Max Schwarz

Benedikt T. Imbusch, Max Schwarz, Sven Behnke

The usefulness of deep learning models in robotics is largely dependent on the availability of training data. Manual annotation of training data is often infeasible. Synthetic data is a viable alternative, but suffers from domain gap. We propose a multi-step method to obtain training data without manual annotation effort: From 3D object meshes, we generate images using a modern synthesis pipeline. We utilize a state-of-the-art image-to-image translation method to adapt the synthetic images to the real domain, minimizing the domain gap in a learned manner. The translation network is trained from unpaired images, i.e. just requires an un-annotated collection of real images. The generated and refined images can then be used to train deep learning models for a particular task. We also propose and evaluate extensions to the translation method that further increase performance, such as patch-based training, which shortens training time and increases global consistency. We evaluate our method and demonstrate its effectiveness on two robotic datasets. We finally give insight into the learned refinement operations.

Andre Rochow, Max Schwarz, Michael Schreiber et al.

VR Facial Animation is necessary in applications requiring clear view of the face, even though a VR headset is worn. In our case, we aim to animate the face of an operator who is controlling our robotic avatar system. We propose a real-time capable pipeline with very fast adaptation for specific operators. In a quick enrollment step, we capture a sequence of source images from the operator without the VR headset which contain all the important operator-specific appearance information. During inference, we then use the operator keypoint information extracted from a mouth camera and two eye cameras to estimate the target expression and head pose, to which we map the appearance of a source still image. In order to enhance the mouth expression accuracy, we dynamically select an auxiliary expression frame from the captured sequence. This selection is done by learning to transform the current mouth keypoints into the source camera space, where the alignment can be determined accurately. We, furthermore, demonstrate an eye tracking pipeline that can be trained in less than a minute, a time efficient way to train the whole pipeline given a dataset that includes only complete faces, show exemplary results generated by our method, and discuss performance at the ANA Avatar XPRIZE semifinals.

Martin Link, Max Schwarz, Sven Behnke

We present a novel approach to estimating physical properties of objects from video. Our approach consists of a physics engine and a correction estimator. Starting from the initial observed state, object behavior is simulated forward in time. Based on the simulated and observed behavior, the correction estimator then determines refined physical parameters for each object. The method can be iterated for increased precision. Our approach is generic, as it allows for the use of an arbitrary - not necessarily differentiable - physics engine and correction estimator. For the latter, we evaluate both gradient-free hyperparameter optimization and a deep convolutional neural network. We demonstrate faster and more robust convergence of the learned method in several simulated 2D scenarios focusing on bin situations.

Arul Selvam Periyasamy, Catherine Capellen, Max Schwarz et al.

Object pose estimation is a key perceptual capability in robotics. We propose a fully-convolutional extension of the PoseCNN method, which densely predicts object translations and orientations. This has several advantages such as improving the spatial resolution of the orientation predictions -- useful in highly-cluttered arrangements, significant reduction in parameters by avoiding full connectivity, and fast inference. We propose and discuss several aggregation methods for dense orientation predictions that can be applied as a post-processing step, such as averaging and clustering techniques. We demonstrate that our method achieves the same accuracy as PoseCNN on the challenging YCB-Video dataset and provide a detailed ablation study of several variants of our method. Finally, we demonstrate that the model can be further improved by inserting an iterative refinement module into the middle of the network, which enforces consistency of the prediction.

Mayara E. Bonani, Max Schwarz, Sven Behnke

Domain adaptation is especially important for robotics applications, where target domain training data is usually scarce and annotations are costly to obtain. We present a method for self-supervised domain adaptation for the scenario where annotated source domain data (e.g. from synthetic generation) is available, but the target domain data is completely unannotated. Our method targets the semantic segmentation task and leverages a segmentation foundation model (Segment Anything Model) to obtain segment information on unannotated data. We take inspiration from recent advances in unsupervised local feature learning and propose an invariance-variance loss over the detected segments for regularizing feature representations in the target domain. Crucially, this loss structure and network architecture can handle overlapping segments and oversegmentation as produced by Segment Anything. We demonstrate the advantage of our method on the challenging YCB-Video and HomebrewedDB datasets and show that it outperforms prior work and, on YCB-Video, even a network trained with real annotations. Additionally, we provide insight through model ablations and show applicability to a custom robotic application.

Marius Beul, Max Schwarz, Jan Quenzel et al.

The Mohamed Bin Zayed International Robotics Challenge (MBZIRC) 2020 posed diverse challenges for unmanned aerial vehicles (UAVs). We present our four tailored UAVs, specifically developed for individual aerial-robot tasks of MBZIRC, including custom hardware- and software components. In Challenge 1, a target UAV is pursued using a high-efficiency, onboard object detection pipeline to capture a ball from the target UAV. A second UAV uses a similar detection method to find and pop balloons scattered throughout the arena. For Challenge 2, we demonstrate a larger UAV capable of autonomous aerial manipulation: Bricks are found and tracked from camera images. Subsequently, they are approached, picked, transported, and placed on a wall. Finally, in Challenge 3, our UAV autonomously finds fires using LiDAR and thermal cameras. It extinguishes the fires with an onboard fire extinguisher. While every robot features task-specific subsystems, all UAVs rely on a standard software stack developed for this particular and future competitions. We present our mostly open-source software solutions, including tools for system configuration, monitoring, robust wireless communication, high-level control, and agile trajectory generation. For solving the MBZIRC 2020 tasks, we advanced the state of the art in multiple research areas like machine vision and trajectory generation. We present our scientific contributions that constitute the foundation for our algorithms and systems and analyze the results from the MBZIRC competition 2020 in Abu Dhabi, where our systems reached second place in the Grand Challenge. Furthermore, we discuss lessons learned from our participation in this complex robotic challenge.

Andre Rochow, Max Schwarz, Michael Weinmann et al.

Novel view synthesis is required in many robotic applications, such as VR teleoperation and scene reconstruction. Existing methods are often too slow for these contexts, cannot handle dynamic scenes, and are limited by their explicit depth estimation stage, where incorrect depth predictions can lead to large projection errors. Our proposed method runs in real time on live streaming data and avoids explicit depth estimation by efficiently warping input images into the target frame for a range of assumed depth planes. The resulting plane sweep volume (PSV) is directly fed into our network, which first estimates soft PSV masks in a self-supervised manner, and then directly produces the novel output view. This improves efficiency and performance on transparent, reflective, thin, and feature-less scene parts. FaDIV-Syn can perform both interpolation and extrapolation tasks at 540p in real-time and outperforms state-of-the-art extrapolation methods on the large-scale RealEstate10k dataset. We thoroughly evaluate ablations, such as removing the Soft-Masking network, training from fewer examples as well as generalization to higher resolutions and stronger depth discretization. Our implementation is available.

Andre Rochow, Max Schwarz, Sven Behnke

The task of face reenactment is to transfer the head motion and facial expressions from a driving video to the appearance of a source image, which may be of a different person (cross-reenactment). Most existing methods are CNN-based and estimate optical flow from the source image to the current driving frame, which is then inpainted and refined to produce the output animation. We propose a transformer-based encoder for computing a set-latent representation of the source image(s). We then predict the output color of a query pixel using a transformer-based decoder, which is conditioned with keypoints and a facial expression vector extracted from the driving frame. Latent representations of the source person are learned in a self-supervised manner that factorize their appearance, head pose, and facial expressions. Thus, they are perfectly suited for cross-reenactment. In contrast to most related work, our method naturally extends to multiple source images and can thus adapt to person-specific facial dynamics. We also propose data augmentation and regularization schemes that are necessary to prevent overfitting and support generalizability of the learned representations. We evaluated our approach in a randomized user study. The results indicate superior performance compared to the state-of-the-art in terms of motion transfer quality and temporal consistency.

Andre Rochow, Max Schwarz, Sven Behnke

Facial animation in virtual reality environments is essential for applications that necessitate clear visibility of the user's face and the ability to convey emotional signals. In our scenario, we animate the face of an operator who controls a robotic Avatar system. The use of facial animation is particularly valuable when the perception of interacting with a specific individual, rather than just a robot, is intended. Purely keypoint-driven animation approaches struggle with the complexity of facial movements. We present a hybrid method that uses both keypoints and direct visual guidance from a mouth camera. Our method generalizes to unseen operators and requires only a quick enrolment step with capture of two short videos. Multiple source images are selected with the intention to cover different facial expressions. Given a mouth camera frame from the HMD, we dynamically construct the target keypoints and apply an attention mechanism to determine the importance of each source image. To resolve keypoint ambiguities and animate a broader range of mouth expressions, we propose to inject visual mouth camera information into the latent space. We enable training on large-scale speaking head datasets by simulating the mouth camera input with its perspective differences and facial deformations. Our method outperforms a baseline in quality, capability, and temporal consistency. In addition, we highlight how the facial animation contributed to our victory at the ANA Avatar XPRIZE Finals.

Anas Gouda, Max Schwarz, Christopher Reining et al.

Foundation models are a strong trend in deep learning and computer vision. These models serve as a base for applications as they require minor or no further fine-tuning by developers to integrate into their applications. Foundation models for zero-shot object segmentation such as Segment Anything (SAM) output segmentation masks from images without any further object information. When they are followed in a pipeline by an object identification model, they can perform object detection without training. Here, we focus on training such an object identification model. A crucial practical aspect for an object identification model is to be flexible in input size. As object identification is an image retrieval problem, a suitable method should handle multi-query multi-gallery situations without constraining the number of input images (e.g. by having fixed-size aggregation layers). The key solution to train such a model is the centroid triplet loss (CTL), which aggregates image features to their centroids. CTL yields high accuracy, avoids misleading training signals and keeps the model input size flexible. In our experiments, we establish a new state of the art on the ArmBench object identification task, which shows general applicability of our model. We furthermore demonstrate an integrated unseen object detection pipeline on the challenging HOPE dataset, which requires fine-grained detection. There, our pipeline matches and surpasses related methods which have been trained on dataset-specific data.

Peer Schüett, Max Schwarz, Sven Behnke

Augmented Reality is a promising technique for human-machine interaction. Especially in robotics, which always considers systems in their environment, it is highly beneficial to display visualizations and receive user input directly in exactly that environment. We explore this idea using the Microsoft HoloLens, with which we capture indoor environments and display interaction cues with known object classes. The 3D mesh recorded by the HoloLens is annotated on-line, as the user moves, with semantic classes using a projective approach, which allows us to use a state-of-the-art 2D semantic segmentation method. The results are fused onto the mesh; prominent object segments are identified and displayed in 3D to the user. Finally, the user can trigger actions by gesturing at the object. We both present qualitative results and analyze the accuracy and performance of our method in detail on an indoor dataset.

Max Schwarz, Christian Lenz, Andre Rochow et al.

Robotic avatars promise immersive teleoperation with human-like manipulation and communication capabilities. We present such an avatar system, based on the key components of immersive 3D visualization and transparent force-feedback telemanipulation. Our avatar robot features an anthropomorphic bimanual arm configuration with dexterous hands. The remote human operator drives the arms and fingers through an exoskeleton-based operator station, which provides force feedback both at the wrist and for each finger. The robot torso is mounted on a holonomic base, providing locomotion capability in typical indoor scenarios, controlled using a 3D rudder device. Finally, the robot features a 6D movable head with stereo cameras, which stream images to a VR HMD worn by the operator. Movement latency is hidden using spherical rendering. The head also carries a telepresence screen displaying a synthesized image of the operator with facial animation, which enables direct interaction with remote persons. We evaluate our system successfully both in a user study with untrained operators as well as a longer and more complex integrated mission. We discuss lessons learned from the trials and possible improvements.

Max Schwarz, Sven Behnke

We present a method for real-time stereo scene capture and remote VR visualization that allows a human operator to freely move their head and thus intuitively control their perspective during teleoperation. The stereo camera is mounted on a 6D robotic arm, which follows the operator's head pose. Existing VR teleoperation systems either induce high latencies on head movements, leading to motion sickness, or use scene reconstruction methods to allow re-rendering of the scene from different perspectives, which cannot handle dynamic scenes effectively. Instead, we present a decoupled approach which renders captured camera images as spheres, assuming constant distance. This allows very fast re-rendering on head pose changes while keeping the resulting temporary distortions during head translations small. We present qualitative examples, quantitative results in the form of lab experiments and a small user study, showing that our method outperforms other visualization methods.

Arul Selvam Periyasamy, Max Schwarz, Sven Behnke

We present SynPick, a synthetic dataset for dynamic scene understanding in bin-picking scenarios. In contrast to existing datasets, our dataset is both situated in a realistic industrial application domain -- inspired by the well-known Amazon Robotics Challenge (ARC) -- and features dynamic scenes with authentic picking actions as chosen by our picking heuristic developed for the ARC 2017. The dataset is compatible with the popular BOP dataset format. We describe the dataset generation process in detail, including object arrangement generation and manipulation simulation using the NVIDIA PhysX physics engine. To cover a large action space, we perform untargeted and targeted picking actions, as well as random moving actions. To establish a baseline for object perception, a state-of-the-art pose estimation approach is evaluated on the dataset. We demonstrate the usefulness of tracking poses during manipulation instead of single-shot estimation even with a naive filtering approach. The generator source code and dataset are publicly available.

Jan Quenzel, Malte Splietker, Dmytro Pavlichenko et al.

Every day, burning buildings threaten the lives of occupants and first responders trying to save them. Quick action is of essence, but some areas might not be accessible or too dangerous to enter. Robotic systems have become a promising addition to firefighting, but at this stage, they are mostly manually controlled, which is error-prone and requires specially trained personal. We present two systems for autonomous firefighting from air and ground we developed for the Mohamed Bin Zayed International Robotics Challenge (MBZIRC) 2020. The systems use LiDAR for reliable localization within narrow, potentially GNSS-restricted environments while maneuvering close to obstacles. Measurements from LiDAR and thermal cameras are fused to track fires, while relative navigation ensures successful extinguishing. We analyze and discuss our successful participation during the MBZIRC 2020, present further experiments, and provide insights into our lessons learned from the competition.

Christian Lenz, Jan Quenzel, Arul Selvam Periyasamy et al.

Autonomous robotic systems for various applications including transport, mobile manipulation, and disaster response are becoming more and more complex. Evaluating and analyzing such systems is challenging. Robotic competitions are designed to benchmark complete robotic systems on complex state-of-the-art tasks. Participants compete in defined scenarios under equal conditions. We present our UGV solution developed for the Mohamed Bin Zayed International Robotics Challenge 2020. Our hard- and software components to address the challenge tasks of wall building and fire fighting are integrated into a fully autonomous system. The robot consists of a wheeled omnidirectional base, a 6 DoF manipulator arm equipped with a magnetic gripper, a highly efficient storage system to transport box-shaped objects, and a water spraying system to fight fires. The robot perceives its environment using 3D LiDAR as well as RGB and thermal camera-based perception modules, is capable of picking box-shaped objects and constructing a pre-defined wall structure, as well as detecting and localizing heat sources in order to extinguish potential fires. A high-level planner solves the challenge tasks using the robot skills. We analyze and discuss our successful participation during the MBZIRC 2020 finals, present further experiments, and provide insights to our lessons learned.

Christian Lenz, Max Schwarz, Andre Rochow et al.

Constructing large structures with robots is a challenging task with many potential applications that requires mobile manipulation capabilities. We present two systems for autonomous wall building that we developed for the Mohamed Bin Zayed International Robotics Challenge 2020. Both systems autonomously perceive their environment, find bricks, and build a predefined wall structure. While the UGV uses a 3D LiDAR-based perception system which measures brick poses with high precision, the UAV employs a real-time camera-based system for visual servoing. We report results and insights from our successful participation at the MBZIRC 2020 Finals, additional lab experiments, and discuss the lessons learned from the competition.

Max Schwarz, Sven Behnke

Training data is the key ingredient for deep learning approaches, but difficult to obtain for the specialized domains often encountered in robotics. We describe a synthesis pipeline capable of producing training data for cluttered scene perception tasks such as semantic segmentation, object detection, and correspondence or pose estimation. Our approach arranges object meshes in physically realistic, dense scenes using physics simulation. The arranged scenes are rendered using high-quality rasterization with randomized appearance and material parameters. Noise and other transformations introduced by the camera sensors are simulated. Our pipeline can be run online during training of a deep neural network, yielding applications in life-long learning and in iterative render-and-compare approaches. We demonstrate the usability by learning semantic segmentation on the challenging YCB-Video dataset without actually using any training frames, where our method achieves performance comparable to a conventionally trained model. Additionally, we show successful application in a real-world regrasping system.

Umashankar Deekshith, Nishit Gajjar, Max Schwarz et al.

Correspondence estimation is one of the most widely researched and yet only partially solved area of computer vision with many applications in tracking, mapping, recognition of objects and environment. In this paper, we propose a novel way to estimate dense correspondence on an RGB image where visual descriptors are learned from video examples by training a fully convolutional network. Most deep learning methods solve this by training the network with a large set of expensive labeled data or perform labeling through strong 3D generative models using RGB-D videos. Our method learns from RGB videos using contrastive loss, where relative labeling is estimated from optical flow. We demonstrate the functionality in a quantitative analysis on rendered videos, where ground truth information is available. Not only does the method perform well on test data with the same background, it also generalizes to situations with a new background. The descriptors learned are unique and the representations determined by the network are global. We further show the applicability of the method to real-world videos.

Catherine Capellen, Max Schwarz, Sven Behnke

6D object pose estimation is a prerequisite for many applications. In recent years, monocular pose estimation has attracted much research interest because it does not need depth measurements. In this work, we introduce ConvPoseCNN, a fully convolutional architecture that avoids cutting out individual objects. Instead we propose pixel-wise, dense prediction of both translation and orientation components of the object pose, where the dense orientation is represented in Quaternion form. We present different approaches for aggregation of the dense orientation predictions, including averaging and clustering schemes. We evaluate ConvPoseCNN on the challenging YCB-Video Dataset, where we show that the approach has far fewer parameters and trains faster than comparable methods without sacrificing accuracy. Furthermore, our results indicate that the dense orientation prediction implicitly learns to attend to trustworthy, occlusion-free, and feature-rich object regions.

Arul Selvam Periyasamy, Max Schwarz, Sven Behnke

Robotic systems often require precise scene analysis capabilities, especially in unstructured, cluttered situations, as occurring in human-made environments. While current deep-learning based methods yield good estimates of object poses, they often struggle with large amounts of occlusion and do not take inter-object effects into account. Vision as inverse graphics is a promising concept for detailed scene analysis. A key element for this idea is a method for inferring scene parameter updates from the rasterized 2D scene. However, the rasterization process is notoriously difficult to invert, both due to the projection and occlusion process, but also due to secondary effects such as lighting or reflections. We propose to remove the latter from the process by mapping the rasterized image into an abstract feature space learned in a self-supervised way from pixel correspondences. Using only a light-weight inverse rendering module, this allows us to refine 6D object pose estimations in highly cluttered scenes by optimizing a simple pixel-wise difference in the abstract image representation. We evaluate our approach on the challenging YCB-Video dataset, where it yields large improvements and demonstrates a large basin of attraction towards the correct object poses.

Dmytro Pavlichenko, Diego Rodriguez, Christian Lenz et al.

In human-made scenarios, robots need to be able to fully operate objects in their surroundings, i.e., objects are required to be functionally grasped rather than only picked. This imposes very strict constraints on the object pose such that a direct grasp can be performed. Inspired by the anthropomorphic nature of humanoid robots, we propose an approach that first grasps an object with one hand, obtaining full control over its pose, and performs the functional grasp with the second hand subsequently. Thus, we develop a fully autonomous pipeline for dual-arm functional regrasping of novel familiar objects, i.e., objects never seen before that belong to a known object category, e.g., spray bottles. This process involves semantic segmentation, object pose estimation, non-rigid mesh registration, grasp sampling, handover pose generation and in-hand pose refinement. The latter is used to compensate for the unpredictable object movement during the first grasp. The approach is applied to a human-like upper body. To the best knowledge of the authors, this is the first system that exhibits autonomous bimanual functional regrasping capabilities. We demonstrate that our system yields reliable success rates and can be applied on-line to real-world tasks using only one off-the-shelf RGB-D sensor.

Tobias Klamt, Diego Rodriguez, Lorenzo Baccelliere et al.

Mobile manipulation robots have high potential to support rescue forces in disaster-response missions. Despite the difficulties imposed by real-world scenarios, robots are promising to perform mission tasks from a safe distance. In the CENTAURO project, we developed a disaster-response system which consists of the highly flexible Centauro robot and suitable control interfaces including an immersive tele-presence suit and support-operator controls on different levels of autonomy. In this article, we give an overview of the final CENTAURO system. In particular, we explain several high-level design decisions and how those were derived from requirements and extensive experience of Kerntechnische Hilfsdienst GmbH, Karlsruhe, Germany (KHG). We focus on components which were recently integrated and report about a systematic evaluation which demonstrated system capabilities and revealed valuable insights.

Patrick Stotko, Stefan Krumpen, Max Schwarz et al.

Applications like disaster management and industrial inspection often require experts to enter contaminated places. To circumvent the need for physical presence, it is desirable to generate a fully immersive individual live teleoperation experience. However, standard video-based approaches suffer from a limited degree of immersion and situation awareness due to the restriction to the camera view, which impacts the navigation. In this paper, we present a novel VR-based practical system for immersive robot teleoperation and scene exploration. While being operated through the scene, a robot captures RGB-D data that is streamed to a SLAM-based live multi-client telepresence system. Here, a global 3D model of the already captured scene parts is reconstructed and streamed to the individual remote user clients where the rendering for e.g. head-mounted display devices (HMDs) is performed. We introduce a novel lightweight robot client component which transmits robot-specific data and enables a quick integration into existing robotic systems. This way, in contrast to first-person exploration systems, the operators can explore and navigate in the remote site completely independent of the current position and view of the capturing robot, complementing traditional input devices for teleoperation. We provide a proof-of-concept implementation and demonstrate the capabilities as well as the performance of our system regarding interactive object measurements and bandwidth-efficient data streaming and visualization. Furthermore, we show its benefits over purely video-based teleoperation in a user study revealing a higher degree of situation awareness and a more precise navigation in challenging environments.

Tobias Klamt, Max Schwarz, Christian Lenz et al.

Solving mobile manipulation tasks in inaccessible and dangerous environments is an important application of robots to support humans. Example domains are construction and maintenance of manned and unmanned stations on the moon and other planets. Suitable platforms require flexible and robust hardware, a locomotion approach that allows for navigating a wide variety of terrains, dexterous manipulation capabilities, and respective user interfaces. We present the CENTAURO system which has been designed for these requirements and consists of the Centauro robot and a set of advanced operator interfaces with complementary strength enabling the system to solve a wide range of realistic mobile manipulation tasks. The robot possesses a centaur-like body plan and is driven by torque-controlled compliant actuators. Four articulated legs ending in steerable wheels allow for omnidirectional driving as well as for making steps. An anthropomorphic upper body with two arms ending in five-finger hands enables human-like manipulation. The robot perceives its environment through a suite of multimodal sensors. The resulting platform complexity goes beyond the complexity of most known systems which puts the focus on a suitable operator interface. An operator can control the robot through a telepresence suit, which allows for flexibly solving a large variety of mobile manipulation tasks. Locomotion and manipulation functionalities on different levels of autonomy support the operation. The proposed user interfaces enable solving a wide variety of tasks without previous task-specific training. The integrated system is evaluated in numerous teleoperated experiments that are described along with lessons learned.

Dmytro Pavlichenko, Diego Rodriguez, Max Schwarz et al.

Autonomous dual-arm manipulation is an essential skill to deploy robots in unstructured scenarios. However, this is a challenging undertaking, particularly in terms of perception and planning. Unstructured scenarios are full of objects with different shapes and appearances that have to be grasped in a very specific manner so they can be functionally used. In this paper we present an integrated approach to perform dual-arm pick tasks autonomously. Our method consists of semantic segmentation, object pose estimation, deformable model registration, grasp planning and arm trajectory optimization. The entire pipeline can be executed on-board and is suitable for on-line grasping scenarios. For this, our approach makes use of accumulated knowledge expressed as convolutional neural network models and low-dimensional latent shape spaces. For manipulating objects, we propose a stochastic trajectory optimization that includes a kinematic chain closure constraint. Evaluation in simulation and on the real robot corroborates the feasibility and applicability of the proposed methods on a task of picking up unknown watering cans and drills using both arms.

Arul Selvam Periyasamy, Max Schwarz, Sven Behnke

Object pose estimation is a crucial prerequisite for robots to perform autonomous manipulation in clutter. Real-world bin-picking settings such as warehouses present additional challenges, e.g., new objects are added constantly. Most of the existing object pose estimation methods assume that 3D models of the objects is available beforehand. We present a pipeline that requires minimal human intervention and circumvents the reliance on the availability of 3D models by a fast data acquisition method and a synthetic data generation procedure. This work builds on previous work on semantic segmentation of cluttered bin-picking scenes to isolate individual objects in clutter. An additional network is trained on synthetic scenes to estimate object poses from a cropped object-centered encoding extracted from the segmentation results. The proposed method is evaluated on a synthetic validation dataset and cluttered real-world scenes.

Max Schwarz, David Droeschel, Christian Lenz et al.

The Mohamed Bin Zayed International Robotics Challenge (MBZIRC) 2017 has defined ambitious new benchmarks to advance the state-of-the-art in autonomous operation of ground-based and flying robots. In this article, we describe our winning entry to MBZIRC Challenge 2: the mobile manipulation robot Mario. It is capable of autonomously solving a valve manipulation task using a wrench tool detected, grasped, and finally employed to turn a valve stem. Mario's omnidirectional base allows both fast locomotion and precise close approach to the manipulation panel. We describe an efficient detector for medium-sized objects in 3D laser scans and apply it to detect the manipulation panel. An object detection architecture based on deep neural networks is used to find and select the correct tool from grayscale images. Parametrized motion primitives are adapted online to percepts of the tool and valve stem in order to turn the stem. We report in detail on our winning performance at the challenge and discuss lessons learned.

Max Schwarz, Christian Lenz, Germán Martín García et al.

Robotic picking from cluttered bins is a demanding task, for which Amazon Robotics holds challenges. The 2017 Amazon Robotics Challenge (ARC) required stowing items into a storage system, picking specific items, and packing them into boxes. In this paper, we describe the entry of team NimbRo Picking. Our deep object perception pipeline can be quickly and efficiently adapted to new items using a custom turntable capture system and transfer learning. It produces high-quality item segments, on which grasp poses are found. A planning component coordinates manipulation actions between two robot arms, minimizing execution time. The system has been demonstrated successfully at ARC, where our team reached second places in both the picking task and the final stow-and-pick task. We also evaluate individual components.

Max Schwarz, Tobias Rodehutskors, David Droeschel et al.

Robots that solve complex tasks in environments too dangerous for humans to enter are desperately needed, e.g. for search and rescue applications. We describe our mobile manipulation robot Momaro, with which we participated successfully in the DARPA Robotics Challenge. It features a unique locomotion design with four legs ending in steerable wheels, which allows it both to drive omnidirectionally and to step over obstacles or climb. Furthermore, we present advanced communication and teleoperation approaches, which include immersive 3D visualization, and 6D tracking of operator head and arm motions. The proposed system is evaluated in the DARPA Robotics Challenge, the DLR SpaceBot Cup Qualification and lab experiments. We also discuss the lessons learned from the competitions.

Max Schwarz, Anton Milan, Arul Selvam Periyasamy et al.

Autonomous robotic manipulation in clutter is challenging. A large variety of objects must be perceived in complex scenes, where they are partially occluded and embedded among many distractors, often in restricted spaces. To tackle these challenges, we developed a deep-learning approach that combines object detection and semantic segmentation. The manipulation scenes are captured with RGB-D cameras, for which we developed a depth fusion method. Employing pretrained features makes learning from small annotated robotic data sets possible. We evaluate our approach on two challenging data sets: one captured for the Amazon Picking Challenge 2016, where our team NimbRo came in second in the Stowing and third in the Picking task, and one captured in disaster-response scenarios. The experiments show that object detection and semantic segmentation complement each other and can be combined to yield reliable object perception.

Marcell Missura, Cedrick Münstermann, Philipp Allgeuer et al.

Over the past few years, soccer-playing humanoid robots have advanced significantly. Elementary skills, such as bipedal walking, visual perception, and collision avoidance have matured enough to allow for dynamic and exciting games. When two robots are fighting for the ball, they frequently push each other and balance recovery becomes crucial. In this paper, we report on insights we gained from systematic push experiments performed on a bipedal model and outline an online learning method we used to improve its push-recovery capabilities. In addition, we describe how the localization ambiguity introduced by the uniform goal color was resolved and report on the results of the RoboCup 2013 competition.

Philipp Allgeuer, Max Schwarz, Julio Pastrana et al.

Over the past few years, a number of successful humanoid platforms have been developed, including the Nao and the DARwIn-OP, both of which are used by many research groups for the investigation of bipedal walking, full-body motions, and human-robot interaction. The NimbRo-OP is an open humanoid platform under development by team NimbRo of the University of Bonn. Significantly larger than the two aforementioned humanoids, this platform has the potential to interact with a more human-scale environment. This paper describes a software framework for the NimbRo-OP that is based on the Robot Operating System (ROS) middleware. The software provides functionality for hardware abstraction, visual perception, and behavior generation, and has been used to implement basic soccer skills. These were demonstrated at RoboCup 2013, as part of the winning team of the Humanoid League competition.

Max Schwarz, Julio Pastrana, Philipp Allgeuer et al.

In recent years, the introduction of affordable platforms in the KidSize class of the Humanoid League has had a positive impact on the performance of soccer robots. The lack of readily available larger robots, however, severely affects the number of participants in Teen- and AdultSize and consequently the progress of research that focuses on the challenges arising with robots of larger weight and size. This paper presents the first hardware release of a low cost Humanoid TeenSize open platform for research, the first software release, and the current state of ROS-based software development. The NimbRo-OP robot was designed to be easily manufactured, assembled, repaired, and modified. It is equipped with a wide-angle camera, ample computing power, and enough torque to enable full-body motions, such as dynamic bipedal locomotion, kicking, and getting up.

Tobias Klamt, Diego Rodriguez, Max Schwarz et al.

Mobile manipulation tasks are one of the key challenges in the field of search and rescue (SAR) robotics requiring robots with flexible locomotion and manipulation abilities. Since the tasks are mostly unknown in advance, the robot has to adapt to a wide variety of terrains and workspaces during a mission. The centaur-like robot Centauro has a hybrid legged-wheeled base and an anthropomorphic upper body to carry out complex tasks in environments too dangerous for humans. Due to its high number of degrees of freedom, controlling the robot with direct teleoperation approaches is challenging and exhausting. Supervised autonomy approaches are promising to increase quality and speed of control while keeping the flexibility to solve unknown tasks. We developed a set of operator assistance functionalities with different levels of autonomy to control the robot for challenging locomotion and manipulation tasks. The integrated system was evaluated in disaster response scenarios and showed promising performance.