Oliver Brock

Eduardo Sebastián, Adrian Pfisterer, Vito Mengers et al.

Robot learning must produce policies that generalize to new combinations of constraints, teammates, and environments. To achieve this, we must structurally factor the policy, which is a choice that dictates what generalizes, what requires retraining, and what remains entangled. Existing methods span a wide spectrum, from expecting structure to emerge from data scaling, to hand-designing it via hierarchies, skill libraries or learned specializations. In this paper, we study what we argue is the most fundamental factorization in robotics: separating the world from the task. We investigate the conditions under which this factorization is principled. World factors are properties of the embodied system and the environment; they exist independently of intent. Task factors are defined by the task's logic over what the world admits. We formalize this asymmetry through Bayesian model evidence: it aligns with the data-generating process, maintains high likelihood through an analytical world model, and reduces the Occam razor's penalty on task parameters. We instantiate this factorization by pairing AICON, a differentiable graph of recursive estimators and interconnections that is compositional, operates without task-specific data, and propagates cost gradients to actuators, with a compact, learned policy that modulates gradient paths. Gradients serve as the interface between the two factors: they carry world structure through the graph and task structure through costs, enabling low-dimensional learning while preserving structural generalization. We test the world/task factorization across three problems that encompass heterogeneous robots, environments, task logic and sensorimotor modalities. Our framework outperforms end-to-end baselines and analytical heuristics in all settings, generalizes zero-shot to out-of-distribution configurations, and transfers to real hardware without retraining.

Aravind Battaje, Malte Bernhard, Vito Mengers et al.

Generalization in robotics requires prior knowledge about how the world is structured, yet this structure changes from one situation to the next. This paper investigates the proposition that generalization arises from adaptively composing regularities -- predictable relationships within the robot-environment system -- into situation-appropriate structures for behavior generation. We examine this proposition by analyzing the mechanism in AICON (Active InterCONnect), a framework representing regularities as interacting processes in a differentiable network, where sensory feedback realizes composition and gradient descent generates behavior. To isolate adaptive composition as the key mechanism, we study a simple simulated problem in which all relevant regularities can be identified. We expose the resulting model to a wide range of novel conditions not considered during design, and we find that it generates context-appropriate behavior in all but one case, where encoded regularities are provably insufficient. Ablations reveal that the network automatically modulates which regularities influence behavior based on their informativeness. These results suggest that adaptive composition of regularities constitutes a powerful inductive bias for building generalization into behavior generation.

Aravind Battaje, Oliver Brock

A gaze-fixating robot perceives distance to the fixated object and relative positions of surrounding objects immediately, accurately, and robustly. We show how fixation, which is the act of looking at one object while moving, exploits regularities in the geometry of 3D space to obtain this information. These regularities introduce rotation-translation couplings that are not commonly used in structure from motion. To validate, we use a Franka Emika Robot with an RGB camera. We a) find that error in distance estimate is less than 5 mm at a distance of 15 cm, and b) show how relative position can be used to find obstacles under challenging scenarios. We combine accurate distance estimates and obstacle information into a reactive robot behavior that is able to pick up objects of unknown size, while impeded by unforeseen obstacles. Project page: https://oxidification.com/p/one-object-at-a-time/ .

Manuel Baum, Oliver Brock

Robotic manipulation behavior should be robust to disturbances that violate high-level task-structure. Such robustness can be achieved by constantly monitoring the environment to observe the discrete high-level state of the task. This is possible because different phases of a task are characterized by different sensor patterns and by monitoring these patterns a robot can decide which controllers to execute in the moment. This relaxes assumptions about the temporal sequence of those controllers and makes behavior robust to unforeseen disturbances. We implement this idea as probabilistic filter over discrete states where each state is direcly associated with a controller. Based on this framework we present a robotic system that is able to open a drawer and grasp tennis balls from it in a surprisingly robust way.

Vito Mengers, Oliver Brock

Reactive control is often considered insufficient for multi-objective tasks because conflicting objectives give rise to local minima. We argue this limitation is not inherent but arises from static encodings that fail to reflect how objectives currently interact. We exploit the interaction structure encoded in a graph-based world model by extending it with nullspace projections: conflicts are resolved where they arise by projecting lower-priority gradients into the nullspace of higher-priority ones, with priorities determined continuously from the current state. We demonstrate this in two domains where conflicts between objectives are central: navigation around non-convex obstacles, where static potential fields fundamentally fail, and planar pushing of non-convex objects, where our method achieves $100\%$ success across one-hundred configurations versus $0\%$ for the steepest-descent baseline and ${\sim}55\%$ for diffusion policy, without demonstrations or retraining. The same formulation transfers directly to a real robot with additional perceptual and kinematic constraints, accommodating them through the same mechanism.

Xing Li, Manuel Baum, Oliver Brock

We introduce a Learning from Demonstration (LfD) approach for contact-rich manipulation tasks with articulated mechanisms. The extracted policy from a single human demonstration generalizes to different mechanisms of the same type and is robust against environmental variations. The key to achieving such generalization and robustness from a single human demonstration is to autonomously augment the initial demonstration to gather additional information through purposefully interacting with the environment. Our real-world experiments on complex mechanisms with multi-DOF demonstrate that our approach can reliably accomplish the task in a changing environment. Videos are available at the: https://sites.google.com/view/rbosalfdec/home

Oussama Zenkri, Oliver Brock

Large Language Models are increasingly proposed as cognitive components for robotic systems, yet their opaque decision processes make it difficult to explain success or failure in closed-loop embodied tasks. Following an empirical AI methodology, we study embodied LLM agents behaviorally by varying the information available to the agent and measuring the resulting changes in behavior. Using the Lockbox, a sequential mechanical puzzle with hidden interdependencies, we evaluate LLMs across RGB, RGB-D, and ground-truth symbolic observations in a physical robotic setup and use controlled simulation to probe the resulting behavior. Counterintuitively, agents perform best under raw RGB input and worst under perfect ground-truth observations. In simulation, we probe this effect by randomly flipping perceived action outcomes and find that moderate noise improves performance, peaking at a 40% flip probability with a 2.85-fold success rate increase over the noise-free baseline. Further analysis links this gain to a reduction in repetitive action loops. These findings suggest that success rates alone are insufficient for evaluating LLMs, as measured performance may reflect the interaction between perceptual errors and reasoning failures rather than robust problem solving.

Xing Li, Oliver Brock

We present a Learning from Demonstration (LfD) framework that achieves one-shot generalization in multi-stage, contact-rich manipulation tasks. Central to our approach is the utilization of environmental constraints as the inductive bias. By representing a demonstration as a sequence of behaviors that exploit environmental constraints, the robot separates task-general structure -- the constraint types and their transitions -- from instance-specific details such as exact demonstration trajectories, poses, and local geometries. Our four-stage pipeline builds a complete policy on this representation: the robot first abstracts a single demonstration into environmental-constraint primitives, then disambiguates them through self-guided exploration, next assimilates targeted human corrections that handle out-of-distribution variations, and finally recovers the abstracted-away details online through compliant interaction. Because the resulting policy follows constraints rather than mimics trajectories, it generalizes across object poses, local geometries, and unmodeled contact dynamics. We validate our approach on seven real-world multi-stage contact-rich manipulation tasks and achieve over 90% success. These extensive experimental results establish environmental constraints as fundamental building blocks for efficient generalization in learning from demonstration.

Michael Migacev, Vito Mengers, Antonia Köngeter et al.

Understanding why some sequential planning problems are harder than others requires models that go beyond average performance. They should capture the specific pattern of which problems are hard, and ideally fail in the same way people do when planning capacity is reduced. We apply AICON, a reactive gradient-descent framework developed for robotic manipulation, to the Tower of London test, a cognitive test used to assess planning in Parkinson's disease, mild cognitive impairment, and stroke. Without any lookahead planning or knowledge of human cognition, AICON reproduces the fine-grained human difficulty ordering across 24 problems better than structural task parameters and generalizes to held-out problems in a leave-two-out evaluation. Crucially, AICON outperforms a planning baseline for groups with reduced planning capacity while the planning baseline better captures healthy controls. This dissociation was predicted by the original AICON paper, which noted that the model's failure modes resemble those of Parkinson's patients who struggle with goal hierarchies but not move counts. This suggests that as planning capacity is reduced, human behavior shifts toward the reactive mode AICON models. The finding extends a broader pattern: AICON, originally built for robotics, now captures aspects of biological behavior across perception, eye movements, and sequential planning, suggesting its core abstraction reflects something real about how biological systems are organized.

Vito Mengers, Bao Duc Cao, Oliver Brock

Collective behavior in animals has long been modeled through self-propelled particle models, which reproduce striking group-level phenomena through abstract interaction forces. Yet these models are fundamentally descriptive: they leave open the question of how collective behavior is actually produced. Recent empirical work makes this gap concrete: locusts do not align with neighbors, sensory and cognitive mechanisms mediate interaction instead. A mechanistic model must therefore operate at the sensorimotor level, grounded in what individual organisms can actually perceive, estimate, and physically execute. We present such a model based on a modeling framework from robotics, extended here to collective motion. Each agent perceives neighbors through bearing and apparent-size cues within a limited field of view, maintains uncertain internal state estimates, and selects actions through gradient descent on a desired social distance -- without any prescribed interaction forces. This simple model produces diverse collective behaviors including polarized motion, milling, ring formations, and subgroup fragmentation. A global sensitivity analysis shows that behavioral transitions are governed by sensorimotor parameters corresponding to measurable biological quantities: field of view geometry, sensory noise, turning agility, and memory. Collective behavior can therefore be understood as the emergent outcome of interacting sensorimotor regularities, and differences across species as the emergent outcome of differences in embodiment and environment.

Vito Mengers, Nicolas Roth, Oliver Brock et al.

The objects we perceive guide our eye movements when observing real-world dynamic scenes. Yet, gaze shifts and selective attention are critical for perceiving details and refining object boundaries. Object segmentation and gaze behavior are, however, typically treated as two independent processes. Here, we present a computational model that simulates these processes in an interconnected manner and allows for hypothesis-driven investigations of distinct attentional mechanisms. Drawing on an information processing pattern from robotics, we use a Bayesian filter to recursively segment the scene, which also provides an uncertainty estimate for the object boundaries that we use to guide active scene exploration. We demonstrate that this model closely resembles observers' free viewing behavior on a dataset of dynamic real-world scenes, measured by scanpath statistics, including foveation duration and saccade amplitude distributions used for parameter fitting and higher-level statistics not used for fitting. These include how object detections, inspections, and returns are balanced and a delay of returning saccades without an explicit implementation of such temporal inhibition of return. Extensive simulations and ablation studies show that uncertainty promotes balanced exploration and that semantic object cues are crucial to forming the perceptual units used in object-based attention. Moreover, we show how our model's modular design allows for extensions, such as incorporating saccadic momentum or pre-saccadic attention, to further align its output with human scanpaths.

Rico Jonschkowski, Divyam Rastogi, Oliver Brock

We present differentiable particle filters (DPFs): a differentiable implementation of the particle filter algorithm with learnable motion and measurement models. Since DPFs are end-to-end differentiable, we can efficiently train their models by optimizing end-to-end state estimation performance, rather than proxy objectives such as model accuracy. DPFs encode the structure of recursive state estimation with prediction and measurement update that operate on a probability distribution over states. This structure represents an algorithmic prior that improves learning performance in state estimation problems while enabling explainability of the learned model. Our experiments on simulated and real data show substantial benefits from end-to- end learning with algorithmic priors, e.g. reducing error rates by ~80%. Our experiments also show that, unlike long short-term memory networks, DPFs learn localization in a policy-agnostic way and thus greatly improve generalization. Source code is available at https://github.com/tu-rbo/differentiable-particle-filters .

Apoorv Vaish, Oliver Brock

Co-designing a robot's morphology and control can ensure synergistic interactions between them, prevalent in biological organisms. However, co-design is a high-dimensional search problem. To make this search tractable, we need a systematic method for identifying inductive biases tailored to its structure. In this paper, we analyze co-design landscapes for soft locomotion and manipulation tasks and identify three patterns that are consistent across regions of their co-design spaces. We observe that within regions of co-design space, quality varies along a low-dimensional manifold. Higher-quality regions exhibit variations spread across more dimensions, while tightly coupling morphology and control. We leverage these insights to devise an efficient co-design algorithm. Since the precise instantiation of this structure varies across tasks and is not known a priori, our algorithm infers it from information gathered during search and adapts to each task's specific structure. This yields $36\%$ more improvement than benchmark algorithms. Moreover, our algorithm achieved more than two orders of magnitude in sample efficiency compared to these benchmark algorithms, demonstrating the effectiveness of leveraging inductive biases to co-design.

Patrick Lowin, Vito Mengers, Oliver Brock

Robotic affordance estimation is challenging due to visual, geometric, and semantic ambiguities in sensory input. We propose a method that disambiguates these signals using two coupled recursive estimators for sub-aspects of affordances: graspable and movable regions. Each estimator encodes property-specific regularities to reduce uncertainty, while their coupling enables bidirectional information exchange that focuses attention on regions where both agree, i.e., affordances. Evaluated on a real-world dataset, our method outperforms three recent affordance estimators (Where2Act, Hands-as-Probes, and HRP) by 308%, 245%, and 257% in precision, and remains robust under challenging conditions such as low light or cluttered environments. Furthermore, our method achieves a 70% success rate in our real-world evaluation. These results demonstrate that coupling complementary estimators yields precise, robust, and embodiment-appropriate affordance predictions.

Aditya Bhatt, Adrian Sieler, Steffen Puhlmann et al.

We present in-hand manipulation skills on a dexterous, compliant, anthropomorphic hand. Even though these skills were derived in a simplistic manner, they exhibit surprising robustness to variations in shape, size, weight, and placement of the manipulated object. They are also very insensitive to variation of execution speeds, ranging from highly dynamic to quasi-static. The robustness of the skills leads to compositional properties that enable extended and robust manipulation programs. To explain the surprising robustness of the in-hand manipulation skills, we performed a detailed, empirical analysis of the skills' performance. From this analysis, we identify three principles for skill design: 1) Exploiting the hardware's innate ability to drive hard-to-model contact dynamics. 2) Taking actions to constrain these interactions, funneling the system into a narrow set of possibilities. 3) Composing such action sequences into complex manipulation programs. We believe that these principles constitute an important foundation for robust robotic in-hand manipulation, and possibly for manipulation in general.

Steffen Puhlmann, Jason Harris, Oliver Brock

We present the RBO Hand 3, a highly capable and versatile anthropomorphic soft hand based on pneumatic actuation. The RBO Hand 3 is designed to enable dexterous manipulation, to facilitate transfer of insights about human dexterity, and to serve as a robust research platform for extensive real-world experiments. It achieves these design goals by combining many degrees of actuation with intrinsic compliance, replicating relevant functioning of the human hand, and by combining robust components in a modular design. The RBO Hand 3 possesses 16 independent degrees of actuation, implemented in a dexterous opposable thumb, two-chambered fingers, an actuated palm, and the ability to spread the fingers. In this work, we derive the design objectives that are based on experimentation with the hand's predecessors, observations about human grasping, and insights about principles of dexterity. We explain in detail how the design features of the RBO Hand 3 achieve these goals and evaluate the hand by demonstrating its ability to achieve the highest possible score in the Kapandji test for thumb opposition, to realize all 33 grasp types of the comprehensive GRASP taxonomy, to replicate common human grasping strategies, and to perform dexterous in-hand manipulation.

Tessa J. Pannen, Steffen Puhlmann, Oliver Brock

Soft robotics is an emerging field that yields promising results for tasks that require safe and robust interactions with the environment or with humans, such as grasping, manipulation, and human-robot interaction. Soft robots rely on intrinsically compliant components and are difficult to equip with traditional, rigid sensors which would interfere with their compliance. We propose a highly flexible tactile sensor that is low-cost and easy to manufacture while measuring contact pressures independently from 14 taxels. The sensor is built from piezoresistive fabric for highly sensitive, continuous responses and from a custom-designed flexible printed circuit board which provides a high taxel density. From these taxels, location and intensity of contact with the sensor can be inferred. In this paper, we explain the design and manufacturing of the proposed sensor, characterize its input-output relation, evaluate its effects on compliance when equipped to the silicone-based pneumatic actuators of the soft robotic RBO Hand 2, and demonstrate that the sensor provides rich and useful feedback for learning-based in-hand object recognition.

Nicholas Roy, Ingmar Posner, Tim Barfoot et al.

Machine learning has long since become a keystone technology, accelerating science and applications in a broad range of domains. Consequently, the notion of applying learning methods to a particular problem set has become an established and valuable modus operandi to advance a particular field. In this article we argue that such an approach does not straightforwardly extended to robotics -- or to embodied intelligence more generally: systems which engage in a purposeful exchange of energy and information with a physical environment. In particular, the purview of embodied intelligent agents extends significantly beyond the typical considerations of main-stream machine learning approaches, which typically (i) do not consider operation under conditions significantly different from those encountered during training; (ii) do not consider the often substantial, long-lasting and potentially safety-critical nature of interactions during learning and deployment; (iii) do not require ready adaptation to novel tasks while at the same time (iv) effectively and efficiently curating and extending their models of the world through targeted and deliberate actions. In reality, therefore, these limitations result in learning-based systems which suffer from many of the same operational shortcomings as more traditional, engineering-based approaches when deployed on a robot outside a well defined, and often narrow operating envelope. Contrary to viewing embodied intelligence as another application domain for machine learning, here we argue that it is in fact a key driver for the advancement of machine learning technology. In this article our goal is to highlight challenges and opportunities that are specific to embodied intelligence and to propose research directions which may significantly advance the state-of-the-art in robot learning.

Roberto Martín-Martín, Clemens Eppner, Oliver Brock

We present a dataset with models of 14 articulated objects commonly found in human environments and with RGB-D video sequences and wrenches recorded of human interactions with them. The 358 interaction sequences total 67 minutes of human manipulation under varying experimental conditions (type of interaction, lighting, perspective, and background). Each interaction with an object is annotated with the ground truth poses of its rigid parts and the kinematic state obtained by a motion capture system. For a subset of 78 sequences (25 minutes), we also measured the interaction wrenches. The object models contain textured three-dimensional triangle meshes of each link and their motion constraints. We provide Python scripts to download and visualize the data. The data is available at https://tu-rbo.github.io/articulated-objects/ and hosted at https://zenodo.org/record/1036660/.

Niko Sünderhauf, Oliver Brock, Walter Scheirer et al.

The application of deep learning in robotics leads to very specific problems and research questions that are typically not addressed by the computer vision and machine learning communities. In this paper we discuss a number of robotics-specific learning, reasoning, and embodiment challenges for deep learning. We explain the need for better evaluation metrics, highlight the importance and unique challenges for deep robotic learning in simulation, and explore the spectrum between purely data-driven and model-driven approaches. We hope this paper provides a motivating overview of important research directions to overcome the current limitations, and help fulfill the promising potentials of deep learning in robotics.

Jeannette Bohg, Karol Hausman, Bharath Sankaran et al.

Recent approaches in robotics follow the insight that perception is facilitated by interaction with the environment. These approaches are subsumed under the term of Interactive Perception (IP). It provides the following benefits: (i) interaction with the environment creates a rich sensory signal that would otherwise not be present and (ii) knowledge of the regularity in the combined space of sensory data and action parameters facilitate the prediction and interpretation of the signal. In this survey we postulate this as a principle and collect evidence in support by analyzing and categorizing existing work in this area. We also provide an overview of the most important applications of Interactive Perception. We close this survey by discussing remaining open questions. Thereby, we hope to define a field and inspire future work.

Nikolaus Correll, Kostas E. Bekris, Dmitry Berenson et al.

This paper presents a overview of the inaugural Amazon Picking Challenge along with a summary of a survey conducted among the 26 participating teams. The challenge goal was to design an autonomous robot to pick items from a warehouse shelf. This task is currently performed by human workers, and there is hope that robots can someday help increase efficiency and throughput while lowering cost. We report on a 28-question survey posed to the teams to learn about each team's background, mechanism design, perception apparatus, planning and control approach. We identify trends in this data, correlate it with each team's success in the competition, and discuss observations and lessons learned based on survey results and the authors' personal experiences during the challenge.

Rico Jonschkowski, Sebastian Höfer, Oliver Brock

Supervised, semi-supervised, and unsupervised learning estimate a function given input/output samples. Generalization of the learned function to unseen data can be improved by incorporating side information into learning. Side information are data that are neither from the input space nor from the output space of the function, but include useful information for learning it. In this paper we show that learning with side information subsumes a variety of related approaches, e.g. multi-task learning, multi-view learning and learning using privileged information. Our main contributions are (i) a new perspective that connects these previously isolated approaches, (ii) insights about how these methods incorporate different types of prior knowledge, and hence implement different patterns, (iii) facilitating the application of these methods in novel tasks, as well as (iv) a systematic experimental evaluation of these patterns in two supervised learning tasks.