Current projects

ConfidentAM

BMWK, 2023 – 2026

Continuous Fiber-Reinforced Low-Density Thermoplastic Additive Manufacturing

Fiber-reinforced plastics offer significant lightweight potential due to their high stiffness and strength. However, their cost-effective production relies on full automation, which is currently limited to large-scale applications due to high investment requirements. ConfidentAM aims to develop an additive manufacturing chain that combines 3D printing and 3D skeleton winding technology to enable the economic production of fiber-reinforced lightweight structures, even in small quantities and custom-made products.

VADER

BMWK, 2023 – 2025

Networked Ditigal Assistant for the Data-Driven Engineering of Robot Workcells

The use of robots has significant potential to be a universal and flexible core technology for collaborative or fully automated production. However, a lack of technologies to realize economically feasible advanced sensor-based applications prevents robots from realizing their full potential. In VADER, we develop a digital engineering assistant for data-driven automation of industrial production processes with robots, that fits into existing Industry 4.0 infrastructures. The assistant offers selected assistance functions for crucial steps in the entire planning to commissioning process as a service. Research focuses on deployment strategies in the cloud or edge, allowing the user retain full sovereignty over their data.

 

Next2OEM

BMWK, 2023 – 2025

A Digital and Automated Value Chain for Wiring Harness Manufacturing

Next2OEM aims to develop a digitized and automated value chain ranging from the development to the next-to-OEM manufacturing of wire harnesses and assembly in the car body at the OEM. We focus on software solutions for automating the high-precision handling of deformable cables, and the development of advanced vision and force control algorithms.

 

Source: © ArtiMinds Robotics GmbH

EASY

BMWK, 2023 – 2025

Efficient analysis and control in a dynamic edge-cloud-continuum for industrial production

Edge computing enables the neartime processing of data directly at the source. This reduces latencies and allows powerful machine learning and data analysis to be applied in production, when short cycle times are crucial. EASY aims to create an easily usable edge-cloud continuum to provide a runtime environment and services for AI-based analysis and process control. In EASY, ArtiMinds develops technologies for efficiently analyzing large amounts of robot and sensor data as well as deploying machine learning solutions to optimize robotic production processes in neartime.

 

Robot measuring the position and orientation of a cable tip for insertion

GANResilRob

BMWI, 2022 – 2025

Generative Adversarial Networks and semantics for resilient, flexible robotic production

In GANResilRob, we explore the use of generative models for automatic robot program synthesis. The project aims to take advantage of the power of GANs to learn process knowledge in the domains of assembly and disassembly. GAN-based assembly plans are then mapped to semantic robot tasks and integrated into an intuitive robot programming SW suite. Alongside the research and engineering components of the project, we aim to design a „Robot-as-a-Service“ (RaaS) concept, which will provide tools for the manufacturing industry to trailor pre-trained generative models to their concrete use cases.

 

RoboGrind

Invest BW, 2021 – 2023

Hybrid AI for flexible robotic surface treatment

The sustainable refabrication of hardware components usually requires a large amount of manual labor. The wear and tear of used components impacts obejct geometries and surface characteristics in unique ways which are difficult to predict. This renders automation difficult and makes refabrication – despite being ecologically superior – less economically viable. RoboGrind unites ecological and economic sustainability via the use of artificial intelligence. Machine learning techniques permit an automated, fully individualized treatment of single workpieces, which increases the economic viability of refabrication.
 
 

An industrial robot grinding a metal surface

KARL

BMBF, 2021 – 2025

Competency Center Artificial Intelligence for Working and Living in the Karlsruhe Region

The competency center KARL aims at developing human-centered and transparent AI-assisted systems for working and learning, and to make them tangible in concrete, practical applications. The development of the AI-assisted systems is accompanied by the drafting of rules, guidelines and demoable concepts of such working and learning systems. Emphasis is placed on harnessing and evaluating the advantages of both human and machine intelligence in practice. KARL is focused on four application domains, which highlight the technological and scientific strengths of the Karlsruhe region.

Koala-Grasp

BMBF, 2021 – 2024

Cognitively assisted laparoscopy: A learning robotic assistance system for surgical grasping and holding tasks

Koala-Grasp aims to develop new approaches for planning and executing robot motions to assist human surgeons performing laparoscopic surgery. The main technical contribution will be a novel robot-guided mechatronic laparoscopy tool and the corresponding perception, motion planning and control algorithms, which permit the force-controlled manipulation of organs and tissues by a 6- or 7-DOF robot arm through the narrow trocars typically used in minimally invasive surgery. ArtiMinds is developing novel machine learning-based offline motion planning and online robot control algorithms which explicitly consider the intrinsic and extrinsic kinematic and dynamic constraints imposed by the tool and trocar, and which are robust against the perception and manipulation challenges entailed by in-vivo surgery.

 

Completed projects

HoLLiECares

BMBF, 2020 – 2023

A multifunctional service robot to support care professionals in the hospital

Recent years have highlighted several challenges in the care sector such as shortages of trained care professionals, precarious labor conditions or lack of investment. HoLLiECares aims to refine a prototype service robot for flexible use and adaptive support of care professionals in hospital settings:

  • Identification and realization of tasks which can be delegated to a robot to relieve care professionals
  • Development and evaluation of flexible ready-to use hardware and software modules suitable for the healthcare market
  • Development of cost-effective solutions for force-sensitive robot control, navigation, object manipulation and multimodel user interaction
  • Improvement of robot flexibility by combining motor functions with social interaction via touch, language, gesture and brain interfaces

Visit the project website for more information.

KIRK

BMBF, 2020 – 2022

AI-based robot calibration

KIRK aims to develop robust and efficient algorithms for error compensation and calibration of industrial robots using machine learning methods. The core idea is centered around pretraining deep artificial neural networks with simulated robot data and real-world data collected in the lab, to learn powerful calibration models. In a second step, these learned models are finetuned with data from the particular robot to be calibrated. This approach promises the efficient calibration of robots directly in the plant, without significant interruptions to production. We test our approach on calibration scenarios involving payload- and temperature-induced positioning errors as well as wear and tear.

 

ILIAS

BMBF, 2019 – 2022

Imitation learning from human demonstrations in Virtual Reality for physical human-robot interaction in assistance tasks

In ILIAS, we propose a novel way of programming robotic assistance tasks, which scales better towards open task domains. In this programming approach humans demonstrate how to accomplish assistance tasks in virtual environments where the demonstrations are automatically interpreted and transformed into generalized knowledge bases. The system then uses the symbolic representation together with high-dimensional log data from virtual reality (VR)  demonstrations in order to generate and parameterize complex robot programs using deep learning.

 

ProBot

BMBF, 2019 – 2022

Proactive diagnosis and conception of collaborative robot deployment in small and medium-sized enterprises

Collaborative robots (cobots) have been identified as a highly promising tool to address both the increasing innovation pressure and skilled labor shortage small and medium-sized enterprises (SMEs) are facing today. The goal of ProBot is to develop a virtual, integrated toolbox to aid SMEs in the conception and deployment of cobot solutions. The toolbox comprises an online platform which provides interactive tools and guides to assist SMEs in the planning and layouting of collaborative robot cells and the selection of cobot hardware suitable for the intended use cases. ArtiMinds focuses on the development of cobot skills tailored to collaborative surface treatment (grinding, powder coating) and machine tending.

Visit the project website for more information.

 

ErgoBot

ZIM, 2018 – 2020

Dynamic and ergonomic adaptation of robot motion for human-robot collaboration via machine learning and motion analysis

Collaboration between industrial robots and human workers is a key component for efficient high-mix, low-volume production. In ErgoBot, we leverage the flexibility of industrial robots to improve the ergonomics of repetitive tasks while simultaneously optimizing economic productivity

The project results comprise

  • A novel approach for generating precomputed reachability maps, which allow for very fast real-time reachability queries. Given a Human Robot  Collaboration Model (HRCM) containing a probabilistic representation of human activity, this allows for extremely fast pre-planning and online replanning of robot trajectories
  • A reactive programming concept which permits the online adaptation of robot programs via REST interfaces
  • An optimization algorithm for the multicriterial optimization of robot program parameters with respect to worker ergonomics and economic KPIs

Visit the project website for a list of publications and further information.

 

RoPHa

BMBF, 2017 – 2020

Robust perception skills for robotic household assistance in the context of elderly care

In RoPHa, we developed a concept and implementation of a library of modular, sensor-adaptive manipulation skills for food handling and preparation, such as cutting waffles or spreading butter. On this basis, we could automatically instantiate complex robot programs based on the task context and current sensor inputs. We integrated our solution into a complete perception-action system together with vision sensors from roboception, the RoboSherlock perception framework of the University of Bremen and the Care-O-bot 4 assistance robot built by Fraunhofer IPA.

For more information, visit www.ropha-projekt.de.

 

VRCobot

BMBF,  2017 – 2018

Planning Human-Robot Collaboration in Virtual Reality

In VRCobot, we created and evaluated a Virtual Reality (VR) system for the intuitive programming of industrial robots.
By using VR as a means of visualization of the expected robot movements, the programming task could be greatly simplified. The programming itself is also more accessible compared to classic, desktop-based approaches.

 

MonRob

ZIM, 2017 – 2018

Frameworks and technologies for the monitoring, analysis and online adaption of industrial robotic production processes

Over the course of MonRob, three core technologies for the intuitive analysis and data-driven optimization of robot programs were developed.

  • Database and web-based user interface to collect and persist semantically annotated pose, joint, force-torque and meta-information from robot programs natively executed on the robot controller
  • 2D and 3D visualization solutions for the monitoring and analysis (clustering, statistical methods) of robot data
  • Deep learning-based robot program self-optimization: Prediction of object features with deep recurrent neural networks (RNNs), and online optimization of the robot program based on the predictions
  • Deep learning-based multimodal anomaly detection: Detection of anomalies in assembly processes with deep autoencoders on vision and force-torque data

MonRob laid the foundations of ArtiMinds Learning and Analytics for Robots (LAR), our commercial solution for powerful monitoring, analysis and optimization of robotic production processes.

 

DexBuddy

ECHORD++, 2015 – 2016

Intuitive programming of a flexible, dexterous robotic co-worker for varying cable manipulation tasks in industrial assembly and quality control

DexBuddy demonstrated the overall cost reduction potential of using a flexible gripping device in combination with advanced intuitive programming in automotive manufacturing tasks.

We tested the feasibility of easily programmable dexterous robotic co-workers performing highly dexterous tasks in real industrial scenarios – specifically, cable manipulation tasks in manufacturing, bringing together key skills from 4 partners to produce a novel system.

 

RoBioTool

smart businessIT, 2014 – 2015

An adaptive robotics tool for biomedical research

Goal of RoBioTool was the piloting of an intuitive software solution for the fast and intuitive programming of complex, adaptive robot motions for lab tasks in the context of biomedical research.

We developed user interaction concepts and a software solution to allow lab assistants without robotics experience to create complex robotic automation workflows and conducted studies with lab employees of different qualification levels.