arXiv subjects

Sven Behnke

Publications and source records attributed to Sven Behnke.

At least 37 records · Page 2Linked to original sources

arXiv2025

Leveraging Vision-Language Models for Open-Vocabulary Instance Segmentation and Tracking

Vision-language models (VLMs) excel in visual understanding but often lack reliable grounding capabilities and actionable inference rates. Integrating them with open-vocabulary object detection (OVD), instance segmentation, and tracking leverages their strengths while mitigating these drawbacks. We utilize VLM-generated structured descriptions to identify visible object instances, collect application-relevant attributes, and inform an open-vocabulary detector to extract corresponding bounding boxes that are passed to a video segmentation model providing segmentation masks and tracking. Once initialized, this model directly extracts segmentation masks, processing image streams in real time with minimal computational overhead. Tracks can be updated online as needed by generating new structured descriptions and detections. This combines the descriptive power of VLMs with the grounding capability of OVD and the pixel-level understanding and speed of video segmentation. Our evaluation across datasets and robotics platforms demonstrates the broad applicability of this approach, showcasing its ability to extract task-specific attributes from non-standard objects in dynamic environments. Code, data, videos, and benchmarks are available at https://vlm-gist.github.io

cs.CV↗

arXiv2025

Iterative Motion Compensation for Canonical 3D Reconstruction from UAV Plant Images Captured in Windy Conditions

3D phenotyping of plants plays a crucial role for understanding plant growth, yield prediction, and disease control. We present a pipeline capable of generating high-quality 3D reconstructions of individual agricultural plants. To acquire data, a small commercially available UAV captures images of a selected plant. Apart from placing ArUco markers, the entire image acquisition process is fully autonomous, controlled by a self-developed Android application running on the drone's controller. The reconstruction task is particularly challenging due to environmental wind and downwash of the UAV. Our proposed pipeline supports the integration of arbitrary state-of-the-art 3D reconstruction methods. To mitigate errors caused by leaf motion during image capture, we use an iterative method that gradually adjusts the input images through deformation. Motion is estimated using optical flow between the original input images and intermediate 3D reconstructions rendered from the corresponding viewpoints. This alignment gradually reduces scene motion, resulting in a canonical representation. After a few iterations, our pipeline improves the reconstruction of state-of-the-art methods and enables the extraction of high-resolution 3D meshes. We will publicly release the source code of our reconstruction pipeline. Additionally, we provide a dataset consisting of multiple plants from various crops, captured across different points in time.

cs.CV↗

arXiv2025

Integration of the TIAGo Robot into Isaac Sim with Mecanum Drive Modeling and Learned S-Curve Velocity Profiles

Efficient physics simulation has significantly accelerated research progress in robotics applications such as grasping and assembly. The advent of GPU-accelerated simulation frameworks like Isaac Sim has particularly empowered learning-based methods, enabling them to tackle increasingly complex tasks. The PAL Robotics TIAGo++ Omni is a versatile mobile manipulator equipped with a mecanum-wheeled base, allowing omnidirectional movement and a wide range of task capabilities. However, until now, no model of the robot has been available in Isaac Sim. In this paper, we introduce such a model, calibrated to approximate the behavior of the real robot, with a focus on its omnidirectional drive dynamics. We present two control models for the omnidirectional drive: a physically accurate model that replicates real-world wheel dynamics and a lightweight velocity-based model optimized for learning-based applications. With these models, we introduce a learning-based calibration approach to approximate the real robot's S-shaped velocity profile using minimal trajectory data recordings. This simulation should allow researchers to experiment with the robot and perform efficient learning-based control in diverse environments. We provide the integration publicly at https://github.com/AIS-Bonn/tiago_isaac.

cs.RO↗

arXiv2025

LIAM: Multimodal Transformer for Language Instructions, Images, Actions and Semantic Maps

The availability of large language models and open-vocabulary object perception methods enables more flexibility for domestic service robots. The large variability of domestic tasks can be addressed without implementing each task individually by providing the robot with a task description along with appropriate environment information. In this work, we propose LIAM - an end-to-end model that predicts action transcripts based on language, image, action, and map inputs. Language and image inputs are encoded with a CLIP backbone, for which we designed two pre-training tasks to fine-tune its weights and pre-align the latent spaces. We evaluate our method on the ALFRED dataset, a simulator-generated benchmark for domestic tasks. Our results demonstrate the importance of pre-aligning embedding spaces from different modalities and the efficacy of incorporating semantic maps.

cs.CV↗

arXiv2025

Anticipating Human Behavior for Safe Navigation and Efficient Collaborative Manipulation with Mobile Service Robots

The anticipation of human behavior is a crucial capability for robots to interact with humans safely and efficiently. We employ a smart edge sensor network to provide global observations, future predictions, and goal information to integrate anticipatory behavior for the control of a mobile manipulation robot. We present approaches to anticipate human behavior in the context of safe navigation and collaborative mobile manipulation. First, we anticipate human motion by employing projections of predicted human trajectories from smart edge sensor observations into the planning map of a mobile robot. Second, we anticipate human intentions in a collaborative furniture-carrying task to achieve a given room layout. Our experiments indicate that anticipating human behavior allows for safer navigation and more efficient collaboration. Finally, we showcase an integrated robotic system that anticipates human behavior while collaborating with an operator to achieve a target room layout, including the placement of tables and chairs.

cs.RO↗

arXiv2025

AGILOped: Agile Open-Source Humanoid Robot for Research

With academic and commercial interest for humanoid robots peaking, multiple platforms are being developed. Through a high level of customization, they showcase impressive performance. Most of these systems remain closed-source or have high acquisition and maintenance costs, however. In this work, we present AGILOped - an open-source humanoid robot that closes the gap between high performance and accessibility. Our robot is driven by off-the-shelf backdrivable actuators with high power density and uses standard electronic components. With a height of 110 cm and weighing only 14.5 kg, AGILOped can be operated without a gantry by a single person. Experiments in walking, jumping, impact mitigation and getting-up demonstrate its viability for use in research.

cs.RO↗

arXiv2025

Interactive Shaping of Granular Media Using Reinforcement Learning

Autonomous manipulation of granular media, such as sand, is crucial for applications in construction, excavation, and additive manufacturing. However, shaping granular materials presents unique challenges due to their high-dimensional configuration space and complex dynamics, where traditional rule-based approaches struggle without extensive engineering efforts. Reinforcement learning (RL) offers a promising alternative by enabling agents to learn adaptive manipulation strategies through trial and error. In this work, we present an RL framework that enables a robotic arm with a cubic end-effector and a stereo camera to shape granular media into desired target structures. We show the importance of compact observations and concise reward formulations for the large configuration space, validating our design choices with an ablation study. Our results demonstrate the effectiveness of the proposed approach for the training of visual policies that manipulate granular media including their real-world deployment, significantly outperforming two baseline approaches in terms of target shape accuracy.

cs.RO↗

arXiv2025

SWA-SOP: Spatially-aware Window Attention for Semantic Occupancy Prediction in Autonomous Driving

Perception systems in autonomous driving rely on sensors such as LiDAR and cameras to perceive the 3D environment. However, due to occlusions and data sparsity, these sensors often fail to capture complete information. Semantic Occupancy Prediction (SOP) addresses this challenge by inferring both occupancy and semantics of unobserved regions. Existing transformer-based SOP methods lack explicit modeling of spatial structure in attention computation, resulting in limited geometric awareness and poor performance in sparse or occluded areas. To this end, we propose Spatially-aware Window Attention (SWA), a novel mechanism that incorporates local spatial context into attention. SWA significantly improves scene completion and achieves state-of-the-art results on LiDAR-based SOP benchmarks. We further validate its generality by integrating SWA into a camera-based SOP pipeline, where it also yields consistent gains across modalities.

cs.CV↗

arXiv2025

OC-SOP: Enhancing Vision-Based 3D Semantic Occupancy Prediction by Object-Centric Awareness

Autonomous driving perception faces significant challenges due to occlusions and incomplete scene data in the environment. To overcome these issues, the task of semantic occupancy prediction (SOP) is proposed, which aims to jointly infer both the geometry and semantic labels of a scene from images. However, conventional camera-based methods typically treat all categories equally and primarily rely on local features, leading to suboptimal predictions, especially for dynamic foreground objects. To address this, we propose Object-Centric SOP (OC-SOP), a framework that integrates high-level object-centric cues extracted via a detection branch into the semantic occupancy prediction pipeline. This object-centric integration significantly enhances the prediction accuracy for foreground objects and achieves state-of-the-art performance among all categories on SemanticKITTI.

cs.CV↗

arXiv2025

DiffSSC: Semantic LiDAR Scan Completion using Denoising Diffusion Probabilistic Models

Perception systems play a crucial role in autonomous driving, incorporating multiple sensors and corresponding computer vision algorithms. 3D LiDAR sensors are widely used to capture sparse point clouds of the vehicle's surroundings. However, such systems struggle to perceive occluded areas and gaps in the scene due to the sparsity of these point clouds and their lack of semantics. To address these challenges, Semantic Scene Completion (SSC) jointly predicts unobserved geometry and semantics in the scene given raw LiDAR measurements, aiming for a more complete scene representation. Building on promising results of diffusion models in image generation and super-resolution tasks, we propose their extension to SSC by implementing the noising and denoising diffusion processes in the point and semantic spaces individually. To control the generation, we employ semantic LiDAR point clouds as conditional input and design local and global regularization losses to stabilize the denoising process. We evaluate our approach on autonomous driving datasets, and it achieves state-of-the-art performance for SSC, surpassing most existing methods.

cs.CV↗

arXiv2025

Learning from SAM: Harnessing a Foundation Model for Sim2Real Adaptation by Regularization

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

cs.CV↗

arXiv2025

VideoPCDNet: Video Parsing and Prediction with Phase Correlation Networks

Understanding and predicting video content is essential for planning and reasoning in dynamic environments. Despite advancements, unsupervised learning of object representations and dynamics remains challenging. We present VideoPCDNet, an unsupervised framework for object-centric video decomposition and prediction. Our model uses frequency-domain phase correlation techniques to recursively parse videos into object components, which are represented as transformed versions of learned object prototypes, enabling accurate and interpretable tracking. By explicitly modeling object motion through a combination of frequency domain operations and lightweight learned modules, VideoPCDNet enables accurate unsupervised object tracking and prediction of future video frames. In our experiments, we demonstrate that VideoPCDNet outperforms multiple object-centric baseline models for unsupervised tracking and prediction on several synthetic datasets, while learning interpretable object and motion representations.

cs.CV↗

arXiv2025

PlaySlot: Learning Inverse Latent Dynamics for Controllable Object-Centric Video Prediction and Planning

Predicting future scene representations is a crucial task for enabling robots to understand and interact with the environment. However, most existing methods rely on videos and simulations with precise action annotations, limiting their ability to leverage the large amount of available unlabeled video data. To address this challenge, we propose PlaySlot, an object-centric video prediction model that infers object representations and latent actions from unlabeled video sequences. It then uses these representations to forecast future object states and video frames. PlaySlot allows the generation of multiple possible futures conditioned on latent actions, which can be inferred from video dynamics, provided by a user, or generated by a learned action policy, thus enabling versatile and interpretable world modeling. Our results show that PlaySlot outperforms both stochastic and object-centric baselines for video prediction across different environments. Furthermore, we show that our inferred latent actions can be used to learn robot behaviors sample-efficiently from unlabeled video demonstrations. Videos and code are available on https://play-slot.github.io/PlaySlot/.

cs.CV↗

arXiv2025

Prompt-responsive Object Retrieval with Memory-augmented Student-Teacher Learning

Building models responsive to input prompts represents a transformative shift in machine learning. This paradigm holds significant potential for robotics problems, such as targeted manipulation amidst clutter. In this work, we present a novel approach to combine promptable foundation models with reinforcement learning (RL), enabling robots to perform dexterous manipulation tasks in a prompt-responsive manner. Existing methods struggle to link high-level commands with fine-grained dexterous control. We address this gap with a memory-augmented student-teacher learning framework. We use the Segment-Anything 2 (SAM 2) model as a perception backbone to infer an object of interest from user prompts. While detections are imperfect, their temporal sequence provides rich information for implicit state estimation by memory-augmented models. Our approach successfully learns prompt-responsive policies, demonstrated in picking objects from cluttered scenes. Videos and code are available at https://memory-student-teacher.github.io

cs.RO↗

arXiv2025

Feature-Preserving Mesh Decimation for Normal Integration

Normal integration reconstructs 3D surfaces from normal maps obtained e.g. by photometric stereo. These normal maps capture surface details down to the pixel level but require large computational resources for integration at high resolutions. In this work, we replace the dense pixel grid with a sparse anisotropic triangle mesh prior to normal integration. We adapt the triangle mesh to the local geometry in the case of complex surface structures and remove oversampling from flat featureless regions. For high-resolution images, the resulting compression reduces normal integration runtimes from hours to minutes while maintaining high surface accuracy. Our main contribution is the derivation of the well-known quadric error measure from mesh decimation for screen space applications and its combination with optimal Delaunay triangulation.

cs.CV↗

arXiv2025

Dexterous Pre-grasp Manipulation for Human-like Functional Categorical Grasping: Deep Reinforcement Learning and Grasp Representations

Many objects, such as tools and household items, can be used only if grasped in a very specific way - grasped functionally. Often, a direct functional grasp is not possible, though. We propose a method for learning a dexterous pre-grasp manipulation policy to achieve human-like functional grasps using deep reinforcement learning. We introduce a dense multi-component reward function that enables learning a single policy, capable of dexterous pre-grasp manipulation of novel instances of several known object categories with an anthropomorphic hand. The policy is learned purely by means of reinforcement learning from scratch, without any expert demonstrations. It implicitly learns to reposition and reorient objects of complex shapes to achieve given functional grasps. In addition, we explore two different ways to represent a desired grasp: explicit and more abstract, constraint-based. We show that our method consistently learns to successfully manipulate and achieve desired grasps on previously unseen object instances of known categories using both grasp representations. Training is completed on a single GPU in under three hours.

cs.RO↗

arXiv2025

SOLD: Slot Object-Centric Latent Dynamics Models for Relational Manipulation Learning from Pixels

Learning a latent dynamics model provides a task-agnostic representation of an agent's understanding of its environment. Leveraging this knowledge for model-based reinforcement learning (RL) holds the potential to improve sample efficiency over model-free methods by learning from imagined rollouts. Furthermore, because the latent space serves as input to behavior models, the informative representations learned by the world model facilitate efficient learning of desired skills. Most existing methods rely on holistic representations of the environment's state. In contrast, humans reason about objects and their interactions, predicting how actions will affect specific parts of their surroundings. Inspired by this, we propose Slot-Attention for Object-centric Latent Dynamics (SOLD), a novel model-based RL algorithm that learns object-centric dynamics models in an unsupervised manner from pixel inputs. We demonstrate that the structured latent space not only improves model interpretability but also provides a valuable input space for behavior models to reason over. Our results show that SOLD outperforms DreamerV3 and TD-MPC2 - state-of-the-art model-based RL algorithms - across a range of benchmark robotic environments that require relational reasoning and manipulation capabilities. Videos are available at https://slot-latent-dynamics.github.io/.

cs.LG↗

arXiv2025

Quadrupedal Footstep Planning using Learned Motion Models of a Black-Box Controller

Legged robots are increasingly entering new domains and applications, including search and rescue, inspection, and logistics. However, for such systems to be valuable in real-world scenarios, they must be able to autonomously and robustly navigate irregular terrains. In many cases, robots that are sold on the market do not provide such abilities, being able to perform only blind locomotion. Furthermore, their controller cannot be easily modified by the end-user, requiring a new and time-consuming control synthesis. In this work, we present a fast local motion planning pipeline that extends the capabilities of a black-box walking controller that is only able to track high-level reference velocities. More precisely, we learn a set of motion models for such a controller that maps high-level velocity commands to Center of Mass (CoM) and footstep motions. We then integrate these models with a variant of the A star algorithm to plan the CoM trajectory, footstep sequences, and corresponding high-level velocity commands based on visual information, allowing the quadruped to safely traverse irregular terrains at demand.

cs.RO↗