N. Hendrich

Hamburg University | UHH · Department of Informatics

Existing robotic musicians utilize detailed hand-crafted instrument models to generate or learn policies for playing because model-free or inaccurate policy rollouts might easily damage or wear out fragile instruments. We introduce an approach to characterize geometric models of chordophones and their audio onset responses directly through audio-ta...

We describe a fully integrated blackjack dealing robot system utilizing multimodal input to interact with human players. It can deal cards to players and visually detect which cards have been played. Furthermore, it detects gestures commonly used in blackjack, such as knocking and swiping performed by human players to indicate whether they would li...

Accurate estimation of the relative pose between an object and a robot hand is critical for many manipulation tasks. However, most of the existing object-in-hand pose datasets use two-finger grippers and also assume that the object remains fixed in the hand without any relative movements, which is not representative of real-world scenarios. To addr...

Human-Robot collaboration (HRC) is an important topic for manufacturing and household robotics. It is very challenging to ensure both efficiency and safety in HRC. This paper presents an HRC pipeline that generates efficient and collision-free robot trajectories based on predictions of the human arm and hand (AH) motions. We train a recurrent neura...

While for vision and audio the same mass-produced units can be embedded in many different systems from smartphones to robots, tactile sensors have to be built in application-specific shapes and sizes. To use a commercially available tactile sensor, it can be necessary to develop the entire system around an existing sensor model. We present a set of...

Markerless vision-based teleoperation that leverages innovations in computer vision offers the advantages of allowing natural and noninvasive finger motions for multifingered robot hands. However, current pose estimation methods still face inaccuracy issues due to the self-occlusion of the fingers. Herein, we develop a novel vision-based hand-arm t...

We propose a vision-proprioception model for planar object pushing, efficiently integrating all necessary information from the environment. A Variational Autoencoder (VAE) is used to extract compact representations from the task-relevant part of the image. With the real-time robot state obtained easily from the hardware system, we fuse the latent r...

Tactile perception is an important modality for dexterous manipulation, but faces unique challenges in that the sensor has to wrap around the system. Biological skin solves these problems extremely well. Not only can it detect contacts, but it also grows on complex three-dimensional shapes, stretches with underlying soft tissue and around joints, a...

In this work, we tackle the challenging problem of grasping novel objects using a high-DoF anthropomorphic hand-arm system. Combining fingertip tactile sensing, joint torques and proprioception, a multimodal agent is trained in simulation to learn the finger motions and to determine when to lift an object. Binary contact information and level-based...

Reliability and robustness to external influences are important characteristics to using humanoid robots outside of laboratory conditions This paper proposes a closed-loop system to recover from falling over and self-righten using quintic spline interpolation. By using PID controllers to correct divergence in the IMU readings, faster and more relia...

The key role of tactile sensing for human grasping and manipulation is widely acknowledged, but most industrial robot grippers and even multi-fingered hands are still designed and used without any tactile sensors. While the basic design principles for resistive or capacitive sensors are well known, several factors keep tactile sensing from large-sc...

Screws are often used to connect 3D printed parts to other objects. When screwing directly into printed plastic, the reliability is limited, and the connection can wear out over time. For more reliable connections, standard metal nuts are often inserted into slots designed into the object. This paper presents an approach where nuts and other ferrom...

Robotic manipulation systems frequently utilize RGB-D cameras based on infrared projection to perceive three-dimensional environments. Unfortunately, this technique often fails on transparent objects such as glasses, bottles and plastic containers. We present methods to exploit the perceived infrared camera images to detect and reconstruct volumetr...

The quality of crossmodal perception hinges on two factors: The accuracy of the independent unimodal perception and the ability to integrate information from different sensory systems. In humans, the ability for cognitively demanding crossmodal perception diminishes from young to old age. Here, we propose a new approach to research to which degree...

Established indoor robot navigation frameworks build on the separation between global and local planners. Whereas global planners rely on traditional graph search algorithms, local planners are expected to handle driving dynamics and resolve minor conflicts. We present a system to train neural-network policies for such a local planner component, ex...

Hybrid manufacturing of 3D-printed parts with integrated electronics has made significant progress in recent years. The general idea of depositing conductive material and electronic components during or after the additive manufacturing step was successfully demonstrated for different process types. However, efficient 3D arrangement and wire routing...

Planar pushing remains a challenging research topic, where building the dynamic model of the interaction is the core issue. Even an accurate analytical dynamic model is inherently unstable because physics parameters such as inertia and friction can only be approximated. Data-driven models usually rely on large amounts of training data, but data col...

Force sensing is essential for many manipulation tasks and, more generally, for all robots physically interacting with their environment. While multi-axis force/torque sensors are readily available commercially, their cost and complex integration have so far limited a wide deployment. In this paper, we introduce a modular approach to design and to...

In this paper, we present a multimodal mobile teleoperation system that consists of a novel vision-based hand pose regression network (Transteleop) and an IMU-based arm tracking method. Transteleop observes the human hand through a low-cost depth camera and generates not only joint angles but also depth images of paired robot hand poses through an...

Robust and accurate estimation of liquid height lies as an essential part of pouring tasks for service robots. However, vision-based methods often fail in occluded conditions while audio-based methods cannot work well in a noisy environment. We instead propose a multimodal pouring network (MP-Net) that is able to robustly predict liquid height by c...

The quality of crossmodal perception hinges on two factors: The accuracy of the independent unimodal perception and the ability to integrate information from different sensory systems. In humans, the ability for cognitively demanding crossmodal perception diminishes from young to old age. To research to which degree impediments of these two abiliti...

Additive manufacturing processes are inherently subject to discretization effects. For most technologies, stair-stepping artifacts impair the surface quality of 3D printed objects, especially when the surface slope is close to horizontal. In this paper we propose a novel Fused Deposition Modeling (FDM) slicing approach that combines nonplanar and p...

The need for labeled training data for object recognition in RoboCup increased due to the spread of deep learning approaches. Creating large sets of training images from different environments and annotating the recorded objects is difficult for a single RoboCup team.

In this paper, we focus on the challenging perception problem in robotic pouring. Most of the existing approaches either leverage visual or haptic information. However, these techniques may suffer from poor generalization performances on opaque containers or concerning measuring precision. To tackle these drawbacks, we propose to make use of audio...

We present a data-driven approach to simulate the BioTac tactile fingertip sensor within physics engines. The behavior of the sensor is first captured in an experimental setup that records positions and external forces of contacts as well as the sensor output. This data is then used to fit a non-linear model that maps force-annotated mesh collision...

In this research, a novel and fast memetic evolutionary algorithm is presented which can solve fully-constrained generic inverse kinematics with multiple end effectors and goal objectives, leaving high flexibility for the design of custom cost functions. The algorithm utilises a hybridisation of evolutionary and swarm optimisation, combined with th...

Various approaches to solving inverse kinematics implicitly rely on computing forward kinematics in order to obtain an approximate solution. This work proposes an optimised data structure to efficiently compute these equations by avoiding redundant transformations and calculations. This is particulary relevant for highly articulated kinematic model...

The need for labeled training data for object recognition in RoboCup increased due to the spread of deep learning approaches. Creating large sets of training images from different environments and annotating the recorded objects is difficult for a single RoboCup team. This paper presents our tool ImageTagger which facilitates creating and sharing s...

The increase in software complexity and hardwarecosts forces robot soccer teams to collaborate. Instead of focusingon a specific architecture, we propose a set of common ROSmessages as interface definitions to encourage software exchange.Furthermore, we implemented fundamental utility packages forcollaborative play in mixed teams.

While solving inverse kinematics on serial kine-matic chains is well researched, many methods still seem rather limited in jointly handling more complex geometries, including dexterous multi-finger hands or humanoid robots. In particular, object manipulation and motion tasks would benefit from the ability to define intermediate goals along the kine...

Adaptively computing the layer heights for 3D-printed parts has the potential to achieve high quality results while maintaining a reasonably short printing time. The basic concept, several error measures and variations of the algorithm have been around in the literature for two decades now, but never showed significant impact on widely used slicing...

Solving kinematic motion is a challenging field of research which is relevant for various applications in character animation and robotics. This paper presents a novel and fast hybrid evolutionary algorithm for inverse kinematics which can handle fully constrained and highly articulated geometries with multiple end effectors and individual objectiv...

Various approaches to solving inverse kinematics implicitly rely on computing forward kinematics in order to obtain an approximate solution. This work proposes an optimised data structure to efficiently compute these equations by avoiding redundant transformations and calculations. This is particulary relevant for highly articulated kinematic model...

Sharing software modules between teams in the RoboCup Humanoid League is difficult since all teams use different frameworks. This leads to reimplementation of software which slows the research process. A common framework for the league would resolve this. Therefore, this thesis proposes a ROS-based architecture which is defined by a set of ROS mess...

This paper presents a novel biologically-inspired approach to solving the inverse kinematics problem efficiently on arbitrary joint chains. It provides high accuracy, convincing success rates and is capable of finding suitable solutions for full pose objectives in real-time while incorporating joint constraints. The algorithm tackles the problem by...

Several approaches to the integration of wires and electronic components into almost every existing additive fabrication process have been successfully demonstrated by a number of research groups in the last years. While the pure mechanical process of generating conductive wires inside of a printed object has proved to be feasible, the design, inte...

In this paper, we present an efficient and adaptive method for detecting and tracking multiple persons while providing real-time capability and high robustness to outlier noise. Given an RGB-D image data sequence, our algorithm combines two independent approaches for person detection. First, a cluster-based segmentation and classification on RGB-D...

This paper presents a method for gait optimization of modular robotic locomotion. A physics-based simulation system for development and optimization of locomotion patterns is utilized. A graphical user interface (GUI) allows for interaction with the simulation at runtime and to supervise it. Long-term optimizations can be performed as background pr...

Designing and building a mobile service robot is a complex task, involving many design decisions regarding hardware components and software architecture. A robot design always constitutes a delicate compromise between capabilities, performance, safety, reliability and costs. Only the availability of high-quality robot middleware frameworks like ROS...

Systems for ambient assisted living (AAL) that integrate service robots with sensor networks and user monitoring can help elderly people with their daily activities, allowing them to stay in their homes and live active lives for as long as possible. In this paper, we outline the AAL system currently developed in the European project Robot-Era, and...

Purpose – This paper aims to present the design of a modular robot with 3D-printing technology. Design/methodology/approach – The robot consists of a number of autonomous modules coupled by magnetic interfaces. Each module combines 3D-printed mechanical parts with widely available standard electronic components, including a microcontroller and a s...

Purpose – This paper aims to present the design of a modular robot with 3D-printing technology. Design/methodology/approach – The robot consists of a number of autonomous modules coupled by magnetic interfaces. Each module combines 3D-printed mechanical parts with widely available standard electronic components, including a microcontroller and a s...

The process of different human manipulating a specific object in hand obeys very similar operating steps. The hand movement can be modeled and generalized into action gist to guide other human or robots to execute the specific in-hand manipulation task. This paper suggests a kind of action gist similar to the way humans learn to represent the five...

This paper describes a service- and scenario-driven software architecture for the ambient assisted living infrastructure currently under development in the Robot-Era project. Involving the end-users from the start, the project integrates an ambient sensor network with an advanced knowledge-representation and planning system and three different robo...

In this paper we present and discuss the design, implementation and evaluation of a visual programming environment. The proposed application allows authoring of therapeutic training sequences for the CareToy system, a hardware system developed for the purpose of inhome rehabilitation training for infants, aiming to improve the motor skills of prete...

Robust grasping and manipulation using multi-finger hands remains one of the key challenges of service robotics. So far, most theoretical approaches and simulators have concentrated on the search for static stable grasps, but without consideration of the task context or the different grasp types to be used during object manipulation. In this paper...

In this paper we present a method for the generation of hand postural synergies for different precision grasp types to be used in dextrous robot hands. Our method records the robot hand motions while teleoperated by human subjects via a dataglove, doing different grasp types on a series of objects. This exploits the fact that humans automatically c...

The main objective of the presented work is the development of an efficient method for the calculation of the inverse kinematic of anthropomorphic robotic hands. Our algorithms is able to treat all fingers as one set when generating a given grasp configuration or in-hand manipulation trajectories, resulting in a simultaneous solution for the wrist...

Tactile sensing plays an important role in robot hand manipulation. This papers presents an analysis of the raw values produced by a biomimetic tactile sensor and the proposed solution organized in a pipeline processing structure to overcome the signal's noncompliances and issues for further processing. Even if this paper focuses on the data genera...

We consider in-hand manipulation tasks that consists of periodic movements. In order to improve the manipulation learning ability of a robot with a human-like hand, this paper introduces a segmentation method based on the techniques of action gist. Action gist is the key motion information in manipulation with the property of semantics. In the tech...

This paper describes a method to integrate a hardware device into a modular robot control and simulation software and introduces sensor fusion to investigate the current set of control parameters. As a highlevel remote unit for modulation of control algorithms the Wiimote and the usage of its sensors will be introduced. Sensor fusion of the Wiimote...

Dextrous object manipulation with multi-fingered robot hands remains one of the key challenges of service robotics. So far, most theoretical approaches and simulators have concentrated on the search for and evaluation of static stable grasps, but with neither a model of the full hand-arm system nor the system dynamics. GraspIt! is probably the best...

In this paper we present an overview of a multi-sensor setup designed to record and analyse human in-hand manipulation - tasks consisting of several phases of finger motions following the initial grasp. During the experiments all of the hand, finger, and object positions are recorded, as are the contact forces applied to the manipulated objects. Th...

This paper proposes multi sensor fusion based on an effective calibration method for a perception system designed for mobile robots and intended for later object recognition. The perception system consists of a camera and a three-dimensional laser range finder. The three-dimensional laser range finder is based on a two-dimensional laser scanner and...

This paper studies an efficient solution for transporting scalable video over multi-standard integrated multimedia broadcasting networks, namely DVB-H and DTMB. A novel approach, called network-aware stream adaptation (NASA) framework, is developed within this context to deliver the scalable video streams to the end users over the underlying hetero...

This paper describes a novel research approach for multi-standard network convergence, integrating multiple broadcast networks for high-speed download with communications networks for the uplink and feedback channel. In particular, we study the integration of DVB-H and the Chinese digital television standard DTMB, with interaction provided via GSM/...

The ongoing large-scale deployment of digital broadcast networks adds an exciting new dimension to the wireless communication universe. This paper presents a novel global and unifying approach to network convergence with concurrent access and seamless handover between multiple broadcast networks and several communication networks for the uplink and...

While distance learning is attractive for many reasons, it also incurs some specific problems. One of them is the lack of direct interaction between teaching staff and the students, another one the large number of students for some courses. Therefore, the checking and correction of the homework submitted by the students often takes a lot of time an...

While distance learning is attractive for many reasons, it also incurs some specific problems. One of them is the lack of direct interaction between teaching staff and the students, another one the large number of students for some courses. Therefore, the checking and correction of the homework submitted by the students often takes a lot of time an...

This paper presents a new adaptive iterative learning rule for binary couplings networks. Unlike previous approaches, the algorithm adapts to pattern correlations during learning and succeeds to store highly correlated patterns. Also, by supplying a set of default stabilities to the learning rule, the recall properties of the network can be adjuste...

We describe HADES (short for Hamburg Design System), a Java-based and WWW-enabled environment for simulation and visualization. As an application example, our lab-course on computer architecture is presented.

The content-addressability of random patterns stored in neural networks trained with iterative learning rules is studied numerically under the effects of synaptic damage. The model of damage is random dilution of couplings with some probability lambda after learning. A simple model based on the distribution of stabilities rho (K) is proposed, which...