Accurate inertial parameter identification is crucial for the simulation and control of robots encountering intermittent contacts with the environment. Classically, robots' inertial parameters are obtained from CAD models that are not precise (and sometimes not available, e.g., Spot from Boston Dynamics), hence requiring identification. To do that, existing methods require access to contact force measurement, a modality not present in modern quadruped and humanoid robots. This paper presents an alternative technique that utilizes joint current/torque measurements -a standard sensing modality in modern robots- to identify inertial parameters without requiring direct contact force measurements. By projecting the whole-body dynamics into the null space of contact constraints, we eliminate the dependency on contact forces and reformulate the identification problem as a linear matrix inequality that can handle physical and geometrical constraints. We compare our proposed method against a common black-box identification method using a deep neural network and show that incorporating physical consistency significantly improves the sample efficiency and generalizability of the model. Finally, we validate our method on the Spot quadruped robot across various locomotion tasks, showcasing its accuracy and generalizability in real-world scenarios over different gaits.

Fruit monitoring plays an important role in crop management, and rising global fruit consumption combined with labor shortages necessitates automated monitoring with robots. However, occlusions from plant foliage often hinder accurate shape and pose estimation. Therefore, we propose an active fruit shape and pose estimation method that physically manipulates occluding leaves to reveal hidden fruits. This paper introduces a framework that plans robot actions to maximize visibility and minimize leaf damage. We developed a novel scene-consistent shape completion technique to improve fruit estimation under heavy occlusion and utilize a perception-driven deformation graph model to predict leaf deformation during planning. Experiments on artificial and real sweet pepper plants demonstrate that our method enables robots to safely move leaves aside, exposing fruits for accurate shape and pose estimation, outperforming baseline methods. Project page: https://shaoxiongyao.github.io/lmap-ssc/.

In complex scenarios where typical pick-and-place techniques are insufficient, often non-prehensile manipulation can ensure that a robot is able to fulfill its task. However, non-prehensile manipulation is challenging due to its underactuated nature with hybrid-dynamics, where a robot needs to reason about an object’s long-term behavior and contact-switching, while being robust to contact uncertainty. The presence of clutter in the workspace further complicates this task, introducing the need to include more advanced spatial analysis to avoid unwanted collisions. Building upon prior work on reinforcement learning with multimodal categorical exploration for planar pushing, we propose to incorporate location-based attention to enable robust manipulation in cluttered scenes. Unlike previous approaches addressing this obstacle avoiding pushing task, our framework requires no predefined global paths and considers the desired target orientation of the manipulated object. Experimental results in simulation as well as with a real KUKA iiwa robot arm demonstrate that our learned policy manipulates objects successfully while avoiding collisions through complex obstacle configurations, including dynamic obstacles, to reach the desired target pose.

Aligning robot navigation with human preferences is essential for ensuring comfortable, and predictable robot movement in shared spaces. While preference-based learning methods, such as reinforcement learning from human feedback (RLHF), enable this alignment, the choice of the preference collection interface may influence the process. Traditional 2D interfaces provide structured views but lack spatial depth, whereas immersive VR offers richer perception, potentially affecting preference articulation. This study systematically examines how the interface modality impacts human preference collection and navigation policy alignment. We introduce a novel dataset of 2,325 human preference queries collected through both VR and 2D interfaces, revealing significant differences in user experience, preference consistency, and policy outcomes. Our findings highlight the trade-offs between immersion, perception, and preference reliability, emphasizing the importance of interface selection in preference-based robot learning. The dataset is available to support future research.

Preference-Aligned robot navigation in human environments is typically achieved through learning-based approaches, utilizing user feedback or demonstrations for personalization. However, personal preferences are subject to change and might even be context-dependent. Yet traditional reinforcement learning (RL) approaches with static reward functions often fall short in adapting to evolving user preferences, inevitably reflecting demonstrations once training is completed. This paper introduces a structured framework that combines demonstration-based learning with multi-objective reinforcement learning (MORL). To ensure real-world applicability, our approach allows for dynamic adaptation of the robot navigation policy to changing user preferences without retraining. It fluently modulates the amount of demonstration data reflection and other preference-related objectives. Through rigorous evaluations, including a baseline comparison and sim-to-real transfer on two robots, we demonstrate our framework’s capability to adapt to user preferences accurately while achieving high navigational performance in terms of collision avoidance and goal pursuance.

Automating labor-intensive tasks such as crop monitoring with robots is essential for enhancing production and conserving resources. However, autonomously monitoring horticulture crops remains challenging due to their complex structures, which often result in fruit occlusions. Existing view planning methods attempt to reduce occlusions but either struggle to achieve adequate coverage or incur high robot motion costs. We introduce a global optimization approach for view motion planning that aims to minimize robot motion costs while maximizing fruit coverage. To this end, we leverage coverage constraints derived from the set covering problem (SCP) within a shortest Hamiltonian path problem (SHPP) formulation. While both SCP and SHPP are well-established, their tailored integration enables a unified framework that computes a global view path with minimized motion while ensuring full coverage of selected targets. Given the NP-hard nature of the problem, we employ a region-prior-based selection of coverage targets and a sparse graph structure to achieve effective optimization outcomes within a limited time. Experiments in simulation demonstrate that our method detects more fruits, enhances surface coverage, and achieves higher volume accuracy than the motion-efficient baseline with a moderate increase in motion cost, while significantly reducing motion costs compared to the coverage-focused baseline. Real-World experiments further confirm the practical applicability of our approach.

For scene understanding in unstructured environments, an accurate and uncertainty-aware metric-semantic mapping is required to enable informed action selection by autonomous systems. Existing mapping methods often suffer from overconfident semantic predictions, and sparse and noisy depth sensing, leading to inconsistent map representations. In this paper, we therefore introduce EvidMTL, a multitask learning framework that uses evidential heads for depth estimation and semantic segmentation, enabling uncertainty-aware inference from monocular RGB images. To enable uncertainty-calibrated evidential multi-task learning, we propose a novel evidential depth loss function that jointly optimizes the belief strength of the depth prediction in conjunction with evidential segmentation loss. Building on this, we present EvidKimera, an uncertainty-aware semantic surface mapping framework, which uses evidential depth and semantics prediction for improved 3D metric-semantic consistency. We train and evaluate EvidMTL on the NYUDepthV2 and assess its zero-shot performance on ScanNetV2, demonstrating superior uncertainty estimation compared to conventional approaches while maintaining comparable depth estimation and semantic segmentation. In zero-shot mapping tests on ScanNetV2, EvidKimera outperforms Kimera by 30% in semantic surface mapping accuracy and consistency, highlighting the benefits of uncertainty-aware mapping and underscoring its potential for real-world robotic applications.

Autonomous assembly is an essential capability for industrial and service robots, with Peg-in-Hole (PiH) insertion being one of the core tasks. However, PiH assembly in unknown environments is still challenging due to uncertainty in task parameters, such as the hole position and orientation, resulting from sensor noise. Although context-based meta reinforcement learning (RL) methods have been previously presented to adapt to unknown task parameters in PiH assembly tasks, the performance depends on a sample-inefficient procedure or human demonstrations. Thus, to enhance the applicability of meta RL in real-world PiH assembly tasks, we propose to train the agent to use information from the robot’s forward kinematics and an uncalibrated camera. Furthermore, we improve the applicability by efficiently adapting the meta-trained agent to use data from force/torque sensor. Finally, we propose an adaptation procedure for out-of-distribution tasks whose parameters are different from the training tasks. Experiments on simulated and real robots prove that our modifications enhance the sample efficiency during meta training, real-world adaptation performance, and generalization of the context-based meta RL agent in PiH assembly tasks compared to previous approaches.

Robust and accurate proprioceptive state estimation of the main body is crucial for legged robots to execute tasks in extreme environments where exteroceptive sensors, such as LiDARs and cameras, may become unreliable. In this paper, we propose DogLegs, a state estimation system for legged robots that fuses the measurements from a body-mounted inertial measurement unit (Body-IMU), joint encoders, and multiple leg-mounted IMUs (Leg-IMU) using an extended Kalman filter (EKF). The filter system contains the error states of all IMU frames. The Leg-IMUs are used to detect foot contact, thereby providing zero-velocity measurements to update the state of the Leg-IMU frames. Additionally, we compute the relative position constraints between the Body-IMU and Leg-IMUs by the leg kinematics and use them to update the main body state and reduce the error drift of the individual IMU frames. Field experimental results have shown that our proposed DogLegs system achieves better state estimation accuracy compared to the traditional leg odometry method (using only Body-IMU and joint encoders) across various terrains. We make our datasets publicly available to benefit the research community (https://github.com/YibinWu/leg-odometry).

End-to-end robot policies achieve high performance through neural networks trained via reinforcement learning (RL). Yet, their black box nature and abstract reasoning pose challenges for human-robot interaction (HRI), because humans may experience difficulty in understanding and predicting the robot’s navigation decisions, hindering trust development. We present a virtual reality (VR) interface that visualizes explainable AI (XAI) outputs and the robot’s lidar perception to support intuitive interpretation of RL-based navigation behavior. By visually highlighting objects based on their attribution scores, the interface grounds abstract policy explanations in the scene context. This XAI visualization bridges the gap between obscure numerical XAI attribution scores and a human-centric semantic level of explanation. A within-subjects study with 24 participants evaluated the effectiveness of our interface for four visualization conditions combining XAI and lidar. Participants ranked scene objects across navigation scenarios based on their importance to the robot, followed by a questionnaire assessing subjective understanding and predictability. Results show that semantic projection of attributions significantly enhances non-expert users’ objective understanding and subjective awareness of robot behavior. In addition, lidar visualization further improves perceived predictability, underscoring the value of integrating XAI and sensor for transparent, trustworthy HRI.

Mobile robots increasingly operate alongside humans but are often out of sight, so that humans need to rely on the sounds of the robots to recognize their presence. For successful human-robot interaction (HRI), it is therefore crucial to understand how humans perceive robots by their consequential sounds, i.e., operating noise. Prior research suggests that the sound of a quadruped Go1 is more detectable than that of a wheeled Turtlebot. This study builds on this and examines the human ability to localize consequential sounds of three robots (quadruped Go1, wheeled Turtlebot 2i, wheeled HSR) in Virtual Reality. In a within-subjects design, we assessed participants’ localization performance for the robots with and without an acoustic vehicle alerting system (AVAS) for two velocities (0.3, 0.8 m/s) and two trajectories (head-on, radial). In each trial, participants were presented with the sound of a moving robot for 3 s and were tasked to point at its final position (localization task). Localization errors were measured as the absolute angular difference between the participants’ estimated and the actual robot position. Results showed that the robot type significantly influenced the localization accuracy and precision, with the sound of the wheeled HSR (especially without AVAS) performing worst under all experimental conditions. Surprisingly, participants rated the HSR sound as more positive, less annoying, and more trustworthy than the Turtlebot and Go1 sound. This reveals a tension between subjective evaluation and objective auditory localization performance. Our findings highlight consequential robot sounds as a critical factor for designing intuitive and effective HRI, with implications for human-centered robot design and social navigation.

Existing view planning systems either adopt an iterative paradigm using next-best views (NBV) or a one-shot pipeline relying on the set-covering view-planning (SCVP) network. However, neither of these methods can concurrently guarantee both high-quality and high-efficiency reconstruction of 3-D unknown objects. To tackle this challenge, we introduce a crucial hypothesis: with the availability of more information about the unknown object, the prediction quality of the SCVP network improves. There are two ways to provide extra information: first, leveraging perception data obtained from NBVs, and second, training on an expanded dataset of multiview inputs. In this work, we introduce a novel combined pipeline that incorporates a single NBV before activating the proposed multiview-activated (MA-)SCVP network. The MA-SCVP is trained on a multiview dataset generated by our long-tail sampling method, which addresses the issue of unbalanced multiview inputs and enhances the network performance. Extensive simulated experiments substantiate that our system demonstrates a significant surface coverage increase and a substantial 45% reduction in movement cost compared to state-of-the-art systems. Real-world experiments justify the capability of our system for generalization and deployment.

Active object reconstruction using autonomous robots is gaining great interest. A primary goal in this task is to maximize the information of the object to be reconstructed, given limited on-board resources. Previous view planning methods exhibit inefficiency since they rely on an iterative paradigm based on explicit representations, consisting of (1) planning a path to the next-best view only; and (2) requiring a considerable number of less-gain views in terms of surface coverage. To address these limitations, we propose to integrate implicit representations into the One-Shot View Planning (OSVP). The key idea behind our approach is to use implicit representations to obtain the small missing surface areas instead of observing them with extra views. Therefore, we design a deep neural network, named OSVP, to directly predict a set of views given a dense point cloud refined from an initial sparse observation. To train our OSVP network, we generate supervision labels using dense point clouds refined by implicit representations and set covering optimization problems. Simulated experiments show that our method achieves sufficient reconstruction quality, outperforming several baselines under limited view and movement budgets. We further demonstrate the applicability of our approach in a real-world object reconstruction scenario.

Neural Radiance Fields (NeRFs) are gaining significant interest for online active object reconstruction due to their exceptional memory efficiency and requirement for only posed RGB inputs. Previous NeRF-based view planning methods exhibit computational inefficiency since they rely on an iterative paradigm, consisting of (1) retraining the NeRF when new images arrive; and (2) planning a path to the next best view only. To address these limitations, we propose a non-iterative pipeline based on the Prediction of the Required number of Views (PRV). The key idea behind our approach is that the required number of views to reconstruct an object depends on its complexity. Therefore, we design a deep neural network, named PRVNet, to predict the required number of views, allowing us to tailor the data acquisition based on the object complexity and plan a globally shortest path. To train our PRVNet, we generate supervision labels using the ShapeNet dataset. Simulated experiments show that our PRV-based view planning method outperforms baselines, achieving good reconstruction quality while significantly reducing movement cost and planning time. We further justify the generalization ability of our approach in a real-world experiment.

In this paper, we introduce a novel approach to centroidal state estimation, which plays a crucial role in predictive model-based control strategies for dynamic legged locomotion. Our approach uses the Koopman operator theory to transform the robot’s complex nonlinear dynamics into a linear system, by employing dynamic mode decomposition and deep learning for model construction. We evaluate both models on their linearization accuracy and capability to capture both fast and slow dynamic system responses. We then select the most suitable model for estimation purposes, and integrate it within a moving horizon estimator. This estimator is formulated as a convex quadratic program to facilitate robust, real-time centroidal state estimation. Through extensive simulation experiments on a quadruped robot executing various dynamic gaits, our data-driven framework outperforms conventional Extended Kalman Filtering technique based on nonlinear dynamics. Our estimator addresses challenges posed by force/torque measurement noise in highly dynamic motions and accurately recovers the centroidal states, demonstrating the adaptability and effectiveness of the Koopman-based linear representation for complex locomotive behaviors. Importantly, our model based on dynamic mode decomposition, trained with two locomotion patterns (trot and jump), successfully estimates the centroidal states for a different motion (bound) without retraining.

Horticultural tasks such as pruning and selective harvesting are labor intensive and horticultural staff are hard to find. Automating these tasks is challenging due to the semi-structured greenhouse workspaces, changing environmental conditions such as lighting, dense plant growth with many occlusions, and the need for gentle manipulation of non-rigid plant organs. In this work, we present the three-armed system HortiBot, with two arms for manipulation and a third arm as an articulated head for active perception using stereo cameras. Its perception system detects not only peppers, but also peduncles and stems in real time, and performs online data association to build a world model of pepper plants. Collision-aware online trajectory generation allows all three arms to safely track their respective targets for observation, grasping, and cutting. We integrated perception and manipulation to perform selective harvesting of peppers and evaluated the system in lab experiments. Using active perception coupled with end-effector force torque sensing for compliant manipulation, HortiBot achieves high success rates in our indoor pepper plant mock-up.

Object reconstruction is relevant for many autonomous robotic tasks that require interaction with the environment. A key challenge in such scenarios is planning view configurations to collect informative measurements for reconstructing an initially unknown object. One-shot view planning enables efficient data collection by predicting view configurations and planning the globally shortest path connecting all views at once. However, prior knowledge about the object is required to conduct one-shot view planning. In this work, we propose a novel one-shot view planning approach that utilizes the powerful 3D generation capabilities of diffusion models as priors. By incorporating such geometric priors into our pipeline, we achieve effective one-shot view planning starting with only a single RGB image of the object to be reconstructed. Our planning experiments in simulation and real-world setups indicate that our approach balances well between object reconstruction quality and movement cost.

Mobile robots are increasingly being used in noisy environments for social purposes, e.g. to provide support in healthcare or public spaces. Since these robots also operate beyond human sight, the question arises as to how different robot types, ambient noise or cognitive engagement impacts the detection of the robots by their sound. To address this research gap, we conducted a user study measuring auditory detection distances for a wheeled (Turtlebot 2i) and quadruped robot (Unitree Go 1), which emit different consequential sounds when moving. Additionally, we also manipulated background noise levels and participants’ engagement in a secondary task during the study. Our results showed that the quadruped robot sound was detected significantly better (i.e., at a larger distance) than the wheeled one, which demonstrates that the movement mechanism has a meaningful impact on the auditory detectability. The detectability for both robots diminished significantly as background noise increased. But even in high background noise, participants detected the quadruped robot at a significantly larger distance. The engagement in a secondary task had hardly any impact. In essence, these findings highlight the critical role of distinguishing auditory characteristics of different robots to improve the smooth human-centered navigation of mobile robots in noisy environments.

To align mobile robot navigation policies with user preferences through reinforcement learning from human feedback (RLHF), reliable and behavior-diverse user queries are required. However, deterministic policies fail to generate a variety of navigation trajectory suggestions for a given navigation task. In this paper, we introduce EnQuery, a query generation approach using an ensemble of policies that achieve behavioral diversity through a regularization term. For a given navigation task, EnQuery produces multiple navigation trajectory suggestions, thereby optimizing the efficiency of preference data collection with fewer queries. Our methodology demonstrates superior performance in aligning navigation policies with user preferences in low-query regimes, offering enhanced policy convergence from sparse preference queries. The evaluation is complemented with a novel explainability representation, capturing full scene navigation behavior of the mobile robot in a single plot. Our code is available online at https://github.com/hrl-bonn/EnQuery.

In 1997, the very first tour guide robot RHINO was deployed in a museum in Germany. With the ability to navigate autonomously through the environment, the robot gave tours to over 2,000 visitors. Today, RHINO itself has become an exhibit and is no longer operational In this paper, we present RHINO-VR, an interactive museum exhibit using virtual reality (VR) that allows museum visitors to experience the historical robot RHINO in operation in a virtual museum. RHINO-VR, unlike static exhibits, enables users to familiarize themselves with basic mobile robotics concepts without the fear of damaging the exhibit. In the virtual environment, the user is able to interact with RHINO in VR by pointing to a location to which the robot should navigate and observing the corresponding actions of the robot. To include other visitors who cannot use the VR, we provide an external observation view to make RHINO visible to them. We evaluated our system by measuring the frame rate of the VR simulation, comparing the generated virtual 3D models with the originals, and conducting a user study. The user study showed that RHINO-VR improved the visitors’ understanding of the robot’s functionality and that they would recommend experiencing the VR exhibit to others.

Reinforcement learning (RL) has recently proven great success in various domains. Yet, the design of the reward function requires detailed domain expertise and tedious fine-tuning to ensure that agents are able to learn the desired behaviour. Using a sparse reward conveniently mitigates these challenges. However, the sparse reward represents a challenge on its own, often resulting in unsuccessful training of the agent. In this paper, we therefore address the sparse reward problem in RL. Our goal is to find an effective alternative to reward shaping, without using costly human demonstrations, that would also be applicable to a wide range of domains. Hence, we propose to use model predictive control (MPC) as an experience source for training RL agents in sparse reward environments. Without the need for reward shaping, we successfully apply our approach in the field of mobile robot navigation both in simulation and real-world experiments with a Kuboki Turtlebot 2. We furthermore demonstrate great improvement over pure RL algorithms in terms of success rate as well as number of collisions and timeouts. Our experiments show that MPC as an experience source improves the agent's learning process for a given task in the case of sparse rewards.

State-of-the-art legged robots are either capable of measuring torque at the output of their drive systems, or have transparent drive systems which enable the computation of joint torques from motor currents. In either case, this sensor modality is seldom used in state estimation. In this paper, we propose to use joint torque measurements to estimate the centroidal states of legged robots. To do so, we project the whole-body dynamics of a legged robot into the nullspace of the contact constraints, allowing expression of the dynamics independent of the contact forces. Using the constrained dynamics and the centroidal momentum matrix, we are able to directly relate joint torques and centroidal states dynamics. Using the resulting model as the process model of an Extended Kalman Filter (EKF), we fuse the torque measurement in the centroidal state estimation problem. Through real-world experiments on a quadruped robot executing different gaits, we demonstrate that the estimated centroidal states from our torque-based EKF drastically improve the recovery of these quantities compared to direct computation.

Monitoring the traits of plants and fruits is a fundamental task in horticulture. With accurate measurements, farmers can predict the yield of their crops and use this information for making informed management decisions, and breeders can use it for variety selection. Agricultural robotic applications promise to automate this monitoring task. In this paper, we address the problem of monitoring fruit growth and investigate the matching of fruits recorded in commercial greenhouses at different growth stages based on data recorded from terrestrial laser scanners. This is challenging as fruits appear highly similar, change over time, and are subject to severe occlusions. We first propose a fruit descriptor, which captures the topology of the fruit surroundings to facilitate the matching between different points in time. We capture and describe the relationship between a fruit and its neighbors such that our descriptors are less affected by the growth over time. Furthermore, we define a matching cost function and use an optimal assignment algorithm to match the fruit observations taken in different weeks. The experiments show that our descriptor achieves a high spatio-temporal matching accuracy, which is superior to the commonly used geometric point cloud descriptors.

Viewpoint planning is an important task in any application where objects or scenes need to be viewed from different angles to achieve sufficient coverage. The mapping of confined spaces such as shelves is an especially challenging task since objects occlude each other and the scene can only be observed from the front, posing limitations on the possible viewpoints. In this paper, we propose a deep reinforcement learning framework that generates promising views aiming at reducing the map entropy. Additionally, the pipeline extends standard viewpoint planning by predicting adequate minimally invasive push actions to uncover occluded objects and increase the visible space. Using a 2.5D occupancy height map as state representation that can be efficiently updated, our system decides whether to plan a new viewpoint or perform a push. To learn feasible pushes, we use a neural network to sample push candidates on the map based on training data provided by human experts. As simulated and real-world experimental results with a robotic arm show, our system is able to significantly increase the mapped space compared to different baselines, while the executed push actions highly benefit the viewpoint planner with only minor changes to the object configuration.

For the best human-robot interaction experience, the robot's navigation policy should take into account personal preferences of the user. In this paper, we present a learning framework complemented by a perception pipeline to train a depth vision-based, personalized navigation controller from user demonstrations. Our virtual reality interface enables the demonstration of robot navigation trajectories under motion of the user for dynamic interaction scenarios. The novel perception pipeline enrolls a variational autoencoder in combination with a motion predictor. It compresses the perceived depth images to a latent state representation to enable efficient reasoning of the learning agent about the robot's dynamic environment. In a detailed analysis and ablation study, we evaluate different configurations of the perception pipeline. To further quantify the navigation controller's quality of personalization, we develop and apply a novel metric to measure preference reflection based on the Frechet Distance. We discuss the robot's navigation performance in various virtual scenes and demonstrate the first personalized robot navigation controller that solely relies on depth images. A supplemental video highlighting our approach is available online11Full video: hrl.uni-bonn.de/publications/deheuve123iros_learning.mp4.

Active perception for fruit mapping and harvesting is a difficult task since occlusions occur frequently and the location as well as size of fruits change over time. State-of-the-art viewpoint planning approaches utilize computationally expensive ray casting operations to find good viewpoints aiming at maximizing information gain and covering the fruits in the scene. In this paper, we present a novel viewpoint planning approach that explicitly uses information about the predicted fruit shapes to compute targeted viewpoints that observe as yet unobserved parts of the fruits. Furthermore, we formulate the concept of viewpoint dissimilarity to reduce the sampling space for more efficient selection of useful, dissimilar viewpoints. Our simulation experiments with a UR5e arm equipped with an RGB-D sensor provide a quantitative demonstration of the efficacy of our iterative next best view planning method based on shape completion. In comparative experiments with a state-of-the-art viewpoint planner, we demonstrate improvement not only in the estimation of the fruit sizes, but also in their reconstruction, while significantly reducing the planning time. Finally, we show the viability of our approach for mapping sweet pepper plants with a real robotic system in a commercial glasshouse.

Crop monitoring is crucial for maximizing agricultural productivity and efficiency. However, monitoring large and complex structures such as sweet pepper plants presents significant challenges, especially due to frequent occlusions of the fruits. Traditional next-best view planning can lead to unstructured and inefficient coverage of the crops. To address this, we propose a novel view motion planner that builds a graph network of viable view poses and trajectories between nearby poses, thereby considering robot motion constraints. The planner searches the graphs for view sequences with the highest accumulated information gain, allowing for efficient pepper plant monitoring while minimizing occlusions. The generated view poses aim at both sufficiently covering already detected and discovering new fruits. The graph and the corresponding best view pose sequence are computed with a limited horizon and are adaptively updated in fixed time intervals as the system gathers new information. We demonstrate the effectiveness of our approach through simulated and real-world experiments using a robotic arm equipped with an RGB-D camera and mounted on a trolley. As the experimental results show, our planner produces view pose sequences to systematically cover the crops and leads to increased fruit coverage when given a limited time in comparison to a state-of-the-art single next-best view planner.

Collision-free navigation while moving amongst static and dynamic obstacles with a limited sensor range is still a great challenge for modern mobile robots. Therefore, the ability to avoid collisions with obstacles in crowded, partially observable environments is one of the most important indicators to measure the navigation performance of a mobile robot. In this paper, we propose a novel deep reinforcement learning architecture that combines a spatial graph and attention rea-soning to tackle this problem. We take the relative positions and velocities of observed humans as nodes of the spatial graph and robot-human pairs as nodes of the attention graph to capture the spatial relations between the robot and the humans. In this way, our approach enhances the modeling of the relationship between the moving robot, static obstacles, and the people in the surrounding. As a result, our proposed navigation framework significantly outperforms state-of-the-art approaches [1], [2] in crowded scenarios when the robot has only a limited sensor range in terms of a reduced collision rate. Furthermore, we realize a seriously decreased training time by applying parallel Double Deep Q-Learning.

Automated agricultural applications, i.e., fruit picking require spatial information about crops and, especially, their fruits. In this paper, we present a novel deep reinforcement learning (DRL) approach to determine the next best view for automatic exploration of 3D environments with a robotic arm equipped with an RGB-D camera. We process the obtained images into an octree with labeled regions of interest (ROIs), i.e., fruits. We use this octree to generate 3D observation maps that serve as encoded input to the DRL network. We hereby do not only rely on known information about the environment, but explicitly also represent information about the unknown space to force exploration. Our network takes as input the encoded 3D observation map and the temporal sequence of camera view pose changes, and outputs the most promising camera movement direction. Our experimental results show an improved ROI targeted exploration performance resulting from our learned network in comparison to a state-of-the-art method.

Pushing objects through cluttered scenes is a challenging task, especially when the objects to be pushed have initially unknown dynamics and touching other entities has to be avoided to reduce the risk of damage. In this paper, we approach this problem by applying deep reinforcement learning to generate pushing actions for a robotic manipulator acting on a planar surface where objects have to be pushed to goal locations while avoiding other items in the same workspace. With the latent space learned from a depth image of the scene and other observations of the environment, such as contact information between the end effector and the object as well as distance to the goal, our framework is able to learn contact-rich pushing actions that avoid collisions with other objects. As the experimental results with a six degrees of freedom robotic arm show, our system is able to successfully push objects from start to end positions while avoiding nearby objects. Furthermore, we evaluate our learned policy in comparison to a state-of-the-art pushing controller for mobile robots and show that our agent performs better in terms of success rate, collisions with other objects, and continuous object contact in various scenarios.

Autonomously driving vehicles must be able to navigate in dynamic and unpredictable environments in a collision-free manner. So far, this has only been partially achieved in driverless cars and warehouse installations where marked structures such as roads, lanes, and traffic signs simplify the motion planning and collision avoidance problem. We are presenting a new control approach for car-like vehicles that is based on an unprecedentedly fast-paced A* implementation that allows the control cycle to run at a frequency of 30 Hz. This frequency enables us to place our A* algorithm as a low-level replanning controller that is well suited for navigation and collision avoidance in virtually any dynamic environment. Due to an efficient heuristic consisting of rotate-translate-rotate motions laid out along the shortest path to the target, our Short-Term Aborting A* (STAA*) converges fast and can be aborted early in order to guarantee a high and steady control rate. While our STAA* expands states along the shortest path, it takes care of collision checking with the environment including predicted states of moving obstacles, and returns the best solution found when the computation time runs out. Despite the bounded computation time, our STAA* does not get trapped in corners due to the following of the shortest path. In simulated and real-robot experiments, we demonstrate that our control approach eliminates collisions almost entirely and is superior to an improved version of the Dynamic Window Approach with predictive collision avoidance capabilities [1].

For the most comfortable, human-aware robot navigation, subjective user preferences need to be taken into account. This paper presents a novel reinforcement learning framework to train a personalized navigation controller along with an intuitive virtual reality demonstration interface. The conducted user study provides evidence that our personalized approach significantly outperforms classical approaches with more comfortable human-robot experiences. We achieve these results using only a few demonstration trajectories from non-expert users, who predominantly appreciate the intuitive demonstration setup. As we show in the experiments, the learned controller generalizes well to states not covered in the demonstration data, while still reflecting user preferences during navigation. Finally, we transfer the navigation controller without loss in performance to a real robot.

Footstep planning is the dominating approach when it comes to controlling the walk of a humanoid robot, even though a footstep plan is expensive to compute. The most prominent proposals typically spend up to a few seconds of computation time and output a sequence of up to 30 steps all the way to the goal. This way, footstep planning is applicable only in static environments where nothing changes after a plan has been computed. Since uncontrolled environments present challenges such as unforeseen motion of other objects and unexpected disturbances to balance, fast replanning of a footstep plan while the robot is in motion is highly desirable. We present a new way of fast footstep planning - Aborting A* - which is able to guarantee a replanning rate of 50 Hz by aborting an A* search before completion. We make aborting possible by using a novel, obstacle-aware heuristic function that lays out rotate-translate-rotate motions along the shortest path to the goal, enabling us to stop the planning progress prematurely with a target-oriented solution at any time during the search, even after only a few nodes have been expanded. We show in our experiments that despite the bounded computation time, our planner computes good results and does not get stuck in local minima.

The segmentation of a point cloud into planar primitives is a popular approach to first-line scene interpretation and is particularly useful in mobile robotics for the extraction of drivable or walkable surfaces and for tabletop segmentation for manipulation purposes. Unfortunately, the planar segmentation task becomes particularly challenging when the point clouds are obtained from an inherently noisy, robot-mounted sensor that is often in motion, therefor requiring real time processing capabilities. We present a real time-capable plane segmentation technique based on a region growing algorithm that exploits the organized structure of point clouds obtained from RGB-D sensors. In order to counteract the sensor noise, we invest into careful selection of seeds that start the region growing and avoid the computation of surface normals whenever possible. We implemented our algorithm in C++ and thoroughly tested it in both simulated and real-world environments where we are able to compare our approach against existing state-of-the-art methods implemented in the Point Cloud Library. The experiments presented here suggest that our approach is accurate and fast, even in the presence of considerable sensor noise.

We present PATHoBot an autonomous crop surveying and intervention robot for glasshouse environments. The aim of this platform is to autonomously gather high quality data and also estimate key phenotypic parameters. To achieve this we retro-fit an off-the-shelf pipe-rail trolley with an array of multi-modal cameras, navigation sensors and a robotic arm for close surveying tasks and intervention. In this paper we describe PATHoBot design choices made to ensure proper operation in a commercial glasshouse environment. As a surveying platform we collect a number of datasets which include both sweet pepper and tomatoes. We show how PATHoBot enables novel surveillance approaches by first improving our previous work on fruit counting by incorporating wheel odometry and depth information. We find that by introducing re-projection and depth information we are able to achieve an absolute improvement of 20 points over the baseline technique in an "in the wild" situation. Finally, we present a 3D mapping case study, further showcasing PATHoBot’s crop surveying capabilities.

The detection of planar surfaces in a point cloud is a popular technique for the extraction of drivable or walkable surfaces and for tabletop segmentation. Unfortunately, RGB-D sensors are quite noisy and provide incomplete data, which makes the extraction of surfaces more challenging. Also, it is desirable to process the point cloud data in real time, which at a rate of approximately 30 Hz, leaves only a small amount of computation time per frame. We have already developed a real time-capable plane segmentation method [1] that exploits the organized structure of RGB-D point clouds in order to implement a computationally efficient region growing algorithm. It uses the point-plane distance to assign points to their segments rather than inherently unreliable surface normals. Now we are presenting an improvement where we adapt thresholds and other parameters of our algorithm to the measured depth in order to account for an increasing scatter of the points at larger distances from the camera. We estimate a minimum detectable plane size in pixels dependent on the measured depth. This enables us to stride in pixel coordinates with larger steps that are adaptive to the measured depth and to implement more robust sanity checks of depth-dependent size. Apart from a speed-up of the runtime of our algorithm, the segmentation quality also increased. We show a comparison between our improvement, our previous version, and other state-of-the-art methods evaluated on multiple commonly available datasets.

Modern agricultural applications require knowledge about the position and size of fruits on plants. However, occlusions from leaves typically make obtaining this information difficult. We present a novel viewpoint planning approach that builds up an octree of plants with labeled regions of interest (ROIs), i.e., fruits. Our method uses this octree to sample viewpoint candidates that increase the information around the fruit regions and evaluates them using a heuristic utility function that takes into account the expected information gain. Our system automatically switches between ROI targeted sampling and exploration sampling, which considers general frontier voxels, depending on the estimated utility. When the plants have been sufficiently covered with the RGB-D sensor, our system clusters the ROI voxels and estimates the position and size of the detected fruits. We evaluated our approach in simulated scenarios and compared the resulting fruit estimations with the ground truth. The results demonstrate that our combined approach outperforms a sampling method that does not explicitly consider the ROIs to generate viewpoints in terms of the number of discovered ROI cells. Furthermore, we show the real-world applicability by testing our framework on a robotic arm equipped with an RGB-D camera installed on an automated pipe-rail trolley in a capsicum glasshouse.

In this paper, we present a new approach to human-aware robot navigation, which extends our previous proximity-based navigation framework [1] by introducing visibility and predictability as new parameters. We derived these parameters from a user study and incorporated them into a cost function, which models the user’s discomfort with respect to a relative robot position based on proximity, visibility, predictability, and work efficiency. We use this cost function in combination with an A* planner to create a user-preferred robot navigation policy. In comparison to our previous framework, our new cost function results in a 6% increase in social distance compliance, a 6.3% decrease in visibility of the robot as preferred, and an average decrease of orientation changes of 12.6° per meter resulting in better predictability, while maintaining a comparable average path length. We further performed a virtual reality experiment to evaluate the user comfort based on direct human feedback, finding that the participants on average felt comfortable to very comfortable with the resulting robot trajectories from our approach.

Annotated datasets are essential for supervised learning. However, annotating large datasets is a tedious and time-intensive task. This paper addresses active learning in the context of semantic segmentation with the goal of reducing the human labeling effort. Our application is agricultural robotics and we focus on the task of distinguishing between crop and weed plants from image data. A key challenge in this application is the transfer of an existing semantic segmentation CNN to a new field, in which growth stage, weeds, soil, and weather conditions differ. We propose a novel approach that, given a trained model on one field together with rough foreground segmentation, refines the network on a substantially different field providing an effective method of selecting samples to annotate for supporting the transfer. We evaluated our approach on two challenging datasets from the agricultural robotics domain and show that we achieve a higher accuracy with a smaller number of samples compared to random sampling as well as entropy based sampling, which consequently reduces the required human labeling effort.

Foresighted, human-aware navigation is a prerequisite for service robots acting in indoor environments. In this paper, we present a novel human-aware navigation approach that relies on long-term prediction of human movements. In particular, we consider the problem of finding a path from the robot's current position to the initially unknown navigation goal of a moving user to provide timely assistance there. The navigation strategy has to minimize the robot's arrival time and at the same time comply with the user's comfort during the movement. Our solution predicts the user's navigation goal based on the robot's observations and prior knowledge about typical human transitions between objects. Based on the motion prediction, we then compute a time-dependent cost map that encodes the belief about the user's positions at future time steps. Using this map, we solve the time-dependent shortest path problem to find an efficient path for the robot, which still abides by the rules of human comfort. To identify robot navigation actions that are perceived as uncomfortable by humans, we performed user surveys and defined the corresponding constraints. We thoroughly evaluated our navigation system in simulation as well as in real-world experiments. As the results show, our system outperforms existing approaches in terms of human comfort, while still minimizing arrival times of the robot.

Geometric primitives are a compact and versatile representation of the environment and the objects within. From a motion planning perspective, the geometric structure can be leveraged in order to implement potentially faster and smoother motion control algorithms than it has been possible with grid-based occupancy maps so far. In this paper, we introduce a novel perception pipeline that efficiently processes the point cloud obtained from an RGB-D sensor in order to produce a floor-projected 2D map in the field-of-view of the robot where obstacles are represented as polygons rather than cells. These polygons can then be processed by path planning algorithms and obstacle avoidance controllers. Our pipeline includes a ground floor plane detector that performs significantly faster than other contemporary solutions and a grid segmentation algorithm that uses image processing techniques to identify the contours of obstacles in order to convert them to polygons. We demonstrate the performance of our approach in experiments with a wheeled and a humanoid robot and show that our polygonal perception pipeline works robustly even in the presence of the disturbances caused by the shaking of a walking robot.

As service robots are entering more and more homes it gets evermore important to find behavior strategies that ensure a harmonic coexistence between those systems and their users. In this paper, we present a novel approach to enable a mobile robot to provide timely assistance to a user moving in its environment, while simultaneously avoiding unnecessary movements as well as interferences with the user. We developed a framework that uses information about the last object interaction to predict possible future movement destinations of the user and infer where they might need assistance based on prior knowledge. Given this prediction, the robot chooses the best position for itself that minimizes the time until assistance can be provided as well as avoids interferences with other activities of the user. We evaluated our approach in comparison to state-of-the-art methods in simulated environments and performed a user study in a virtual reality environment. Our evaluation demonstrates that our approach is able to decrease both the time until assistance is provided and the travel distance of the robot as well as increases the average distance between the user and the robot in comparison to state-of-the-art systems. Additionally, the robot behavior generated by our method is rated as more pleasant by our study participants than comparable literature approaches.

Perception of the local environment is a precondition for mobile robots to navigate safely in dynamic environments. Most robots, i.e., humanoids and smaller wheeled robots rely on planar regions. For humanoids, a simple 2D occupancy map as environment representation on which a path is planned is hereby not sufficient since they can step over and onto objects and therefore need height information. Considering dynamic obstacles introduces another level of complexity, since they can lead to necessary replanning or collisions at later stages. In this paper, we present a framework that first extracts planar regions in height maps and detects dynamic obstacles. Our system then uses this information to create a set of prediction maps, in which paths can be efficiently planned in real time at low CPU cost. We show in simulation and real-world experiments that our framework keeps run times well under 10ms for one computation cycle and allows for foresighted real-time 3D footstep planning.

Foresighted navigation is an essential skill for robots to rise from rigid factory floor installations to much more versatile mobile robots that partake in our everyday environment. The current state of the art that provides this mobility to some extent is the Dynamic Window Approach combined with a global start-to-target path planner. However, neither the Dynamic Window Approach nor the path planner are equipped to predict the motion of other objects in the environment. We propose a change in the Dynamic Window Approach-a dynamic collision model-that is capable of predicting future collisions with the environment by also taking into account the motion of other objects. We show in simulated experiments that our new way of computing the Dynamic Window Approach significantly reduces the number of collisions in a dynamic setting with nonholonomic vehicles while still being computationally efficient.

With the advent of polygonal maps finding their way into the navigational software of mobile robots, the Visibility Graph can be used to search for the shortest collision-free path. The nature of the Visibility Graph-based shortest path algorithms is such that first the entire graph is computed in a relatively time-consuming manner. Then, the graph can be searched efficiently any number of times for varying start and target state combinations with the A* or the Dijkstra algorithm. However, real-world environments are typically too dynamic for a map to remain valid for a long time. With the goal of obtaining the shortest path quickly in an ever changing environment, we introduce a rapid path finding algorithm-Minimal Construct-that discovers only a necessary portion of the Visibility Graph around the obstacles that actually get in the way. Collision tests are computed only for lines that seem heuristically promising. This way, shortest paths can be found much faster than with a state-of-the-art Visibility Graph algorithm and as our experiments show, even grid-based A* searches are outperformed in most cases with the added benefit of smoother and shorter paths.

Next-best-view algorithms are commonly used for covering known scenes, for example in search, maintenance, and mapping tasks. In this paper, we consider the problem of planning a strategy for covering articulated environments where the robot also has to manipulate objects to inspect obstructed areas. This problem is particularly challenging due to the many degrees of freedom resulting from the articulation. We propose to exploit graphics processing units present in many embedded devices to parallelize the computations of a greedy next-best-view approach. We implemented algorithms for costmap computation, path planning, as well as simulation and evaluation of viewpoint candidates in OpenGL for Embedded Systems and benchmarked the implementations on multiple device classes ranging from smartphones to multi-GPU servers. We introduce a heuristic for estimating a utility map from images rendered with strategically placed spherical cameras and show in simulation experiments that robots can successfully explore complex articulated scenes with our system.

In this paper, we present a visual simultaneous localization and mapping (SLAM) system for humanoid robots. We introduce a new binary descriptor called DLab that exploits the combined information of color, depth, and intensity to achieve robustness with respect to uniqueness, reproducibility, and stability. We use DLab within ORB-SLAM, where we replaced the place recognition module with a modification of FAB-MAP that works with newly built codebooks using our binary descriptor. In experiments carried out in simulation and with a real Nao humanoid equipped with an RGB-D camera, we show that DLab has a superior performance in comparison to other descriptors. The application to feature tracking and place recognition reveal that the new descriptor is able to reliably track features even in sequences with seriously blurred images and that it has a higher percentage of correctly identified similar images. As a result, our new visual SLAM system has a lower absolute trajectory error in comparison to ORB-SLAM and is able to accurately track the robot's trajectory.

In many situations, users walk on typical paths between specific destinations at which the service of a mobile robot is needed. Depending on the environment and the paths, step-by-step following of the human might not be the optimal solution since better paths for the robot exist. We propose to perform a prediction about the human's future movements and use this information in a reinforcement learning framework to generate foresighted navigation actions for the robot. Since frequent occlusions of the human will occur due to obstacles and the robot's constrained field of view, the estimate about the humans's position and the prediction of the next destination are affected by uncertainty. Our approach deals with such situations by explicitly considering occlusions in the reward function such that the robot automatically considers to execute actions to get the human in its field of view. We show in simulated and real-world experiments that our technique leads to significantly shorter paths compared to an approach in which the robot always tries to closely follow the user and, additionally, can handle occlusions.

We present a simple and efficient mathematical framework suitable for generating motion in the context of a variety of robotic motion tasks ranging from low-level motor control up to high-level locomotion planning. Our concept is based on a one-dimensional second-order model that allows analytic computation of its inverse dynamics while respecting physical constraints. This makes it a particularly useful tool for tasks that are expressed only as a start and goal state, such as animation key frames or way points in path planning. By means of time synchronization, the model extends easily to an arbitrary number of dimensions in a way that the target is reached in all dimensions at the same time. The framework excels in terms of execution time, which lies in the microsecond range even for high-dimensional trajectory generation tasks. We demonstrate our method in two different settings - full-body trajectory generation and path planning - and show its benefits in comparison with current state-of-the-art algorithms.

We present an approach to learn optimal navigation actions for assistance tasks in which the robot aims at efficiently reaching the final navigation goal of a human where service has to be provided. Always following the human at a close distance might hereby result in inefficient trajectories, since people regularly do not move on the shortest path to their destination (e.g., they move to grab the phone or make a note). Therefore, a service robot should infer the human's intended navigation goal and compute its own motion based on that prediction. We developed an approach that applies reinforcement learning to get a Q-function that determines for each pair of the robot's and human's relative positions the best navigation action for the robot. Our approach applies a prediction of the human's motion based on a softened Markov decision process (MDP). This MDP is independent from the navigation learning framework and is learned beforehand on previously observed trajectories. We thoroughly evaluated our method in simulation and on a real robot. As the experimental results show, our approach leads to foresighted navigation behavior and significantly reduces the path length and completion time compared to naive following strategies.

To this date, many footstep planning systems rely on external sensors for mapping and traversability analysis or on computationally expensive algorithms that do not allow for real-time calculations. In this paper, we present an approach that analyzes the environment in the vicinity of the robot with an onboard RGBD camera while computing local footstep plans in real time. We achieve this by combining the advantages of grid-based height maps, fast planar region segmentation, and a systematic local footstep search to a local goal point. Using a single CPU core, a full mapping and planning cycle only takes 18 ms on average and thus presents an important step to autonomous humanoid robots in dynamic environments that only rely on their onboard hardware.

Mobile manipulators installed in warehouses and factories for conveying goods between working stations need to meet the requirements of time-critical workflows. Moreover, the systems are expected to deal with changing tasks, cluttered environments and constraints imposed by the goods to be delivered. In this paper, we present a novel planning framework for generating asymptotically optimal paths for mobile manipulators subject to task constraints. Our approach introduces the Bidirectional Informed RRT* (BI2RRT*) that extends the Informed RRT* [1] towards bidirectional search and satisfaction of end-effector task constraints. In various experiments, we demonstrate the efficiency of BI2RRT* for both unconstrained and constrained mobile manipulation planning problems. As the results show, our planning framework finds better solutions than Informed RRT* and Bidirectional RRT* in less planning.

In this paper, we introduce an approach to efficient robot navigation through cluttered indoor environments. We propose to estimate local obstacle densities based on already detected objects and use them to predict traversal costs corresponding to potential obstacles in regions not yet observable by the robot's sensors. By taking into account the predicted costs for path planning, the robot is then able to navigate in a more foresighted manner and reduces the risk of getting stuck in cluttered regions. We thoroughly evaluated our approach in simulated and real-world experiments. As the experimental results demonstrate, our method enables the robot to efficiently navigate through environments containing cluttered regions and achieves significantly shorter completion times compared to a standard approach not using any prediction.

In grasping tasks carried out with humanoids, knowledge about the robot's reachable workspace is important. Without this knowledge, it might be necessary to repeatedly adapt the stance location and call an inverse kinematics solver before a valid robot configuration to reach a given grasping pose can be found. In this paper, we present an approach to select an optimal stance location in SE(2) for a humanoid robot's feet relative to a desired grasp pose. We use a precomputed representation of the robot's reachable workspace that stores quality information in addition to spatial data. By inverting this representation we obtain a so-called inverse reachability map (IRM) containing a collection of potential stance poses for the robot. The generated IRM can subsequently be used to select a statically stable, collision-free stance configuration to reach a given grasping target. We evaluated our approach with a Nao humanoid in simulation and in experiments with the real robot. As the experiments show, using our approach optimal stance poses can easily be obtained. Furthermore, the IRM leads to a substantially increased success rate of reaching grasping poses compared to other meaningful foot placements within the vicinity of the desired grasp.

In this paper, we present a novel approach to accurately calibrate the kinematic model of a humanoid based on observations of its monocular camera. Our technique estimates the parameters of the complete model, consisting of the joint angle offsets of the whole body including the legs, as well as the camera extrinsic and intrinsic parameters. We cast the parameter estimation as a least-squares optimization problem. In the error function, we consider the residuals between camera observations of end-effector markers and their projections into the image based on the estimate of the calibration parameters. Furthermore, we developed an approach to automatically select a subset of configurations for the calibration process that yields a good trade-off between the number of observations and accuracy. As the experiments with a Nao humanoid show, we achieve an accurate calibration for this low-cost platform. Further, our approach to configuration selection yields substantially better optimization results compared to randomly chosen viable configurations. Hence, our system only requires a reduced number of configurations to achieve accurate results. Our optimization is general and the implementation, which is available online, can easily be applied to different humanoids.

Although the neurological impairments of Parkinson's disease (PD) patients are well known to go along with motor control deficits, e.g., tremor, rigidity, and reduced movement, not much is known about the motor control parameters affected by the disease. In this paper, we therefore present a novel approach to human motions analysis using motor control strategies with joint weight parameterization. We record the motions of healthy subjects and PD patients performing a hand coordination task with the whole-body XSens MVN motion capture system. For our motion strategy analysis we then follow a two step approach. First, we perform a complexity reduction by mapping the recorded human motions to a simplified kinematic model of the upper body. Second, we reproduce the recorded motions using a Jacobian weighted damped least squares controller with adaptive joint weights. We developed a method to iteratively learn the joint weights of the controller with the mapped human joint trajectories as reference input. Finally, we use the learned joint weights for a quantitative comparison between the motion control strategies of healthy subjects and PD patients. Other than expected from clinical experience, we found that the joint weights are almost evenly distributed along the arm in the PD group. In contrast to that, the proximal joint weights of the healthy subjects are notably larger than the distal ones.

Controlling the robot with a permanently-updated optimal trajectory, also known as model predictive control, is the Holy Grail of whole-body motion generation. Before obtaining it, several challenges should be faced: computation cost, non-linear local minima, algorithm stability, etc. In this paper, we address the problem of applying the updated optimal control in real-time on the physical robot. In particular, we focus on the problems raised by the delays due to computation and by the differences between the real robot and the simulated model. Based on the optimal-control solver MuJoCo, we implemented a complete model-predictive controller and we applied it in real-time on the physical HRP-2 robot. It is the first time that such a whole-body model predictive controller is applied in real-time on a complex dynamic robot. Aside from the technical contributions cited above, the main contribution of this paper is to report the experimental results of this première implementation.

In this paper, we present a system that enables humanoid robots to imitate complex whole-body motions of humans in real time. In our approach, we use a compact human model and consider the positions of the end-effectors as well as the center of mass as the most important aspects to imitate. Our system actively balances the center of mass over the support polygon to avoid falls of the robot, which would occur when using direct imitation. For every point in time, our approach generates a statically stable pose. Hereby, we do not constrain the configurations to be in double support. Instead, we allow for changes of the support mode according to the motions to imitate. To achieve safe imitation, we use retargeting of the robot's feet if necessary and find statically stable configurations by inverse kinematics. We present experiments using human data captured with an Xsens MVN motion capture system. The results show that a Nao humanoid is able to reliably imitate complex whole-body motions in real time, which also include extended periods of time in single support mode, in which the robot has to balance on one foot.

Humanoid service robots performing complex object manipulation tasks need to plan whole-body motions that satisfy a variety of constraints: The robot must keep its balance, self-collisions and collisions with obstacles in the environment must be avoided and, if applicable, the trajectory of the end-effector must follow the constrained motion of a manipulated object in Cartesian space. These constraints and the high number of degrees of freedom make whole-body motion planning for humanoids a challenging problem. In this paper, we present an approach to whole-body motion planning with a focus on the manipulation of articulated objects such as doors and drawers. Our approach is based on rapidly-exploring random trees in combination with inverse kinematics and considers all required constraints during the search. Models of articulated objects hereby generate hand poses for sampled configurations along the trajectory of the object handle. We thoroughly evaluated our planning system and present experiments with a Nao humanoid opening a drawer, a door, and picking up an object. The experiments demonstrate the ability of our framework to generate solutions to complex planning problems and furthermore show that these plans can be reliably executed even on a low-cost humanoid platform.

In this paper, we present an integrated navigation system that allows humanoid robots to autonomously navigate in unknown, cluttered environments. From the data of an onboard consumer-grade depth camera, our system estimates the robot's pose to compensate for drift of odometry and maintains a heightmap representation of the environment. Based on this model, our system iteratively computes sequences of safe actions including footsteps and whole-body motions, leading the robot to target locations. Hereby, the planner chooses from a set of actions that consists of planar footsteps, step-over actions, as well as parameterized step-onto and step-down actions. To efficiently check for collisions during planning, we developed a new approach that takes into account the shape of the robot and the obstacles. As we demonstrate in experiments with a Nao humanoid, our system leads to robust navigation in cluttered environments and the robot is able to traverse highly challenging passages.

We present a system that enables a humanoid robot to imitate complex whole-body motions of humans in real time. For recording the human motions, any sensor system capable of inferring the joint angle trajectories can be used. In our work, we capture the human data with an Xsens MVN motion capture system consisting of inertial sensors attached to the body. Our framework converts the human joint angles to the robot's joint angles in real time. Here, we use a mapping between the human's and the robot's joints to ensure feasibility of the motion. The focus of our system lies in ensuring static stability when the motions are executed which is a challenging task, depending on the complexity of the movements. To avoid falls of the robot that might occur when using direct imitation of the joint angle trajectories due to the different weight distribution, we developed an approach that actively balances the center of mass over the support polygon of the robot's feet. At every point in time, our approach ensures that the robot is in a statically stable configuration, i.e., that the ground projection of the center of mass lies within the convex hull of the foot contact points. To achieve this, we apply inverse kinematics given valid foot positions that satisfy the stability criterion and generate the corresponding leg joint angles. In more detail, our system first finds valid positions for the robot's feet by determining a target plane and its orientation, so that the feet can be placed planar and the robot's center of mass is over the support polygon. The new positions of the feet are chosen as the projection on the target plane. Afterwards, the corresponding leg joint angles are calculated via inverse kinematics. To determine whether the configuration is in the double support modus, and if not, which foot is the stance foot, we evaluate the position of the center of mass relative to the feet.

In this paper, we consider the problem of efficient path planning for humanoid robots by combining grid-based 2D planning with footstep planning. In this way, we exploit the advantages of both frameworks, namely fast planning on grids and the ability to find solutions in situations where grid-based planning fails. Our method computes a global solution by adaptively switching between fast grid-based planning in open spaces and footstep planning in the vicinity of obstacles. To decide which planning framework to use, our approach classifies the environment into regions of different complexity with respect to the traversability. Experiments carried out in a simulated office environment and with a Nao humanoid show that (i) our approach significantly reduces the planning time compared to pure footstep planning and (ii) the resulting plans are almost as good as globally computed optimal footstep paths.

Collision-free navigation in cluttered environments is essential for any mobile manipulation system. Traditional navigation systems have relied on a 2D grid map projected from a 3D representation for efficiency. This approach, however, prevents navigation close to objects in situations where projected 3D configurations are in collision within the 2D grid map even if actually no collision occurs in the 3D environment. Accordingly, when using such a 2D representation for planning paths of a mobile manipulation robot, the number of planning problems which can be solved is limited and suboptimal robot paths may result. We present a fast, integrated approach to solve path planning in 3D using a combination of an efficient octree-based representation of the 3D world and an anytime search-based motion planner. Our approach utilizes a combination of multi-layered 2D and 3D representations to improve planning speed, allowing the generation of almost real-time plans with bounded sub-optimality. We present extensive experimental results with the two-armed mobile manipulation robot PR2 carrying large objects in a highly cluttered environment. Using our approach, the robot is able to efficiently plan and execute trajectories while transporting objects, thereby often moving through demanding, narrow passageways.

In this work, we present methods that enable a humanoid robot to traverse ramps using only vision and inertial data for sensing. Our video illustrates the method and shows the results obtained with a Nao humanoid. Using the proposed approach, the robot is able to autonomously walk down a 2.10m long ramp at an inclination of 20°.

In this paper, we present an approach to traversability classification solely based on monocular images and odometry estimates. We iteratively estimate the ground plane by detecting and matching features. Since the features are only sparse in the images and do not lead to dense information about traversability, we present a technique to train appearance-based floor detectors. In this way, we achieve a dense classification of the image data. Our approach trains the classifiers online in a self-supervised fashion from the ground plane estimation. During robot navigation, the classifiers are automatically updated and applied to the image stream to decide which areas are traversable. From this information, the robot can compute a two-dimensional occupancy grid map of the environment and use it for planning collision-free paths. As we illustrate in thorough experiments with a real humanoid, the classification results of our approach are highly accurate and the resulting occupancy map enables the robot to reliably avoid obstacles during navigation. Our appearance-based classifiers can also be used to augment stereo or RGBD-data in close ranges where these sensors cannot provide any depth information.

In order to successfully climb challenging stair-cases that consist of many steps and contain difficult parts, humanoid robots need to accurately determine their pose. In this paper, we present an approach that fuses the robot's observations from a 2D laser scanner, a monocular camera, an inertial measurement unit, and joint encoders in order to localize the robot within a given 3D model of the environment. We develop an extension to standard Monte Carlo localization (MCL) that draws particles from an improved proposal distribution to obtain highly accurate pose estimates. Furthermore, we introduce a new observation model based on chamfer matching between edges in camera images and the environment model. We thoroughly evaluate our localization approach and compare it to previous techniques in real-world experiments with a Nao humanoid. The results show that our approach significantly improves the localization accuracy and leads to a considerably more robust robot behavior. Our improved proposal in combination with chamfer matching can be generally applied to improve a range-based pose estimate by a consistent matching of lines obtained from vision.

Humanoid robots possess the capability of step ping over or onto objects, which distinguishes them from wheeled robots. When planning paths for humanoids, one therefore should consider an intelligent placement of footsteps instead of choosing detours around obstacles. In this paper, we present an approach to optimal footstep planning for humanoid robots. Since changes in the environment may appear and a humanoid may deviate from its originally planned path due to imprecise motion execution or slippage on the ground, the robot might be forced to dynamically revise its plans. Thus, efficient methods for planning and replanning are needed to quickly adapt the footstep paths to new situations. We formulate the problem of footstep planning so that it can be solved with the incremental heuristic search method D* Lite and present our extensions, including continuous footstep locations and efficient collision checking for footsteps. In experiments in simulation and with a real Nao humanoid, we demonstrate the effectiveness of the footstep plans computed and revised by our method. Additionally, we evaluate different footstep sets and heuristics to identify the ones leading to the best performance in terms of path quality and planning time. Our D* Lite algorithm for footstep planning is available as open source implementation.

In this paper, we present an approach to obstacle detection for collision-free, efficient humanoid robot navigation based on monocular images and sparse laser range data. To detect arbitrary obstacles in the surroundings of the robot, we analyze 3D data points obtained from a 2D laser range finder installed in the robot's head. Relying only on this laser data, however, can be problematic. While walking, the floor close to the robot's feet is not observable by the laser sensor, which inherently increases the risk of collisions, especially in nonstatic scenes. Furthermore, it is time-consuming to frequently stop walking and tilting the head to obtain reliable information about close obstacles. We therefore present a technique to train obstacle detectors for images obtained from a monocular camera also located in the robot's head. The training is done online based on sparse laser data in a self-supervised fashion. Our approach projects the obstacles identified from the laser data into the camera image and learns a classifier that considers color and texture information. While the robot is walking, it then applies the learned classifiers to the images to decide which areas are traversable. As we illustrate in experiments with a real humanoid, our approach enables the robot to reliably avoid obstacles during navigation. Furthermore, the results show that our technique leads to significantly more efficient navigation compared to extracting obstacles solely based on 3D laser range data acquired while the robot is standing at certain intervals.

In this paper, we present an approach to enable a humanoid robot to autonomously climb up spiral staircases. This task is substantially more challenging than climbing straight stairs since careful repositioning is needed. Our system globally estimates the pose of the robot, which is subsequently refined by integrating visual observations. In this way, the robot can accurately determine its relative position with respect to the next step. We use a 3D model of the environment to project edges corresponding to stair contours into monocular camera images. By detecting edges in the images and associating them to projected model edges, the robot is able to accurately locate itself towards the stairs and to climb them. We present experiments carried out with a Nao humanoid equipped with a 2D laser range finder for global localization and a low-cost monocular camera for short-range sensing. As we show in the experiments, the robot reliably climbs up the steps of a spiral staircase.

Reliable and efficient navigation with a humanoid robot is a difficult task. First, the motion commands are executed rather inaccurately due to backlash in the joints or foot slippage. Second, the observations are typically highly affected by noise due to the shaking behavior of the robot. Thus, the localization performance is typically reduced while the robot moves and the uncertainty about its pose increases. As a result, the reliable and efficient execution of a navigation task cannot be ensured anymore since the robot's pose estimate might not correspond to the true location. In this paper, we present a reinforcement learning approach to select appropriate navigation actions for a humanoid robot equipped with a camera for localization. The robot learns to reach the destination reliably and as fast as possible, thereby choosing actions to account for motion drift and trading off velocity in terms of fast walking movements against accuracy in localization. We present extensive simulated and practical experiments with a humanoid robot and demonstrate that our learned policy significantly outperforms a hand-optimized navigation strategy.

In this paper, we present a localization method for humanoid robots navigating in arbitrary complex indoor environments using only onboard sensing. Reliable and accurate localization for humanoid robots operating in such environments is a challenging task. First, humanoids typically execute motion commands rather inaccurately and odometry can be estimated only very roughly. Second, the observations of the small and lightweight sensors of most humanoids are seriously affected by noise. Third, since most humanoids walk with a swaying motion and can freely move in the environment, e.g., they are not forced to walk on flat ground only, a 6D torso pose has to be estimated. We apply Monte Carlo localization to globally determine and track a humanoid's 6D pose in a 3D world model, which may contain multiple levels connected by staircases. To achieve a robust localization while walking and climbing stairs, we intergrate 2D laser range measurements as well as attitude data and information from the joint encoders. We present simulated as well as real-word experiments with our humanoid and thoroughly evaluate our approach. As the experiments illustrate, the robot is able to globally localize itself and accurately track its 6D pose over time.

A main difficulty that arises in the context of probabilistic localization is the design of an appropriate observation model, i.e., determining the likelihood of a sensor measurement given the pose of the robot and a map of the environment. Many successful approaches to localization rely on data provided by range sensors, e.g., laser range scanners. When using such data one normally has to deal with erroneous maximum-range readings that occur due to poor-reflecting surfaces. In general, these readings cannot be distinguished from readings obtained when no obstacle is within the measurement range of the sensor. Therefore, existing localization techniques treat these readings alike in the observation model. In this paper, we present a novel approach that explicitly considers the reflection properties of surfaces and thus the expectation of valid range measurements. In addition to the expected range measurement, we compute the probability of reflectance for a beam given the relative pose of the robot to the obstacle taking into account the angle of incidence of the beam. We estimate the reflection properties of surfaces using data collected with a mobile robot equipped with a laser range scanner. As we demonstrate in experiments carried out with a real robot, our technique leads to significantly improved localization results compared to a state-of-the-art observation model.

In this paper, we present an approach that allows a robot to observe, generalize, and reproduce tasks observed from multiple demonstrations. Motion capture data is recorded in which a human instructor manipulates a set of objects. In our approach, we learn relations between body parts of the demonstrator and objects in the scene. These relations result in a generalized task description. The problem of learning and reproducing human actions is formulated using a dynamic Bayesian network (DBN). The posteriors corresponding to the nodes of the DBN are estimated by observing objects in the scene and body parts of the demonstrator. To reproduce a task, we seek for the maximum-likelihood action sequence according to the DBN. We additionally show how further constraints can be incorporated online, for example, to robustly deal with unforeseen obstacles. Experiments carried out with a real 6-DoF robotic manipulator as well as in simulation show that our approach enables a robot to reproduce a task carried out by a human demonstrator. Our approach yields a high degree of generalization illustrated by performing a pick-and-place and a whiteboard cleaning task.

Motion blur is a severe problem in images grabbed by legged robots and, in particular, by small humanoid robots. Standard feature extraction and tracking approaches typically fail when applied to sequences of images strongly affected by motion blur. In this paper, we propose a new feature detection and tracking scheme that is robust even to non-uniform motion blur. Furthermore, we developed a framework for visual odometry based on features extracted out of and matched in monocular image sequences. To reliably extract and track the features, we estimate the point spread function (PSF) of the motion blur individually for image patches obtained via a clustering technique and only consider highly distinctive features during matching. We present experiments performed on standard datasets corrupted with motion blur and on images taken by a camera mounted on walking small humanoid robots to show the effectiveness of our approach. The experiments demonstrate that our technique is able to reliably extract and match features and that it is furthermore able to generate a correct visual odometry, even in presence of strong motion blur effects and without the aid of any inertial measurement sensor.

Cameras are popular sensors for robot navigation tasks such as localization as they are inexpensive, lightweight, and provide rich data. However, fast movements of a mobile robot typically reduce the performance of vision-based localization systems due to motion blur. In this paper, we present a reinforcement learning approach to choose appropriate velocity profiles for vision-based navigation. The learned policy minimizes the time to reach the destination and implicitly takes the impact of motion blur on observations into account. To reduce the size of the resulting policies, which is desirable in the context of memory-constrained systems, we compress the learned policy via a clustering approach. Extensive simulated and real-world experiments demonstrate that our learned policy significantly outperforms any policy that uses a constant velocity. We furthermore show, that our policy is applicable to different environments. Additional experiments demonstrate that our compressed policies do not result in a performance loss compared to the originally learned policy.

Wheeled tour guide robots have already been deployed in various museums or fairs worldwide. A key requirement for successful tour guide robots is to interact with people and to entertain them. Most of the previous tour guide robots, however, focused more on the involved navigation task than on natural interaction with humans. Humanoid robots, on the other hand, offer a great potential for investigating intuitive, multimodal interaction between humans and machines. In this paper, we present our mobile full-body humanoid tour guide robot Robotinho. We provide mechanical and electrical details and cover perception, the integration of multiple modalities for interaction, navigation control, and system integration aspects. The multimodal interaction capabilities of Robotinho have been designed and enhanced according to the questionnaires filled out by the people who interacted with the robot at previous public demonstrations. We present experiences we have made during experiments in which untrained users interacted with the robot.

Humanoids have recently become a popular research platform in the robotics community. Such robots offer various fields for new applications. However, they have several drawbacks compared to wheeled vehicles such as stability problems, limited payload capabilities, violation of the flat world assumption, and they typically provide only very rough odometry information, if at all. In this paper, we investigate the problem of learning accurate grid maps with humanoid robots. We present techniques to deal with some of the above-mentioned difficulties. We describe how an existing approach to the simultaneous localization and mapping (SLAM) problem can be adapted to robustly learn accurate maps with a humanoid equipped with a laser range finder. We present an experiment in which our mapping system builds a highly accurate map with a size of around 20 m by 20 m using data acquired with a humanoid in our office environment containing two loops. The resulting maps have a similar accuracy as maps built with a wheeled robot.

Due to their high number of joints, humanoid robots typically have kinematic redundancies to achieve end- effector poses. Examples for such redundancies are the kinematic chains of pitch and yaw joints that allow the robot to turn towards a gaze target. Our humanoid communication robot currently uses its spine, its neck, and its eye joints to direct its cameras towards an object. In this paper, we propose a control strategy that considers three factors, namely tracking error, discomfort, defined at the joint level, and "effort" to control the pitch and yaw joints. Our strategy is based on gradient descent on a cost function. During the optimization, we use different step sizes to reflect the different inertia of the moved parts. Our control scheme produces human-like motions, where smaller, light-weight parts such as the eyes of the robot move quickly towards the target and then move back while the larger joints turn towards the target. We present experiments to evaluate the proposed strategy qualitatively and quantitatively.

We present an algorithm for estimating the fundamental frequency in speech signals. Our approach incorporates models of voiced speech on three levels. First, we estimate the pitch for each time frame based on its harmonic structure using non-negative matrix factorization. The second level utilizes temporal pitch continuity to extract partial pitch contours. Thirdly, we incorporate statistics of the succession of voiced segments to aggregate partial contours to the final contour of an utterance. We evaluate our approach on the Keele database. The experimental results show the robustness of our method for noisy speech, and the good performance for clean speech in comparison with state-of-the-art algorithms.

In this paper, we present an algorithm for robustly estimating the fundamental frequency in speech signals. Our approach is based on pitch-scaled harmonic filtering (PSHF). Following PSHF, we perform a filtering in the frequency domain using the short-time Fourier transform in order to separate the harmonic and non-harmonic parts of the processed signal. We enhance the standard PSHF approach by using a range of window lengths and a cost function that is applied to each window size. This cost function takes into account the energy at the harmonic and non-harmonic frequency coefficients to estimate harmonic energy for a frame. By using energy peaks and applying a cost function that considers the change in pitch in subsequent frames, we then determine the final pitch contour. We evaluated our approach on the Keele database. As the experimental results demonstrate, our methods performs robustly for noisy speech and has a good performance for clean speech in comparison with state-of-the-art algorithms.

In this paper, we present the humanoid communication robot Fritz. Our robot communicates with people in an intuitive, multimodal way. Fritz uses speech, facial expressions, eye-gaze, and gestures to interact with people. Depending on the audio-visual input, our robot shifts its attention between different persons in order to involve them into the conversation. He performs human-like arm gestures during the conversation and also uses pointing gestures generated with eyes, head, and arms to direct the attention of its communication partners towards objects of interest. To express its emotional state, the robot generates facial expressions and adapts the speech synthesis. We discuss experiences made during two public demonstrations of our robot.

An essential capability for a robot designed to interact with humans is to show attention to the people in its surroundings. To enable a robot to involve multiple persons into interaction requires the maintenance of an accurate belief about the people in the environment. In this paper, we use a probabilistic technique to update the knowledge of the robot based on sensory input. In this way, the robot is able to reason about the uncertainty in its belief about people in the vicinity and is able to shift its attention between different persons. Even people who are not the primary conversational partners are included into the interaction. In practical experiments with a humanoid robot, we demonstrate the effectiveness of our approach.

As people move through their environments, they do not move randomly. Instead, they are often engaged in typical motion patterns, related to specific locations they might be interested in approaching. In this paper we propose a method for adapting the behavior of a mobile robot according to the activities of the people in its surrounding. Our approach uses learned models of people's motion behaviors. Whenever the robot detects a person it computes a probabilistic estimate about which motion pattern the person might be engaged in. During the path planning it then uses this belief to improve its navigation behavior. In different practical experiments carried out on a real robot we demonstrate that our approach allows a robot to quickly adapt its navigation plans according to the activities of the persons in its surrounding. We also present experiments illustrating that our approach provides a better behavior than a standard reactive collision avoidance system.

We propose a method for learning models of people's motion behaviors in an indoor environment. As people move through their environments, they do not move randomly. Instead, they often engage in typical motion patterns, related to specific locations that they might be interested in approaching and specific trajectories that they might follow in doing so. Knowledge about such patterns may enable a mobile robot to develop improved people following and obstacle avoidance skills. This paper proposes an algorithm that learns collections of typical trajectories that characterize a person's motion patterns. Data, recorded by mobile robots equipped with laser range finders, is clustered into different types of motion using the popular expectation maximization algorithm, while simultaneously learning multiple motion patterns. Experimental results, obtained using data collected in a domestic residence and in an office building, illustrate that highly predictive models of human motion patterns can be learned.

We propose a method for learning models of people's motion behaviors in indoor environments. As people move through their environments, they do not move randomly. Instead, they often engage in typical motion patterns, related to specific locations that they might be interested in approaching and specific trajectories that they might follow in doing so. Knowledge about such patterns may enable a mobile robot to develop improved people following and obstacle avoidance skills. This paper proposes an algorithm that learns collections of typical trajectories that characterize a person's motion patterns. Data, recorded by mobile robots equipped with laser-range finders, is clustered into different types of motion using the popular expectation maximization algorithm, while simultaneously learning multiple motion patterns. Experimental results, obtained using data collected in a domestic residence and in an office building, illustrate that highly predictive models of human motion patterns can be learned.

The coordination of robot motions is one of the fundamental problems for multi-robot systems. A popular approach to avoid planning in the high-dimensional composite configuration space is the prioritized and decoupled technique. In this paper we present a method for optimizing priority schemes for such prioritized and decoupled planning technique. Our approach performs a randomized search with hill-climbing to find solutions and to minimize the overall path lengths. The technique has been implemented and tested on real robots and in extensive simulation runs. The experimental results demonstrate that our method is able to greatly reduce the number of failures and to significantly reduce the overall path length for different prioritized and decoupled path planning techniques and even for large teams of robots.

Coordinating the motion of multiple mobile robots is one of the fundamental problems in robotics. The predominant algorithms for coordinating teams of robots are decoupled and prioritized, thereby avoiding combinatorially hard planning problems typically faced by centralized approaches. We present a method for finding solvable priority schemes for such prioritized and decoupled planning techniques. Existing approaches apply a single priority scheme which makes them overly prone to failure in cases where valid solutions exists. By searching in the space of priorization schemes, our approach overcomes this limitation. To focus the search, our algorithm is guided by constraints generated from the task specification. To illustrate the appropriateness of this approach, the paper discusses experimental results obtained with real robots and through systematic robot simulation. The experimental results demonstrate that our approach successfully solves many more coordination problems than previous decoupled and prioritized techniques.

This paper describes an interactive tour-guide robot, which was successfully exhibited in a Smithsonian museum. During its two weeks of operation, the robot interacted with thousands of people, traversing more than 44 km at speeds of up to 163 cm/sec. Our approach specifically addresses issues such as safe navigation in unmodified and dynamic environments, and short-term human-robot interaction. It uses learning pervasively at all levels of the software architecture.