Rewarding DINO: Predicting Dense Rewards with Vision Foundation Models

Well-designed dense reward functions in robot manipulation not only indicate whether a task is completed but also encode progress along the way. Generally, designing dense rewards is challenging and usually requires access to privileged state information available only in simulation, not in real-world experiments. This makes reward prediction models that infer task state information from camera images attractive. A common approach is to predict rewards from expert demonstrations based on visual similarity or sequential frame ordering. However, this biases the resulting reward function towards a specific solution and leaves it undefined in states not covered by the demonstrations. In this work, we introduce Rewarding DINO, a method for language-conditioned reward modeling that learns actual reward functions rather than specific trajectories. The model's compact size allows it to serve as a direct replacement for analytical reward functions with comparatively low computational overhead. We train our model on data sampled from 24 Meta-World+ tasks using a rank-based loss and evaluate pairwise accuracy, rank correlation, and calibration. Rewarding DINO achieves competitive performance in tasks from the training set and generalizes to new settings in simulation and the real world, indicating that it learns task semantics. We also test the model with off-the-shelf reinforcement learning algorithms to solve tasks from our Meta-World+ training set.

OGScene3D: Incremental Open-Vocabulary 3D Gaussian Scene Graph Mapping for Scene Understanding

Open-vocabulary scene understanding is crucial for robotic applications, enabling robots to comprehend complex 3D environmental contexts and supporting various downstream tasks such as navigation and manipulation. However, existing methods require pre-built complete 3D semantic maps to construct scene graphs for scene understanding, which limits their applicability in robotic scenarios where environments are explored incrementally. To address this challenge, we propose OGScene3D, an open-vocabulary scene understanding system that achieves accurate 3D semantic mapping and scene graph construction incrementally. Our system employs a confidence-based Gaussian semantic representation that jointly models semantic predictions and their reliability, enabling robust scene modeling. Building on this representation, we introduce a hierarchical 3D semantic optimization strategy that achieves semantic consistency through local correspondence establishment and global refinement, thereby constructing globally consistent semantic maps. Moreover, we design a long-term global optimization method that leverages temporal memory of historical observations to enhance semantic predictions. By integrating 2D-3D semantic consistency with Gaussian rendering contribution, this method continuously refines the semantic understanding of the entire scene.Furthermore, we develop a progressive graph construction approach that dynamically creates and updates both nodes and semantic relationships, allowing continuous updating of the 3D scene graphs. Extensive experiments on widely used datasets and real-world scenes demonstrate the effectiveness of our OGScene3D on open-vocabulary scene understanding.

Enabling Dynamic Tracking in Vision-Language-Action Models via Time-Discrete and Time-Continuous Velocity Feedforward

While vision-language-action (VLA) models have shown great promise for robot manipulation, their deployment on rigid industrial robots remains challenging due to the inherent trade-off between compliance and responsiveness. Standard Behavior Cloning (BC) approaches predict discrete poses at low frequencies, omitting the velocity and acceleration feedforward terms typically used by low-level compliant controllers. This requires to rely on high stiffness for accurate tracking, thereby sacrificing safe contact dynamics. In this paper, we demonstrate the importance of integrating velocity feedforward terms into VLA policies to resolve this trade-off. We propose two methods for extracting velocity targets from VLAs: a time-discrete finite-difference approximation that serves as a highly effective bridge for existing models, and a continuous Cubic B-Spline action space that natively yields $C^2$ continuous trajectories for high-frequency control. Crucially, both approaches are strictly model-agnostic and compatible with any standard action-chunking architecture, requiring modifications only to teleoperation, data processing, and the low-level controller. We fine-tune the $π_{0.5}$ model and evaluate both of our approaches on a demanding, contact-rich cube-in-hole task. Our results indicate that incorporating the velocity feedforward term via finite differences significantly improves task execution speed, while the continuous B-Spline approach maintains high overall success rates and provides a foundation for smoother higher-order derivatives without compromising compliance.

FlowTouch: View-Invariant Visuo-Tactile Prediction

Tactile sensation is essential for contact-rich manipulation tasks. It provides direct feedback on object geometry, surface properties, and interaction forces, enhancing perception and enabling fine-grained control. An inherent limitation of tactile sensors is that readings are available only when an object is touched. This precludes their use during planning and the initial execution phase of a task. Predicting tactile information from visual information can bridge this gap. A common approach is to learn a direct mapping from camera images to the output of vision-based tactile sensors. However, the resulting model will depend strongly on the specific setup and on how well the camera can capture the area where an object is touched. In this work, we introduce FlowTouch, a novel model for view-invariant visuo-tactile prediction. Our key idea is to use an object's local 3D mesh to encode rich information for predicting tactile patterns while abstracting away from scene-dependent details. FlowTouch integrates scene reconstruction and Flow Matching-based models for image generation. Our results show that FlowTouch is able to bridge the sim-to-real gap and generalize to new sensor instances. We further show that the resulting tactile images can be used for downstream grasp stability prediction. Our code, datasets and videos are available at https://flowtouch.github.io/

VLAgents: A Policy Server for Efficient VLA Inference

The rapid emergence of Vision-Language-Action models (VLAs) has a significant impact on robotics. However, their deployment remains complex due to the fragmented interfaces and the inherent communication latency in distributed setups. To address this, we introduce VLAgents, a modular policy server that abstracts VLA inferencing behind a unified Gymnasium-style protocol. Crucially, its communication layer transparently adapts to the context by supporting both zero-copy shared memory for high-speed simulation and compressed streaming for remote hardware. In this work, we present the architecture of VLAgents and validate it by integrating seven policies -- including OpenVLA and Pi Zero. In a benchmark with both local and remote communication, we further demonstrate how it outperforms the default policy servers provided by OpenVLA, OpenPi, and LeRobot. VLAgents is available at https://github.com/RobotControlStack/vlagents

Robot Control Stack: A Lean Ecosystem for Robot Learning at Scale

Vision-Language-Action models (VLAs) mark a major shift in robot learning. They replace specialized architectures and task-tailored components of expert policies with large-scale data collection and setup-specific fine-tuning. In this machine learning-focused workflow that is centered around models and scalable training, traditional robotics software frameworks become a bottleneck, while robot simulations offer only limited support for transitioning from and to real-world experiments. In this work, we close this gap by introducing Robot Control Stack (RCS), a lean ecosystem designed from the ground up to support research in robot learning with large-scale generalist policies. At its core, RCS features a modular and easily extensible layered architecture with a unified interface for simulated and physical robots, facilitating sim-to-real transfer. Despite its minimal footprint and dependencies, it offers a complete feature set, enabling both real-world experiments and large-scale training in simulation. Our contribution is twofold: First, we introduce the architecture of RCS and explain its design principles. Second, we evaluate its usability and performance along the development cycle of VLA and RL policies. Our experiments also provide an extensive evaluation of Octo, OpenVLA, and Pi Zero on multiple robots and shed light on how simulation data can improve real-world policy performance. Our code, datasets, weights, and videos are available at: https://robotcontrolstack.github.io/

MetricNet: Recovering Metric Scale in Generative Navigation Policies

Generative navigation policies have made rapid progress in improving end-to-end learned navigation. Despite their promising results, this paradigm has two structural problems. First, the sampled trajectories exist in an abstract, unscaled space without metric grounding. Second, the control strategy discards the full path, instead moving directly towards a single waypoint. This leads to short-sighted and unsafe actions, moving the robot towards obstacles that a complete and correctly scaled path would circumvent. To address these issues, we propose MetricNet, an effective add-on for generative navigation that predicts the metric distance between waypoints, grounding policy outputs in real-world coordinates. We evaluate our method in simulation with a new benchmarking framework and show that executing MetricNet-scaled waypoints significantly improves both navigation and exploration performance. Beyond simulation, we further validate our approach in real-world experiments. Finally, we propose MetricNav, which integrates MetricNet into a navigation policy to guide the robot away from obstacles while still moving towards the goal.

Multimodal Spatial Language Maps for Robot Navigation and Manipulation

Grounding language to a navigating agent's observations can leverage pretrained multimodal foundation models to match perceptions to object or event descriptions. However, previous approaches remain disconnected from environment mapping, lack the spatial precision of geometric maps, or neglect additional modality information beyond vision. To address this, we propose multimodal spatial language maps as a spatial map representation that fuses pretrained multimodal features with a 3D reconstruction of the environment. We build these maps autonomously using standard exploration. We present two instances of our maps, which are visual-language maps (VLMaps) and their extension to audio-visual-language maps (AVLMaps) obtained by adding audio information. When combined with large language models (LLMs), VLMaps can (i) translate natural language commands into open-vocabulary spatial goals (e.g., "in between the sofa and TV") directly localized in the map, and (ii) be shared across different robot embodiments to generate tailored obstacle maps on demand. Building upon the capabilities above, AVLMaps extend VLMaps by introducing a unified 3D spatial representation integrating audio, visual, and language cues through the fusion of features from pretrained multimodal foundation models. This enables robots to ground multimodal goal queries (e.g., text, images, or audio snippets) to spatial locations for navigation. Additionally, the incorporation of diverse sensory inputs significantly enhances goal disambiguation in ambiguous environments. Experiments in simulation and real-world settings demonstrate that our multimodal spatial language maps enable zero-shot spatial and multimodal goal navigation and improve recall by 50% in ambiguous scenarios. These capabilities extend to mobile robots and tabletop manipulators, supporting navigation and interaction guided by visual, audio, and spatial cues.

Augmented Reality for RObots (ARRO): Pointing Visuomotor Policies Towards Visual Robustness

Visuomotor policies trained on human expert demonstrations have recently shown strong performance across a wide range of robotic manipulation tasks. However, these policies remain highly sensitive to domain shifts stemming from background or robot embodiment changes, which limits their generalization capabilities. In this paper, we present ARRO, a novel calibration-free visual representation that leverages zero-shot open-vocabulary segmentation and object detection models to efficiently mask out task-irrelevant regions of the scene without requiring additional training. By filtering visual distractors and overlaying virtual guides during both training and inference, ARRO improves robustness to scene variations and reduces the need for additional data collection. We extensively evaluate ARRO with Diffusion Policy on several tabletop manipulation tasks in both simulation and real-world environments, and further demonstrate its compatibility and effectiveness with generalist robot policies, such as Octo and OpenVLA. Across all settings in our evaluation, ARRO yields consistent performance gains, allows for selective masking to choose between different objects, and shows robustness even to challenging segmentation conditions. Videos showcasing our results are available at: augmented-reality-for-robots.github.io

LUMOS: Language-Conditioned Imitation Learning with World Models

We introduce LUMOS, a language-conditioned multi-task imitation learning framework for robotics. LUMOS learns skills by practicing them over many long-horizon rollouts in the latent space of a learned world model and transfers these skills zero-shot to a real robot. By learning on-policy in the latent space of the learned world model, our algorithm mitigates policy-induced distribution shift which most offline imitation learning methods suffer from. LUMOS learns from unstructured play data with fewer than 1% hindsight language annotations but is steerable with language commands at test time. We achieve this coherent long-horizon performance by combining latent planning with both image- and language-based hindsight goal relabeling during training, and by optimizing an intrinsic reward defined in the latent space of the world model over multiple time steps, effectively reducing covariate shift. In experiments on the difficult long-horizon CALVIN benchmark, LUMOS outperforms prior learning-based methods with comparable approaches on chained multi-task evaluations. To the best of our knowledge, we are the first to learn a language-conditioned continuous visuomotor control for a real-world robot within an offline world model. Videos, dataset and code are available at http://lumos.cs.uni-freiburg.de.