{ "generated": "2026-11-02", "sources": [ "Semantic Scholar", "arXiv", "Crossref" ], "totalCitations": 1011, "papers": [ { "id": "ascento-2019", "title": "Ascento: A Two-Wheeled Jumping Robot", "authors": [ "Victor Klemm", "Alessandro Morra", "Ciro Salzmann", "Florian Tschopp", "Karen Bodie", "Lionel Gulich", "Nicola Küng", "Dominik Mannhart", "Corentin Pfister", "Marcus Vierneisel", "Florian Weber", "Robin Deuber", "Roland Siegwart" ], "venue": "ICRA", "venueFull": "IEEE International Conference on Robotics and Automation (ICRA)", "year": 2019, "citations": 257, "theme": "wheeled", "firstAuthor": true, "featured": false, "tldr": "Ascento is introduced, a compact wheeled bipedal robot that is able to move quickly on flat terrain, and to overcome obstacles by jumping, as well as the development of various controllers for different scenarios.", "abstract": "Applications of mobile ground robots demand high speed and agility while navigating in complex indoor environments. These present an ongoing challenge in mobile robotics. A system with these specifications would be of great use for a wide range of indoor inspection tasks. This paper introduces Ascento, a compact wheeled bipedal robot that is able to move quickly on flat terrain, and to overcome obstacles by jumping. The mechanical design and overall architecture of the system is presented, as well as the development of various controllers for different scenarios. A series of experiments with the final prototype system validate these behaviors in realistic scenarios.", "links": { "arxiv": "https://arxiv.org/abs/2005.11435", "pdf": "https://arxiv.org/pdf/2005.11435", "doi": "https://doi.org/10.1109/icra.2019.8793792", "project": null, "video": null, "code": null }, "teaser": "assets/teasers/ascento-2019.webp" }, { "id": "lqr-wbc-2020", "title": "LQR-Assisted Whole-Body Control of a Wheeled Bipedal Robot With Kinematic Loops", "authors": [ "Victor Klemm", "Alessandro Morra", "Lionel Gulich", "Dominik Mannhart", "David Rohr", "Mina Kamel", "Yvain de Viragh", "Roland Siegwart" ], "venue": "RA-L", "venueFull": "IEEE Robotics and Automation Letters (RA-L)", "year": 2020, "citations": 180, "theme": "wheeled", "firstAuthor": true, "featured": false, "tldr": "This work presents a hierarchical whole-body controller leveraging the full rigid body dynamics of the wheeled bipedal robot Ascento to derive closed-form expressions for the dynamics of its kinematic loops in a way that readily generalizes to more complex systems.", "abstract": "We present a hierarchical whole-body controller leveraging the full rigid body dynamics of the wheeled bipedal robot Ascento. We derive closed-form expressions for the dynamics of its kinematic loops in a way that readily generalizes to more complex systems. The rolling constraint is incorporated using a compact analytic solution based on rotation matrices. The non-minimum phase balancing dynamics are accounted for by including a linear-quadratic regulator as a motion task. Robustness when driving curves is increased by regulating the lean angle as a function of the zero-moment point. The proposed controller is computationally lightweight and significantly extends the rough-terrain capabilities and robustness of the system, as we demonstrate in several experiments.", "links": { "arxiv": "https://arxiv.org/abs/2005.11431", "pdf": "https://arxiv.org/pdf/2005.11431", "doi": "https://doi.org/10.1109/lra.2020.2979625", "project": null, "video": null, "code": null }, "teaser": "assets/teasers/lqr-wbc-2020.webp" }, { "id": "cami-2020", "title": "CAMI - Analysis, Design and Realization of a Force-Compliant Variable Cam System", "authors": [ "Dominik Mannhart", "Fabio Dubois", "Karen Bodie", "Victor Klemm", "Alessandro Morra", "Marco Hutter" ], "venue": "ICRA", "venueFull": "IEEE International Conference on Robotics and Automation (ICRA)", "year": 2020, "citations": 4, "theme": "wheeled", "firstAuthor": false, "featured": false, "tldr": "This work presents a novel design concept that achieves multi-legged locomotion using a three-dimensional cam system.", "abstract": "This work presents a novel design concept that achieves multi-legged locomotion using a three-dimensional cam system. A computational framework has been developed to analyze and dimension this cam apparatus, that can perform arbitrary end effector motions within its design constraints. The mechanism enables continuous gait transition and inherent force compliance. With only two motors, any trajectory of a continuous set of gaits can be followed. One motor is used to actuate the system and a second one to morph its movement. To illustrate a possible application of this system, a working prototype of a bipedal robot is developed and validated in hardware. It showcases a smooth velocity change by transitioning through different gaits from standing still to walking fast at 124mm/s within 2.0s, while following the given end effector trajectory with an error of only 2.47mm.", "links": { "arxiv": null, "pdf": "https://www.research-collection.ethz.ch/bitstreams/df5b50c6-6140-4ce1-9a64-e84b2c0f6cf6/download", "doi": "https://doi.org/10.1109/icra40945.2020.9197019", "project": null, "video": null, "code": null }, "teaser": null }, { "id": "wheeled-legged-survey-2022", "title": "A Survey of Wheeled-Legged Robots", "authors": [ "Marko Bjelonic", "Victor Klemm", "Joonho Lee", "Marco Hutter" ], "venue": "CLAWAR", "venueFull": "International Conference on Climbing and Walking Robots (CLAWAR)", "year": 2022, "citations": 55, "theme": "wheeled", "firstAuthor": false, "featured": false, "tldr": "A survey of wheeled-legged robots, which pair the efficiency of wheels on flat ground with the versatility of legs on challenging terrain, reviewing locomotion frameworks that cope with the added complexity of hybrid locomotion.", "abstract": "The community in legged robotics focuses on bio-inspired robots, although there are some human inventions that nature could not recreate. One of the most significant examples is the wheel which has made our transportation system more efficient and faster. Inspired by this human-made evolution, we present a survey of wheeled-legged robots, allowing robotic systems to be efficient on flat as well as versatile on challenging terrain. This enhancement, however, comes at the cost of increased complexity due to additional degrees of freedom at the end-effector, which empowers motions along the rolling direction. The missing examples in nature make designing templates that capture the underlying locomotion principles cumbersome, making hybrid locomotion challenging. This paper reviews some of the novel locomotion frameworks overcoming these challenges.", "links": { "arxiv": null, "pdf": null, "doi": "https://doi.org/10.1007/978-3-031-15226-9_11", "project": null, "video": null, "code": null }, "teaser": null }, { "id": "four-bar-leg-2022", "title": "Design Optimization of a Four-Bar Leg Linkage for a Legged-Wheeled Balancing Robot", "authors": [ "Victor Klemm", "Dominik Mannhart", "Roland Siegwart" ], "venue": "CLAWAR", "venueFull": "International Conference on Climbing and Walking Robots (CLAWAR)", "year": 2022, "citations": 3, "theme": "wheeled", "firstAuthor": true, "featured": false, "tldr": "A design optimization procedure for a 1-DOF four-bar leg linkage that keeps the base pitch of a balancing legged-wheeled robot nearly constant over the leg's extension, with gravity compensation from an optimized torsional spring, evaluated on real hardware.", "abstract": "Balancing legged-wheeled robots have gained popularity in recent years due to their locomotive efficiency while still being able to conquer rough terrain and obstacles. Furthermore, as this type of robot maintains ground contact with its wheels for most of the time, passive gravity compensation mechanisms can greatly minimize power consumption. Various designs with different leg configurations have emerged, whereby a 1-DOF mechanism per leg already showed sufficient compliance to adapt to most outdoor terrain. We propose a design optimization procedure for a 1-DOF four-bar leg linkage to ensure minimum pitch angle correction of the robot’s base while varying the leg extension. Gravity compensation is further achieved through an optimized torsional spring. Finally, we evaluate the performance of the leg linkage and gravity compensation mechanism on real hardware.", "links": { "arxiv": null, "pdf": null, "doi": "https://doi.org/10.1007/978-3-031-15226-9_15", "project": null, "video": null, "code": null }, "teaser": null }, { "id": "multi-amp-2023", "title": "Advanced Skills through Multiple Adversarial Motion Priors in Reinforcement Learning", "authors": [ "Eric Vollenweider", "Marko Bjelonic", "Victor Klemm", "Nikita Rudin", "Joonho Lee", "Marco Hutter" ], "venue": "ICRA", "venueFull": "IEEE International Conference on Robotics and Automation (ICRA)", "year": 2023, "citations": 121, "theme": "rl", "firstAuthor": false, "featured": false, "tldr": "This work presents an approach to enhance the concept of adversarial motion prior-based RL, allowing for multiple, discretely switchable motion styles, and demonstrates that multiple styles and skills can be learned simultaneously without significant performance differences, even in combination with motion data-free skills.", "abstract": "In recent years, reinforcement learning (RL) has shown outstanding performance for locomotion control of highly articulated robotic systems. Such approaches typically involve tedious reward function tuning to achieve the desired motion style. Imitation learning approaches such as adversarial motion priors aim to reduce this problem by encouraging a pre-defined motion style. In this work, we present an approach to augment the concept of adversarial motion prior-based RL to allow for multiple, discretely switchable styles. We show that multiple styles and skills can be learned simultaneously without notable performance differences, even in combination with motion data-free skills. Our approach is validated in several real-world experiments with a wheeled-legged quadruped robot showing skills learned from existing RL controllers and trajectory optimization, such as ducking and walking, and novel skills such as switching between a quadrupedal and humanoid configuration. For the latter skill, the robot is required to stand up, navigate on two wheels, and sit down. Instead of tuning the sit-down motion, we verify that a reverse playback of the stand-up movement helps the robot discover feasible sit-down behaviors and avoids tedious reward function tuning.", "links": { "arxiv": "https://arxiv.org/abs/2203.14912", "pdf": "https://arxiv.org/pdf/2203.14912", "doi": "https://doi.org/10.1109/icra48891.2023.10160751", "project": null, "video": null, "code": null }, "teaser": "assets/teasers/multi-amp-2023.webp" }, { "id": "rayen-2023", "title": "RAYEN: Imposition of Hard Convex Constraints on Neural Networks", "authors": [ "Jesús Tordesillas", "Victor Klemm", "Jonathan P. How", "Marco Hutter" ], "venue": "arXiv", "venueFull": "arXiv preprint", "year": 2023, "citations": 46, "theme": "planning", "firstAuthor": false, "featured": false, "tldr": "RAYEN is presented, a framework for imposing hard convex constraints on the output or latent variables of a neural network, and its ability to enforce actuator constraints on a learned locomotion policy is demonstrated by validating constraint satisfaction in both simulation and real-world experiments on a quadruped robot.", "abstract": "Despite the numerous applications of convex constraints in Robotics, enforcing them within learning-based frameworks remains an open challenge. Existing techniques either fail to guarantee satisfaction at all times, or incur prohibitive computational costs. This paper presents RAYEN, a framework for imposing hard convex constraints on the output or latent variables of a neural network. RAYEN guarantees constraint satisfaction during both training and testing, for any input and any network weights. Unlike prior approaches, RAYEN avoids computationally expensive orthogonal projections, soft constraints, conservative approximations of the feasible set, and slow iterative corrections. RAYEN supports any combination of linear, convex quadratic, second-order cone (SOC), and linear matrix inequality (LMI) constraints, with negligible overhead compared to unconstrained networks. For instance, it imposes 1K quadratic constraints on a 1K-dimensional variable with only 8 ms of overhead compared to a network that does not enforce these constraints. An LMI constraint with 300x300 dense matrices on a 10K-dimensional variable can be guaranteed with only 12 ms additional overhead. When used in neural networks that approximate the solution of constrained trajectory optimization problems, RAYEN runs 20 to 7468 times faster than state-of-the-art algorithms, while guaranteeing constraint satisfaction at all times and achieving a near-optimal cost (<1.5% optimality gap). Finally, we demonstrate RAYEN's ability to enforce actuator constraints on a learned locomotion policy by validating constraint satisfaction in both simulation and real-world experiments on a quadruped robot. The code is available at https://github.com/leggedrobotics/rayen", "links": { "arxiv": "https://arxiv.org/abs/2307.08336", "pdf": "https://arxiv.org/pdf/2307.08336", "doi": null, "project": null, "video": null, "code": "https://github.com/leggedrobotics/rayen" }, "teaser": "assets/teasers/rayen-2023.webp" }, { "id": "curiosity-locomanip-2023", "title": "Curiosity-Driven Learning of Joint Locomotion and Manipulation Tasks", "authors": [ "Clemens Schwarke", "Victor Klemm", "Matthijs van der Boon", "Marko Bjelonic", "Marco Hutter" ], "venue": "CoRL", "venueFull": "Conference on Robot Learning (CoRL)", "year": 2023, "citations": 51, "theme": "rl", "firstAuthor": false, "featured": false, "tldr": "This paper adapts the idea of Random Network Distillation to the robotics domain to learn holistic motion control policies involving simultaneous locomotion and manipulation and investigates opening doors as an exemplary task for robotic applications.", "abstract": "Learning complex locomotion and manipulation tasks presents significant challenges, often requiring extensive engineering of, e.g., reward functions or curricula to provide meaningful feedback to the Reinforcement Learning (RL) algorithm. This paper proposes an intrinsically motivated RL approach to reduce task-specific engineering. The desired task is encoded in a single sparse reward, i.e., a reward of “+1” is given if the task is achieved. Intrinsic motivation enables learning by guiding exploration toward the sparse reward signal. Specifically, we adapt the idea of Random Network Distillation (RND) to the robotics domain to learn holistic motion control policies involving simultaneous locomotion and manipulation. We investigate opening doors as an exemplary task for robotic applications. A second task involving package manipulation from a table to a bin highlights the generalization capabilities of the presented approach. Finally, the resulting RL policies are executed in real-world experiments on a wheeled-legged robot in biped mode. We experienced no failure in our experiments, which consisted of opening push doors (over 15 times in a row) and manipulating packages (over 5 times in a row).", "links": { "arxiv": null, "pdf": "https://proceedings.mlr.press/v229/schwarke23a/schwarke23a.pdf", "doi": null, "project": null, "video": null, "code": null }, "teaser": null }, { "id": "blind-stairs-2024", "title": "Reinforcement Learning for Blind Stair Climbing with Legged and Wheeled-Legged Robots", "authors": [ "Simon Chamorro", "Victor Klemm", "Miguel de la Iglesia Valls", "Christopher Pal", "Roland Siegwart" ], "venue": "ICRA", "venueFull": "IEEE International Conference on Robotics and Automation (ICRA)", "year": 2024, "citations": 47, "theme": "wheeled", "firstAuthor": false, "featured": false, "tldr": "A position-based reinforcement learning formulation with an asymmetric actor-critic yields one blind stair-climbing controller for legged and wheeled-legged robots, shown on quadrupeds and bipeds in simulation and on the balancing robot Ascento, which climbs 15 cm stairs in the real world, a task previously impossible for it.", "abstract": "In recent years, legged and wheeled-legged robots have gained prominence for tasks in environments predominantly created for humans across various domains. One significant challenge faced by many of these robots is their limited capability to navigate stairs, which hampers their functionality in multi-story environments. This study proposes a method aimed at addressing this limitation, employing reinforcement learning to develop a versatile controller applicable to a wide range of robots. In contrast to the conventional velocity-based controllers, our approach builds upon a position-based formulation of the RL task, which we show to be vital for stair climbing. Furthermore, the methodology leverages an asymmetric actor-critic structure, enabling the utilization of privileged information from simulated environments during training while eliminating the reliance on exteroceptive sensors during real-world deployment. Another key feature of the proposed approach is the incorporation of a boolean observation within the controller, enabling the activation or deactivation of a stair-climbing mode. We present our results on different quadrupeds and bipedal robots in simulation and showcase how our method allows the balancing robot Ascento to climb 15cm stairs in the real world, a task that was previously impossible for this robot.", "links": { "arxiv": "https://arxiv.org/abs/2402.06143", "pdf": "https://arxiv.org/pdf/2402.06143", "doi": "https://doi.org/10.1109/icra57147.2024.10610069", "project": null, "video": "https://youtu.be/Ec6ar8BVJh4", "code": null }, "teaser": "assets/teasers/blind-stairs-2024.webp" }, { "id": "symmetry-2024", "title": "Symmetry Considerations for Learning Task Symmetric Robot Policies", "authors": [ "Mayank Mittal", "Nikita Rudin", "Victor Klemm", "Arthur Allshire", "Marco Hutter" ], "venue": "ICRA", "venueFull": "IEEE International Conference on Robotics and Automation (ICRA)", "year": 2024, "citations": 63, "theme": "rl", "firstAuthor": false, "featured": false, "tldr": "This paper investigates two approaches to incorporate symmetry invariance into deep RL (data augmentation and a mirror loss function), provides a theoretical foundation for augmented samples in on-policy training, and shows that this achieves faster convergence and better learned behaviors in challenging robotic tasks, from climbing boxes with a quadruped to dexterous manipulation.", "abstract": "Symmetry is a fundamental aspect of many real-world robotic tasks. However, current deep reinforcement learning (DRL) approaches can seldom harness and exploit symmetry effectively. Often, the learned behaviors fail to achieve the desired transformation invariances and suffer from motion artifacts. For instance, a quadruped may exhibit different gaits when commanded to move forward or backward, even though it is symmetrical about its torso. This issue becomes further pronounced in high-dimensional or complex environments, where DRL methods are prone to local optima and fail to explore regions of the state space equally. Past methods on encouraging symmetry for robotic tasks have studied this topic mainly in a single-task setting, where symmetry usually refers to symmetry in the motion, such as the gait patterns. In this paper, we revisit this topic for goal-conditioned tasks in robotics, where symmetry lies mainly in task execution and not necessarily in the learned motions themselves. In particular, we investigate two approaches to incorporate symmetry invariance into DRL -- data augmentation and mirror loss function. We provide a theoretical foundation for using augmented samples in an on-policy setting. Based on this, we show that the corresponding approach achieves faster convergence and improves the learned behaviors in various challenging robotic tasks, from climbing boxes with a quadruped to dexterous manipulation.", "links": { "arxiv": "https://arxiv.org/abs/2403.04359", "pdf": "https://arxiv.org/pdf/2403.04359", "doi": "https://doi.org/10.1109/icra57147.2024.10611493", "project": null, "video": null, "code": null }, "teaser": "assets/teasers/symmetry-2024.webp" }, { "id": "constrained-rl-2024", "title": "Exploring Constrained Reinforcement Learning Algorithms for Quadrupedal Locomotion", "authors": [ "Joonho Lee", "Lukas Schroth", "Victor Klemm", "Marko Bjelonic", "Alexander Reske", "Marco Hutter" ], "venue": "IROS", "venueFull": "IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)", "year": 2024, "citations": 27, "theme": "rl", "firstAuthor": false, "featured": false, "tldr": "This paper experiments with the Constrained Markov Decision Process (CMDP) framework instead of the conventional unconstrained RL for robotic applications, and performs a comparative study of different constrained policy optimization algorithms to identify suitable methods for practical implementation.", "abstract": "Shifting from traditional control strategies to Deep Reinforcement Learning (RL) for legged robots poses inherent challenges, especially when addressing real-world physical constraints during training. While high-fidelity simulations provide significant benefits, they often bypass these essential physical limitations. In this paper, we experiment with the Constrained Markov Decision Process (CMDP) framework instead of the conventional unconstrained RL for robotic applications. We perform a comparative study of different constrained policy optimization algorithms to identify suitable methods for practical implementation. Our robot experiments demonstrate the critical role of incorporating physical constraints, yielding successful sim-to-real transfers, and reducing operational errors on physical systems. The CMDP formulation streamlines the training process by separately handling constraints from rewards. Our findings underscore the potential of constrained RL for the effective development and deployment of learned controllers in robotics.", "links": { "arxiv": "https://arxiv.org/abs/2309.15430", "pdf": "https://arxiv.org/pdf/2309.15430", "doi": "https://doi.org/10.1109/iros58592.2024.10801341", "project": null, "video": null, "code": null }, "teaser": "assets/teasers/constrained-rl-2024.webp" }, { "id": "diffsim2real-2024", "title": "DiffSim2Real: Deploying Quadrupedal Locomotion Policies Purely Trained in Differentiable Simulation", "authors": [ "Joshua Bagajo", "Clemens Schwarke", "Victor Klemm", "Ignat Georgiev", "Jean-Pierre Sleiman", "Jesús Tordesillas", "Animesh Garg", "Marco Hutter" ], "venue": "arXiv", "venueFull": "arXiv preprint", "year": 2024, "citations": 3, "theme": "diffsim", "firstAuthor": false, "featured": false, "tldr": "This work demonstrates that locomotion policies trained with analytic gradients from a differentiable simulator can be successfully transferred to the real world, marking the first time a real quadrupedal robot is able to locomote after training exclusively in a differentiable simulation.", "abstract": "Differentiable simulators provide analytic gradients, enabling more sample-efficient learning algorithms and paving the way for data intensive learning tasks such as learning from images. In this work, we demonstrate that locomotion policies trained with analytic gradients from a differentiable simulator can be successfully transferred to the real world. Typically, simulators that offer informative gradients lack the physical accuracy needed for sim-to-real transfer, and vice-versa. A key factor in our success is a smooth contact model that combines informative gradients with physical accuracy, ensuring effective transfer of learned behaviors. To the best of our knowledge, this is the first time a real quadrupedal robot is able to locomote after training exclusively in a differentiable simulation.", "links": { "arxiv": "https://arxiv.org/abs/2411.02189", "pdf": "https://arxiv.org/pdf/2411.02189", "doi": null, "project": null, "video": null, "code": null }, "teaser": "assets/teasers/diffsim2real-2024.webp" }, { "id": "llm-ctf-2024", "title": "Dataset and Lessons Learned from the 2024 SaTML LLM Capture-the-Flag Competition", "authors": [ "Edoardo Debenedetti", "Javier Rando", "Daniel Paleka", "Fineas Silaghi", "Dragos Albastroiu", "Niv Cohen", "Yuval Lemberg", "Reshmi Ghosh", "Rui Wen", "Ahmed Salem", "Giovanni Cherubin", "Santiago Zanella-Beguelin", "Robin Schmid", "Victor Klemm", "Takahiro Miki", "Chenhao Li", "Stefan Kraft", "Mario Fritz", "Florian Tramèr", "Sahar Abdelnabi", "Lea Schönherr" ], "venue": "NeurIPS", "venueFull": "Conference on Neural Information Processing Systems (NeurIPS), Datasets and Benchmarks Track", "year": 2024, "citations": 27, "theme": "other", "firstAuthor": false, "featured": false, "tldr": "A capture-the-flag competition at IEEE SaTML 2024 found that all defenses were bypassed at least once, highlighting the difficulty of designing a successful defense and the necessity for additional research to protect LLM systems.", "abstract": "Large language model systems face important security risks from maliciously crafted messages that aim to overwrite the system's original instructions or leak private data. To study this problem, we organized a capture-the-flag competition at IEEE SaTML 2024, where the flag is a secret string in the LLM system prompt. The competition was organized in two phases. In the first phase, teams developed defenses to prevent the model from leaking the secret. During the second phase, teams were challenged to extract the secrets hidden for defenses proposed by the other teams. This report summarizes the main insights from the competition. Notably, we found that all defenses were bypassed at least once, highlighting the difficulty of designing a successful defense and the necessity for additional research to protect LLM systems. To foster future research in this direction, we compiled a dataset with over 137k multi-turn attack chats and open-sourced the platform.", "links": { "arxiv": "https://arxiv.org/abs/2406.07954", "pdf": "https://arxiv.org/pdf/2406.07954", "doi": "https://doi.org/10.52202/079017-1164", "project": null, "video": null, "code": null }, "teaser": "assets/teasers/llm-ctf-2024.webp" }, { "id": "nonsmooth-to-2024", "title": "Nonsmooth Trajectory Optimization for Wheeled Balancing Robots With Contact Switches and Impacts", "authors": [ "Victor Klemm", "Yvain de Viragh", "David Rohr", "Roland Siegwart", "Marco Tognon" ], "venue": "T-RO", "venueFull": "IEEE Transactions on Robotics (T-RO)", "year": 2025, "citations": 25, "theme": "wheeled", "firstAuthor": true, "featured": false, "tldr": "A fast trajectory optimizer for wheeled-legged balancing robots that explicitly models contact switches and impacts, letting the wheeled biped Ascento traverse steps, ramps and entire stairways, jump and stand up in real-world experiments.", "abstract": "Recent years have seen a steady rise in the abilities of wheeled–legged balancing robots. Yet, their use is still severely restricted by the lack of efficient control algorithms for overcoming obstacles such as stairs. We take a considerable step toward closing this gap by presenting a fast trajectory optimizer for generating trajectories over a large class of challenging terrains. By limiting the underlying modeling to the planar, nonlinear rigid-body dynamics and subdividing the terrain into contact-phases, a tractable nonlinear programming problem is obtained. The model explicitly accounts for contact switches and impacts, traction limits, and actuation bounds. By introducing an arc-length-related parametrization, the trajectories are rendered inherently contact constraint-consistent. We apply our method to the specific case of the wheeled bipedal robot Ascento, for which we derive closed-form expressions of the dynamics equations, including the kinematic loops. To track the trajectories, we propose a simple LQR-based controller. The approach is validated in real-world experiments where we show the execution of trajectories for traversing steps, driving up ramps, jumping, standing up, and driving up entire stairways. To the best of our knowledge, enabling the latter by means of trajectory optimization (TO) is a novelty for wheeled–legged robots.", "links": { "arxiv": null, "pdf": null, "doi": "https://doi.org/10.1109/tro.2023.3326334", "project": null, "video": null, "code": null }, "teaser": null }, { "id": "leva-2025", "title": "LEVA: A High-Mobility Logistic Vehicle with Legged Suspension", "authors": [ "Marco Arnold", "Lukas Hildebrandt", "Kaspar Janssen", "Efe Ongan", "Pascal Bürge", "Ádám Gyula Gábriel", "James Kennedy", "Rishi Lolla", "Quanisha Oppliger", "Micha Schaaf", "Joseph Church", "Michael Fritsche", "Victor Klemm", "Turcan Tuna", "Giorgio Valsecchi", "Cedric Weibel", "Michael Wüthrich", "Marco Hutter" ], "venue": "ICRA", "venueFull": "IEEE International Conference on Robotics and Automation (ICRA)", "year": 2025, "citations": 12, "theme": "wheeled", "firstAuthor": false, "featured": false, "tldr": "LEVA is a high-payload, high-mobility logistics robot with a parallel-kinematics legged suspension and steerable wheels that climbs stairs with an RL controller and autonomously loads and carries up to 85 kg over uneven terrain.", "abstract": "The autonomous transportation of materials over challenging terrain is a challenge with major economic implications and remains unsolved. This paper introduces LEVA, a high-payload, high-mobility robot designed for autonomous logistics across varied terrains, including those typical in agriculture, construction, and search and rescue operations. LEVA uniquely integrates an advanced legged suspension system using parallel kinematics. It is capable of traversing stairs using an RL controller, has steerable wheels, and includes a specialized box pickup mechanism that enables autonomous payload loading as well as precise and reliable cargo transportation of up to 85 kg across uneven surfaces, steps and inclines while maintaining a cost of transport (CoT) as low as 0.15. Through extensive experimental validation, LEVA demonstrates its off-road capabilities and reliability regarding payload loading and transport.", "links": { "arxiv": "https://arxiv.org/abs/2503.10028", "pdf": "https://arxiv.org/pdf/2503.10028", "doi": "https://doi.org/10.1109/icra55743.2025.11128847", "project": null, "video": null, "code": null }, "teaser": "assets/teasers/leva-2025.webp" }, { "id": "marladona-2025", "title": "MARLadona - Towards Cooperative Team Play Using Multi-Agent Reinforcement Learning", "authors": [ "Zichong Li", "Filip Bjelonic", "Victor Klemm", "Marco Hutter" ], "venue": "ICRA", "venueFull": "IEEE International Conference on Robotics and Automation (ICRA)", "year": 2025, "citations": 13, "theme": "rl", "firstAuthor": false, "featured": false, "tldr": "A decentralized multi-agent reinforcement learning (MARL) training pipeline capable of producing agents with sophisticated team play behavior, bridging the shortcomings of heuristic methods is introduced.", "abstract": "Robot soccer, in its full complexity, poses an unsolved research challenge. Current solutions heavily rely on engineered heuristic strategies, which lack robustness and adaptability. Deep reinforcement learning has gained significant traction in various complex robotics tasks such as locomotion, manipulation, and competitive games (e.g., AlphaZero, OpenAI Five), making it a promising solution to the robot soccer problem. This paper introduces MARLadona. A decentralized multi-agent reinforcement learning (MARL) training pipeline capable of producing agents with sophisticated team play behavior, bridging the shortcomings of heuristic methods. Furthermore, we created an open-source multi-agent soccer environment. Utilizing our MARL framework and a modified global entity encoder (GEE) as our core architecture, our approach achieves a 66.8% win rate against HELIOS agent, which employs a state-of-the-art heuristic strategy. In addition, we provided an in-depth analysis of the policy behavior and interpreted the agent's intention using the critic network.", "links": { "arxiv": "https://arxiv.org/abs/2409.20326", "pdf": "https://arxiv.org/pdf/2409.20326", "doi": "https://doi.org/10.1109/icra55743.2025.11128256", "project": null, "video": null, "code": null }, "teaser": "assets/teasers/marladona-2025.webp" }, { "id": "residual-diffsim-2025", "title": "Residual Policy Learning for Perceptive Quadruped Control Using Differentiable Simulation", "authors": [ "Jing Yuan Luo", "Yunlong Song", "Victor Klemm", "Fan Shi", "Davide Scaramuzza", "Marco Hutter" ], "venue": "ICRA", "venueFull": "IEEE International Conference on Robotics and Automation (ICRA)", "year": 2025, "citations": 26, "theme": "diffsim", "firstAuthor": false, "featured": false, "tldr": "For quadruped locomotion, it is found that the role of residual policy learning in FoPG-based training (FoPG RPL) is primarily to improve asymptotic rewards, compared to improving sample efficiency for model-free RL.", "abstract": "First-order Policy Gradient (FoPG) algorithms such as Backpropagation through Time and Analytical Policy Gradients leverage local simulation physics to accelerate policy search, significantly improving sample efficiency in robot control compared to standard model-free reinforcement learning. However, FoPG algorithms can exhibit poor learning dynamics in contact-rich tasks like locomotion. Previous approaches address this issue by alleviating contact dynamics via algorithmic or simulation innovations. In contrast, we propose guiding the policy search by learning a residual over a simple baseline policy. For quadruped locomotion, we find that the role of residual policy learning in FoPG-based training (FoPG RPL) is primarily to improve asymptotic rewards, compared to improving sample efficiency for model-free RL. Additionally, we provide insights on applying FoPG's to pixel-based local navigation, training a point-mass robot to convergence within seconds. Finally, we showcase the versatility of FoPG RPL by using it to train locomotion and perceptive navigation end-to-end on a quadruped in minutes.", "links": { "arxiv": "https://arxiv.org/abs/2410.03076", "pdf": "https://arxiv.org/pdf/2410.03076", "doi": "https://doi.org/10.1109/icra55743.2025.11127448", "project": null, "video": null, "code": null }, "teaser": "assets/teasers/residual-diffsim-2025.webp" }, { "id": "diffsim-locomotion-2024", "title": "Learning Deployable Locomotion Control via Differentiable Simulation", "authors": [ "Clemens Schwarke", "Victor Klemm", "Joshua Bagajo", "Jean-Pierre Sleiman", "Ignat Georgiev", "Jesús Tordesillas", "Marco Hutter" ], "venue": "CoRL", "venueFull": "Conference on Robot Learning (CoRL)", "year": 2025, "citations": 15, "theme": "diffsim", "firstAuthor": false, "featured": false, "tldr": "A differentiable contact model with informative gradients and high physical fidelity enables the first sim-to-real transfer of a legged locomotion policy learned entirely in a differentiable simulator.", "abstract": "Differentiable simulators promise to improve sample efficiency in robot learning by providing analytic gradients of the system dynamics. Yet, their application to contact-rich tasks like locomotion is complicated by the inherently non-smooth nature of contact, impeding effective gradient-based optimization. Existing works thus often rely on soft contact models that provide smooth gradients but lack physical accuracy, constraining results to simulation. To address this limitation, we propose a differentiable contact model designed to provide informative gradients while maintaining high physical fidelity. We demonstrate the efficacy of our approach by training a quadrupedal locomotion policy within our differentiable simulator leveraging analytic gradients and successfully transferring the learned policy zero-shot to the real world. To the best of our knowledge, this represents the first successful sim-to-real transfer of a legged locomotion policy learned entirely within a differentiable simulator, establishing the feasibility of using differentiable simulation for real-world locomotion control.", "links": { "arxiv": "https://arxiv.org/abs/2404.02887", "pdf": "https://arxiv.org/pdf/2404.02887", "doi": null, "project": null, "video": null, "code": null }, "teaser": "assets/teasers/diffsim-locomotion-2024.webp" }, { "id": "foci-2025", "title": "FOCI: Trajectory Optimization on Gaussian Splats", "authors": [ "Mario Gomez Andreu", "Maximum Wilder-Smith", "Victor Klemm", "Vaishakh Patil", "Jesús Tordesillas", "Marco Hutter" ], "venue": "IROS", "venueFull": "IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)", "year": 2025, "citations": 6, "theme": "planning", "firstAuthor": false, "featured": false, "tldr": "This work proposes FOCI (Field Overlap Collision Integral), an algorithm that is able to optimize trajectories directly on the Gaussians themselves, using the notion of the overlap integral between Gaussians in 3D Gaussian Splats.", "abstract": "3D Gaussian Splatting (3DGS) has recently gained popularity as a faster alternative to Neural Radiance Fields (NeRFs) in 3D reconstruction and view synthesis methods. Leveraging the spatial information encoded in 3DGS, this work proposes FOCI (Field Overlap Collision Integral), an algorithm that is able to optimize trajectories directly on the Gaussians themselves. FOCI leverages a novel and interpretable collision formulation for 3DGS using the notion of the overlap integral between Gaussians. Contrary to other approaches, which represent the robot with conservative bounding boxes that underestimate the traversability of the environment, we propose to represent the environment and the robot as Gaussian Splats. This not only has desirable computational properties, but also allows for orientation-aware planning, allowing the robot to pass through very tight and narrow spaces. We extensively test our algorithm in both synthetic and real Gaussian Splats, showcasing that collision-free trajectories for the ANYmal legged robot that can be computed in a few seconds, even with hundreds of thousands of Gaussians making up the environment. The project page and code are available at https://rffr.leggedrobotics.com/works/foci/", "links": { "arxiv": "https://arxiv.org/abs/2505.08510", "pdf": "https://arxiv.org/pdf/2505.08510", "doi": "https://doi.org/10.1109/iros60139.2025.11245845", "project": "https://rffr.leggedrobotics.com/works/foci/", "video": null, "code": null }, "teaser": "assets/teasers/foci-2025.webp" }, { "id": "motion-embedding-2025", "title": "Multi-Domain Motion Embedding: Expressive Real-Time Mimicry for Legged Robots", "authors": [ "Matthias Heyrman", "Chenhao Li", "Victor Klemm", "Dongho Kang", "Stelian Coros", "Marco Hutter" ], "venue": "arXiv", "venueFull": "arXiv preprint", "year": 2025, "citations": 2, "theme": "rl", "firstAuthor": false, "featured": false, "tldr": "Multi-Domain Motion Embedding (MDME) combines a wavelet-based encoder for structured, periodic motion patterns with a probabilistic embedding for aperiodic ones; robot policies conditioned on it imitate unseen human and animal motions in real time without retargeting on humanoid and quadruped platforms, outperforming prior approaches in motion reproduction and generalization.", "abstract": "Effective motion representation is crucial for enabling robots to imitate expressive behaviors in real time, yet existing motion controllers often ignore inherent patterns in motion. Previous efforts in representation learning do not attempt to jointly capture human and animal movements through structured periodic patterns and variational aperiodic descriptions. To address this, we present Multi-Domain Motion Embedding (MDME), a motion representation that unifies the complementary embedding of structured and unstructured features using a wavelet-based encoder and a probabilistic embedding in parallel. This produces a rich representation of reference motions from a minimal input set that generalizes across diverse motion styles. We evaluate MDME on retargeting-free motion imitation at deployment by conditioning robot control policies on the learned embeddings to reconstruct ideal retargeted states on the robot, demonstrating accurate reproduction of long-horizon trajectories on both humanoid and quadruped platforms. Our comparative studies confirm that MDME outperforms prior approaches in motion reproduction and generalization to unseen motions. Furthermore, we demonstrate real-time zero-shot deployment on unseen motions, removing per-motion tuning and online retargeting. These results show that MDME provides a generalizable and structure-aware foundation for scalable real-time robot imitation.", "links": { "arxiv": "https://arxiv.org/abs/2512.07673", "pdf": "https://arxiv.org/pdf/2512.07673", "doi": null, "project": null, "video": null, "code": null }, "teaser": "assets/teasers/motion-embedding-2025.webp" }, { "id": "ame2-2026", "title": "Agile and Generalized Legged Locomotion via Attention-Based Neural Map Encoding", "authors": [ "Chong Zhang", "Victor Klemm", "Fan Yang", "Marco Hutter" ], "venue": "T-RO", "venueFull": "IEEE Transactions on Robotics (T-RO)", "year": 2026, "citations": 27, "theme": "rl", "firstAuthor": false, "featured": false, "tldr": "A unified reinforcement learning (RL) framework for agile and generalized locomotion that incorporates a novel attention-based map encoder in the control policy and a learning-based mapping pipeline that provides fast, uncertainty-aware terrain representations robust to noise and occlusions, serving as policy inputs.", "abstract": "Achieving agile and generalized legged locomotion across terrains requires tight integration of perception and control, especially under occlusions and sparse footholds. Existing methods have demonstrated agility on parkour courses but often rely on end-to-end sensorimotor models with limited generalization and interpretability. By contrast, methods targeting generalized locomotion typically exhibit limited agility and struggle with visual occlusions. We introduce a unified reinforcement learning (RL) framework for agile and generalized locomotion that incorporates a novel attention-based map encoder in the control policy. This encoder extracts local and global mapping features and uses attention mechanisms to focus on salient regions, producing an interpretable and generalized embedding for RL-based control. We further propose a learning-based mapping pipeline that provides fast, uncertainty-aware terrain representations robust to noise and occlusions, serving as policy inputs. It uses neural networks to convert depth observations into local elevations with uncertainties, and fuses them with odometry. The pipeline also integrates with parallel simulation so that we can train controllers with online mapping, aiding sim-to-real transfer. We validate our framework with the proposed mapping pipeline on a quadruped and a biped robot, and the resulting controllers demonstrate strong agility and generalization to unseen terrains in simulation and in real-world experiments.", "links": { "arxiv": "https://arxiv.org/abs/2601.08485", "pdf": "https://arxiv.org/pdf/2601.08485", "doi": "https://doi.org/10.1109/tro.2026.3742243", "project": null, "video": null, "code": null }, "teaser": "assets/teasers/ame2-2026.webp" }, { "id": "beam-search-2026", "title": "An Efficient Beam Search Algorithm for Active Perception in Mobile Robotics", "authors": [ "Kaixian Qu", "Han Wang", "Victor Klemm", "César Cadena", "Marco Hutter" ], "venue": "IJRR", "venueFull": "The International Journal of Robotics Research (IJRR)", "year": 2026, "citations": 1, "theme": "planning", "firstAuthor": false, "featured": false, "tldr": "A node-wise beam search algorithm for active perception that keeps the top-B candidates per node, combined with the rapidly-exploring random annulus graph (RRAG), outperforms state-of-the-art methods by at least 20% in one or more of three representative tasks.", "abstract": "Active perception is a fundamental problem in autonomous robotics in which the robot must decide where to move and what to sense in order to obtain the most informative observations for accomplishing its mission. Existing approaches either solve a computationally expensive traveling salesman problem over heuristically selected informative nodes, or adopt a more efficient but overly constrained shortest path tree formulation. To address these limitations, we explore beam search algorithms as scalable alternatives. While the standard beam search provides scalability by preserving the top-B paths at each depth level, it is prone to local optima and exhibits parameter sensitivity. Our first contribution is a node-wise beam search (NBS) algorithm, which maintains top-B candidates per node to enable more effective exploration of the solution space. Systematic benchmarking on graphs shows that NBS consistently outperforms other baselines and maintains strong performance even at low beam widths. As a second contribution, we integrate the concept of frontiers into the path selection criterion, introducing the expected gain metric, which better balances exploration and exploitation compared to existing alternatives. Our third contribution proposes the rapidly-exploring random annulus graph (RRAG), a novel graph construction method that preserves full orientation sampling and ensures connectivity in cluttered environments through a fallback local sampling-based planner. Extensive experiments demonstrate that NBS combined with RRAG achieves the highest performance across all three representative active perception tasks, outperforming state-of-the-art algorithms by at least 20% in one or more tasks. We further validate the approach on real robotic platforms in different scenarios.", "links": { "arxiv": "https://arxiv.org/abs/2604.23327", "pdf": "https://arxiv.org/pdf/2604.23327", "doi": "https://doi.org/10.1177/02783649261455911", "project": "https://efficient-beam-search.github.io/", "video": null, "code": null }, "teaser": "assets/teasers/beam-search-2026.webp" }, { "id": "inline-skating-2026", "title": "Reinforcement Learning-Based Control for an Inline Skating Humanoid Robot", "authors": [ "Ethan Marot", "Thomas Bi", "Clemens Schwarke", "Victor Klemm", "Marco Hutter", "Raffaello D'Andrea" ], "venue": "IROS", "venueFull": "IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)", "year": 2026, "citations": 0, "theme": "rl", "firstAuthor": false, "featured": false, "tldr": "This paper trains and deploys a reinforcement learning control policy that enables novel locomotion strategies for a humanoid robot modified to equip consumer inline skates instead of conventional feet, and demonstrates up to a 50% reduction in Cost of Transport (CoT) compared to standard walking gaits.", "abstract": "As humanoid robots become increasingly dynamic, coupling them with reinforcement learning offers a promising approach to solving the complex, underactuated mechanics of passive inline skating. Equipping a humanoid robot with passive inline skating wheels presents an opportunity to combine the versatile agility of humanoids with the high-speed, energy-efficient locomotion strategies utilized by human skaters. In this paper, we train and deploy a reinforcement learning control policy that enables novel locomotion strategies for a humanoid robot modified to equip consumer inline skates instead of conventional feet. Unlike previous work limited to quadrupedal robots or actively driven wheels, our system allows for precise 6-DoF control of the skates to execute dynamic, edge-driven propulsion strategies. Our skating strategies emerge entirely from our reward structure, without reliance on human motion data, imitation learning, or kinematic priors. We overcome the inherent instability of passive wheels and simulation contact artifacts by utilizing different geometric wheel models (spherical and ellipsoidal) during training and validation, along with a custom success-based command curriculum and a specialized rolling reward. Consequently, our policy demonstrates up to a 50% reduction in Cost of Transport (CoT) compared to standard walking gaits. The resulting policy successfully transfers zero-shot to the physical Booster T1 hardware. Real-world deployments demonstrate dynamic balance, the ability to reject active physical perturbations, and agile locomotion strategies capable of turning at speed. A video of our results can be found at https://www.youtube.com/watch?v=-_APcOS7uFo.", "links": { "arxiv": "https://arxiv.org/abs/2606.31807", "pdf": "https://arxiv.org/pdf/2606.31807", "doi": null, "project": null, "video": "https://www.youtube.com/watch?v=-_APcOS7uFo", "code": null }, "teaser": "assets/teasers/inline-skating-2026.webp" }, { "id": "voronoi-curriculum-2026", "title": "Bidirectional Voronoi-biased Exploration Curriculum for Reinforcement Learning", "authors": [ "Juri Pfammatter", "Kaixian Qu", "Clemens Schwarke", "Victor Klemm", "Marco Hutter" ], "venue": "arXiv", "venueFull": "arXiv preprint", "year": 2026, "citations": 0, "theme": "rl", "firstAuthor": false, "featured": false, "tldr": "BVER, a bidirectional Voronoi-biased exploration curriculum for goal-conditioned RL inspired by bidirectional RRT planning, grows start states outward from the goal and goals outward from the initial state distribution, biases both toward unexplored task space, and steers them toward each other, training one goal-conditioned policy on both.", "abstract": "Long-horizon tasks with sparse rewards pose an exploration bottleneck for goal-conditioned reinforcement learning: a policy started from the initial state rarely reaches the goal and receives no learning signal. Reference motions, hand-designed curricula, and shaped rewards supply this signal but require demonstrations or task-specific engineering; automatic start-state and goal curricula avoid this but typically expand from one side only, so the full distance to the target must be covered from that side. We propose the Bidirectional Voronoi-biased Exploration curriculum for Reinforcement learning (BVER), which expands from both ends at once. Inspired by bidirectional RRT planning, BVER grows start states outward from the goal and goals outward from the initial state distribution, biases both toward unexplored task space, and steers them toward each other, training one goal-conditioned policy on both. On point-mass mazes, quadrupedal box climbing, and robot-arm ring-on-peg transfer, BVER learns faster than all compared reference-free curricula. On box climbing, it reaches 95% success on a 0.4 m box in roughly 65% fewer iterations than the best of them, is the only one of them to learn to climb a 0.7 m box, and yields a policy robust to start, goal, and yaw variation. Without a demonstration, it approaches the sample efficiency of reference-based curricula on the 0.4 m box and on ring-on-peg transfer. Ablations show that expanding from both ends outperforms either direction alone.", "links": { "arxiv": "https://arxiv.org/abs/2610.03395", "pdf": "https://arxiv.org/pdf/2610.03395", "doi": null, "project": null, "video": null, "code": null }, "teaser": "assets/teasers/voronoi-curriculum-2026.webp" }, { "id": "rubiks-cube-2026", "title": "Solving a Giant Rubik's Cube with a Quadruped Robot", "authors": [ "Lukas Amiet", "Gabriel Sánchez-Blanco Repgen", "Mayank Mittal", "Victor Klemm", "Andrei Cramariuc", "Marco Hutter", "César Cadena" ], "venue": "arXiv", "venueFull": "arXiv preprint", "year": 2026, "citations": 0, "theme": "rl", "firstAuthor": false, "featured": false, "tldr": "A quadruped autonomously solves a giant Rubik's Cube using only its feet, combining onboard RGB-D perception of the cube's pose and color state, a planner that turns the solve into skill sequences, and RL-trained whole-body pedipulation skills; the system completed three autonomous hardware solves.", "abstract": "Pedipulating a freestanding articulated object requires controlling its pose and internal configuration while maintaining balance. We demonstrate this capability by autonomously solving a giant Rubik's Cube with a quadruped using only its feet. The task demands repeated precise interactions, since unintended layer motion can invalidate subsequent operations. We develop an onboard RGB-D perception pipeline that estimates the cube's metric pose and its full color state from partial views acquired through locomotion and flipping. A planner converts the computed solve into robot-executable skill sequences. Learned whole-body skills reorient the cube while preserving its layer configuration and turn its layers with one foot pinning the cube as the other pushes. We train these skills using reinforcement learning with privileged observations and distill them into policies that map onboard depth and proprioception to joint targets. The deployed policies tolerate variations in relative robot-cube placement and recover from external disturbances. The system completes three autonomous hardware solves, each about 30 steps over 8-9 minutes. These results show that combining onboard perception with coordinated pedipulation enables a quadruped to complete extended multi-step physical tasks. Additional material can be found at: https://leggedrobotics.github.io/rubik-skills/", "links": { "arxiv": "https://arxiv.org/abs/2610.24610", "pdf": "https://arxiv.org/pdf/2610.24610", "doi": null, "project": "https://leggedrobotics.github.io/rubik-skills/", "video": null, "code": null }, "teaser": "assets/teasers/rubiks-cube-2026.webp" } ] }