robots possess human-like movement and manipulation capabilities, making them particularly well-suited for collaborative tasks such as pick-and-place operations, material handling, and workplace assistance. Through the use of reinforcement learning, the humanoid robot learns optimal behaviors through experience and adapts to dynamic conditions in real time. The focus is on enabling safe, efficient, and intuitive collaboration between humans and robots. Such systems have the potential to support the manufacturing, logistics, healthcare, and service industries by combining human flexibility with robotic precision and endurance.
Contacts
Mustafa Talal, Lukas Ott, Kanak Pandit
Retrieval-Augmented Generation (RAG) for Industrial Knowledge Management
Retrieval-Augmented Generation (RAG) combines large language models with external knowledge sources to provide accurate, context-aware, and traceable answers. Unlike traditional AI assistants, which rely solely on pre-trained knowledge, RAG systems draw relevant information from documents, databases, technical manuals, and corporate repositories before generating responses. This technology enables intelligent knowledge management in industrial environments by supporting engineers, operators, and decision-makers with immediate access to critical information. Potential applications include maintenance support, troubleshooting assistance, searching technical documentation, quality management, and training. The approach improves the accuracy of responses while ensuring that the generated content is based on verified corporate knowledge.
Contact persons
Lateef Khan, Ilche Georgievski
Large Language Models (LLMs) for Industrial and Research Applications
Large language models (LLMs) are transforming the way companies handle information, automate processes, and support decision-making. These AI systems can understand, generate, summarize, and analyze natural language, enabling a wide range of applications in manufacturing, research, and business operations. In industrial settings, LLMs can support intelligent assistants, automated reporting, predictive maintenance analytics, technical documentation, code generation, process optimization, and human-machine interaction. When combined with domain-specific data, digital twins, IoT systems, and knowledge bases, LLMs become powerful tools for accelerating innovation and improving operational efficiency. The technology represents a key building block for future AI-driven smart factories and intelligent enterprises.
Contacts
Tanay Rajendrakumar Khilare, Ulrike Pado
Trustworthy and transparent AI for real-time video analysis
This project develops trustworthy and explainable AI (XAI) for real-time video analysis, enabling accurate object detection and identification while providing transparent explanations for AI decisions. To ensure data privacy and compliance with security regulations, the system automatically blurs and anonymizes human faces while continuing to monitor activities and detect relevant objects. The solution supports applications in the fields of smart manufacturing, logistics, surveillance, and security monitoring by combining real-time object detection, data privacy, and explainable decision-making.
Contacts
Riffaqat Hussain, Kanak Pandit, Kerem Kayaba
AI-assisted creation of CAD models
This project explores the use of generative AI and machine learning techniques to automatically create and optimize CAD models based on design requirements, sketches, text specifications, or existing technical data. This approach accelerates product development, supports rapid prototyping, and enables engineers to generate innovative design alternatives while reducing manual modeling effort and development cycle times.
Contacts
Nabiha Anwar
Generating synthetic data for training artificial intelligence
This activity focuses on generating high-quality synthetic datasets using AI, simulation, and digital twin technologies to support the development of machine learning models. Synthetic data helps overcome data scarcity, privacy concerns, and annotation challenges while improving the robustness and accuracy of AI systems. Applications include object recognition, robotics, quality inspection, autonomous systems, and industrial automation.
Contacts
Maaz Ahmed, Anum Iqbal, Ateeqa Siddiqui, Mustafa Talal, Kanak Pandit, Ananya Kulkarni
Quantum Machine Learning (QML)
This project explores the integration of quantum computing and machine learning to develop next-generation AI solutions for complex industrial and scientific challenges. By employing quantum-enhanced algorithms in conjunction with classical computing resources, “Quantum Machine Learning” aims to improve optimization, pattern recognition, prediction, and decision-making processes. The research explores hybrid quantum-classical workflows and evaluates their application potential in the fields of manufacturing, logistics, robotics, digital twins, and data-driven engineering to pave the way for future AI systems based on quantum technologies.
Contact persons
Anum Iqbal, Ateeqa Siddiqui, Maaz Ahmed, Kanak Pandit, Ananya Kulkarni, Georg Siedel
Tactile sensors for intelligent robotic handling
This project uses the Contactile Tactile Gripper to enhance robots’ handling capabilities through advanced touch and force sensors. By providing real-time tactile feedback, the gripper enables robots to detect object properties, adjust gripping force, and handle delicate or irregularly shaped objects with greater precision and reliability. The technology supports applications such as bin picking, assembly, quality inspection, and human-robot collaboration, enabling robotic systems to perform complex handling tasks in dynamic industrial environments more safely and intelligently.
Contacts
Johannes Scheu, Kanak Pandit, Ateeqa Siddiqui, Riffaqat Hussain, Ons Laroussi, Heba Khamis
Objective
Projects in the Industrial AI research focus area drive the development of AI-supported automation solutions for industry. Our aim is to research innovative, efficient and safe AI technologies for production and to develop them in an application-oriented manner. The research field is aimed at companies, research institutions and start-ups that are interested in integrating AI into production processes. Industrial partners and scientific institutions are among the initiators.