Loco-manipulation has recently shown promising capabilities; however, achieving high-precision control, managing the high-dimensional action space induced by many degrees of freedom (DoFs), and fully exploiting the inherent redundancy of whole-body systems remain challenging. In this paper, we propose a novel whole-body control framework that effectively addresses these challenges by decomposing the complex loco-manipulation problem into partial reference motion generation and low-level imitation control. We introduce a new Kinematic Normalizing Flow (KNF) model, trained on a large-scale kinematic dataset, that generates diverse yet feasible partial reference motions. A high-level controller is then trained to navigate the KNF's latent space to exploit redundant solutions, while a low-level controller ensures physically feasible and accurate motion execution. We validate our approach on the quadrupedal robot equipped with a six-DoF robotic arm. In simulation, experimental results show that our approach significantly outperforms state-of-the-art methods in terms of tracking accuracy and feasible workspace coverage. For hardware deployment, we evaluate the system over 24 episodes across 8 different mobile loco-manipulation tasks. The system achieves end-effector pose-tracking errors of 4.5 cm and 0.14 rad, while maintaining accurate locomotion tracking with linear and angular velocity errors of 0.1 m/s and 0.01 rad/s, respectively, outperforming competitive baselines. Our method represents a practical and powerful solution for accurate and generalized whole-body loco-manipulation in high-DoF robotic systems, with promising potential for diverse downstream robotic tasks.

Stable quadrupedal locomotion on sparse terrain requires selecting state-relevant terrain evidence for precise foot placement. Model-based foothold planners provide precise foothold selection but rely heavily on explicit model assumptions. Recent attention-based map encoding (AME) studies show that end-to-end reinforcement learning (RL) can learn implicit foothold guidance. However, the computational cost of dense AME encoding grows with map resolution, limiting its scalability to fine-grained sparse terrain. We propose DELTA, a Deformable Elevation-Based Local Terrain Attention encoder. DELTA predicts state-conditioned sampling locations, forms terrain evidence tokens from adaptive local elevation patches, and attends only to a fixed-size token set. With fixed sampling and patch settings, DELTA's encoder cost is independent of map resolution. Experiments show that DELTA achieves final traversal performance comparable to AME at the standard resolution while improving learning efficiency. This fixed encoder cost enables the use of higher-resolution terrain maps, improving traversal on fine-grained sparse terrain. DELTA also demonstrates strong generalization to unseen mixed evaluation courses composed of continuous and discrete terrain elements. Beyond simulation, DELTA demonstrates successful sim-to-real transfer on RAIBO2. Analysis of the learned sampling offsets and attention weights shows that DELTA samples steppable regions and attends to terrain evidence relevant to future touchdowns without foothold labels or attention supervision.

Legged manipulators extend robotic capabilities beyond static manipulation by integrating agile locomotion with versatile arm control. However, achieving precise manipulation while maintaining coordinated locomotion remains a major challenge. This work presents a hierarchical reinforcement learning framework for dynamic pick-and-place tasks using a quadruped equipped with a 6-DOF robotic arm. The framework incorporates an explicit mass estimation module enabling adaptive whole-body control for objects with varying weights. In simulation, the system achieves an 86.05% success rate with payloads up to 2.3 kg. The approach is further validated through real-world experiments across six representative scenarios with controlled variations in object physical properties (size and mass) and task heights. Specifically, within a wide vertical workspace ranging from ground level to 1.1 m-high tabletops, the system demonstrates an average success rate of 73.3% for payloads up to 1.3 kg, with an average execution time of 4.06 s. Unlike prior works that handle lightweight objects and execute pick-and-place motions with slow, piecewise motions, the proposed framework exploits concurrent locomotion and manipulation for dynamic, continuous execution. These results demonstrate the potential of quadrupedal mobile manipulators for adaptive, whole-body pick-and-place with heavier payloads and extended workspaces.

The simulation-to-reality (sim-to-real) transfer of large-scale hydraulic robots presents a significant challenge in robotics because of the inherent slow control response and complex fluid dynamics. The complex dynamics result from the multiple interconnected cylinder structure and the difference in fluid rates of the cylinders. These characteristics complicate detailed simulation for all joints, making it unsuitable for reinforcement learning (RL) applications. In this work, we propose an analytical actuator model driven by hydraulic dynamics to represent the complicated actuators. The model predicts joint torques for all 12 actuators in under 1 microsecond, allowing rapid processing in RL environments. We compare our model with neural network-based actuator models and demonstrate the advantages of our model in data-limited scenarios. The locomotion policy trained in RL with our model is deployed on a hydraulic quadruped robot, which is over 300 kg. This work is the first demonstration of a successful transfer of stable and robust command-tracking locomotion with RL on a heavy hydraulic quadruped robot, demonstrating advanced sim-to-real transferability.

In this paper, we present TRACE (Tokenized Robust Attention for Contact-Aware Estimation), an end-to-end learned proprioceptive odometry estimator for legged robots under unreliable contact conditions. The proposed estimator directly predicts relative displacement, relative rotation, and body-frame velocity from a recent history of onboard inertial and joint measurements. To improve robustness under unreliable contact conditions, we introduce a foot-aware cross-attention module that adaptively weights IMU and leg-wise kinematic tokens without relying on manually defined contact or slip thresholds. The estimator is trained with direct supervision and two physics-inspired auxiliary losses that promote kinematic consistency and reliable use of leg information. To reduce policy-specific overfitting and consequently improve sim-to-real transfer, simulation training incorporates policy randomization, followed by partial real-world fine-tuning of the temporal encoder and prediction head. Experiments across diverse indoor and outdoor terrains demonstrate consistent reductions in position drift compared with classical filtering-based, hybrid, and purely learning-based baselines. Ablation studies further validate the contributions of the proposed training objectives, policy randomization, and real-world fine-tuning, particularly under unreliable contacts and sim-to-real mismatch.

High-speed legged navigation in discrete and geometrically complex environments is a challenging task because of the high–degree-of-freedom dynamics and long-horizon, nonconvex nature of the optimization problem. In this work, we propose a hierarchical navigation pipeline for legged robots that can traverse such environments at high speed. The proposed pipeline consists of a planner and tracker module. The planner module finds physically feasible foothold plans by sampling-based optimization with fast sequential filtering using heuristics and a neural network. Subsequently, rollouts are performed in a physics simulation to identify the best foothold plan regarding the engineered cost function and to confirm its physical consistency. This hierarchical planning module is computationally efficient and physically accurate at the same time. The tracker aims to accurately step on the target footholds from the planning module. During the training stage, the foothold target distribution is given by a generative model that is trained competitively with the tracker. This process ensures that the tracker is trained in an environment with the desired difficulty. The resulting tracker can overcome terrains that are more difficult than what the previous methods could manage. We demonstrated our approach using Raibo, our in-house dynamic quadruped robot. The results were dynamic and agile motions: Raibo is capable of running on vertical walls, jumping a 1.3-meter gap, running over stepping stones at 4 meters per second, and autonomously navigating on terrains full of 30° ramps, stairs, and boxes of various sizes.

This paper introduces a novel proprioceptive state estimator for legged robots that combines model-based filters with deep neural networks. In environments where vision systems are not reliable, proprioceptive state estimators become indispensable. Traditionally, proprioceptive state estimators are based on model-based approaches, which rely solely on contact foot kinematics as measurements. In contrast, learning-based approaches have obtained new measurements, such as displacement and covariance, by leveraging real-world data in a supervised manner. In this work, we develop a state estimation framework that trains a neural measurement network (NMN) to estimate the base's linear velocity and foot contact probability, which are then employed as measurements in an invariant extended Kalman filter. Our approach relies solely on simulation data for training, as it allows us to obtain extensive data easily. We address the sim-to-real gap by adapting existing learning techniques and regularization. To validate our proposed method, we conduct hardware experiments using a quadruped robot on four types of terrain: flat, debris, soft, and slippery. In our experiments, the proposed method demonstrates significant improvements over the model-based state estimator, achieving an average reduction in Absolute Trajectory Error (ATE) by 61.8% for position and 8.5% for velocity.

For reliable autonomous robot navigation in urban settings, the robot must have the ability to identify semantically traversable terrains in the image based on the semantic understanding of the scene. This reasoning ability is based on semantic traversability, which is frequently achieved using semantic segmentation models fine-tuned on the testing domain. This fine-tuning process often involves manual data collection with the target robot and annotation by human labelers which is prohibitively expensive and unscalable. In this work, we present an effective methodology for training a semantic traversability estimator using egocentric videos and an automated annotation process. Egocentric videos are collected from a camera mounted on a pedestrian's chest. The dataset for training the semantic traversability estimator is then automatically generated by extracting semantically traversable regions in each video frame using a recent foundation model in image segmentation and its prompting technique. Extensive experiments with videos taken across several countries and cities, covering diverse urban scenarios, demonstrate the high scalability and generalizability of the proposed annotation method. Furthermore, performance analysis and real-world deployment for autonomous robot navigation showcase that the trained semantic traversability estimator is highly accurate, able to handle diverse camera viewpoints, computationally light, and real-world applicable. The summary video is available at https://youtu.be/EUVoH-wA-lA .

In this research, we investigated the innovative use of a manipulator as a tail in quadruped robots to augment their physical capabilities. Previous studies have primarily focused on enhancing various abilities by attaching robotic tails that function solely as tails on quadruped robots. While these tails improve the performance of the robots, they come with several disadvantages, such as increased overall weight and higher costs. To mitigate these limitations, we propose the use of a 6-DoF manipulator as a tail, allowing it to serve both as a tail and as a manipulator. To control this highly complex robot, we developed a controller based on reinforcement learning for the robot equipped with the manipulator. Our experimental results demonstrate that robots equipped with a manipulator outperform those without a manipulator in tasks such as rapid turning, aerial reorientation, and balancing. These results indicate that the manipulator can improve the agility and stability of quadruped robots, similar to a tail, in addition to its manipulation capabilities.

This article presents Platform Adaptive Locomotion (PAL), a unified control method for quadrupedal robots with different morphologies and dynamics. We leverage deep reinforcement learning to train a single locomotion policy on procedurally generated robots. The policy maps proprioceptive robot state information and base velocity commands into desired joint actuation targets, which are conditioned using a latent embedding of the temporally local system dynamics. We explore two conditioning strategies - one using a GRU-based dynamics encoder and another using a morphology-based property estimator - and show that morphology-aware conditioning outperforms temporal dynamics encoding regarding velocity task tracking for our hardware test on ANYmal C. Our results demonstrate that both approaches achieve robust zero-shot transfer across multiple unseen simulated quadrupeds. Furthermore, we demonstrate the need for careful robot reference modelling during training: exposing the policy to a diverse set of robot morphologies and dynamics leads to improved generalization, reducing the velocity tracking error by up to 30% compared to the baseline method. Despite PAL not surpassing the best-performing reference-free controller in all cases, our analysis uncovers critical design choices and informs improvements to the state of the art.

Robots consisting of many articulated parts performing complex movements are challenging to design. We showcase an interactive system for exploring shapes and structures of robots through 3D sketching, generating plausible movements of robots through AI, and reviewing and refining them in VR. Such immersive prototyping in the early stages can help reduce the time and cost associated with trial and error in later stages, contributing to shortening and streamlining of the robot development process.

Several earlier studies have shown impressive control performance in complex robotic systems by designing the controller using a neural network and training it with model-free reinforcement learning. However, these outstanding controllers with natural motion style and high task performance are developed through extensive reward engineering, which is a highly laborious and time-consuming process of designing numerous reward terms and determining suitable reward coefficients. In this work, we propose a novel reinforcement learning framework for training neural network controllers for complex robotic systems consisting of both rewards and constraints. To let the engineers appropriately reflect their intent to constraints and handle them with minimal computation overhead, two constraint types and an efficient policy optimization algorithm are suggested. The learning framework is applied to train locomotion controllers for several legged robots with different morphology and physical attributes to traverse challenging terrains. Extensive simulation and real-world experiments demonstrate that performant controllers can be trained with significantly less reward engineering, by tuning only a single reward coefficient. Furthermore, a more straightforward and intuitive engineering process can be utilized, thanks to the interpretability and generalizability of constraints. The summary video is available at https://youtu.be/KAlm3yskhvM.

We propose a learning-based system for enabling quadrupedal robots to manipulate large, heavy objects using their whole body. Our system is based on a hierarchical control strategy that uses the deep latent variable embedding which captures manipulation-relevant information from interactions, proprioception, and action history, allowing the robot to implicitly understand object properties. We evaluate our framework in both simulation and real-world scenarios. In the simulation, it achieves a success rate of 93.6 $\%$ in accurately re-positioning and re-orienting various objects within a tolerance of 0.03 $\text{m}$ and 5 $^\circ$ . Real-world experiments demonstrate the successful manipulation of objects such as a 19.2 $\text{kg}$ water-filled drum and a 15.3 $\text{kg}$ plastic box filled with heavy objects while the robot weighs 27 $\text{kg}$ . Unlike previous works that focus on manipulating small and light objects using prehensile manipulation, our framework illustrates the possibility of using quadrupeds for manipulating large and heavy objects that are ungraspable with the robot's entire body. Our method does not require explicit object modeling and offers significant computational efficiency compared to optimization-based methods.

In the realm of autonomous mobile robots, safe navigation through unpaved outdoor environments remains a challenging task. Due to the high-dimensional nature of sensor data, extracting relevant information becomes a complex problem, which hinders adequate perception and path planning. Previous works have shown promising performances in extracting global features from full-sized images. However, they often face challenges in capturing essential local information. In this letter, we propose Crop-LSTM, which iteratively takes cropped image patches around the current robot's position and predicts the future position, orientation, and bumpiness. Our method performs local feature extraction by paying attention to corresponding image patches along the predicted robot trajectory in the 2D image plane. This enables more accurate predictions of the robot's future trajectory. With our wheeled mobile robot platform Raicart, we demonstrated the effectiveness of Crop-LSTM for point-goal navigation in an unpaved outdoor environment. Our method enabled safe and robust navigation using RGBD images in challenging unpaved outdoor terrains.

We present ArtiGrasp, a novel method to synthesize bi-manual hand-object interactions that include grasping and articulation. This task is challenging due to the diversity of the global wrist motions and the precise finger control that are necessary to articulate objects. ArtiGrasp leverages reinforcement learning and physics simulations to train a policy that controls the global and local hand pose. Our framework unifies grasping and articulation within a single policy guided by a single hand pose reference. Moreover, to facilitate the training of the precise finger control required for articulation, we present a learning curriculum with increasing difficulty. It starts with single-hand manipulation of stationary objects and continues with multi-agent training including both hands and non-stationary objects. To evaluate our method, we introduce Dynamic Object Grasping and Articulation, a task that involves bringing an object into a target articulated pose. This task requires grasping, relocation, and articulation. We show our method's efficacy towards this task. We further demonstrate that our method can generate motions with noisy hand-object pose estimates from an off-the-shelf image-based regressor.

Control of legged robots is a challenging problem that has been investigated by different approaches, such as model-based control and learning algorithms. This work proposes a novel Imitating and Finetuning Model Predictive Control (IFM) framework to take the strengths of both approaches. Our framework first develops a conventional model predictive controller (MPC) using Differential Dynamic Programming and Raibert heuristic, which serves as an expert policy. Then we train a clone of the MPC using imitation learning to make the controller learnable. Finally, we leverage deep reinforcement learning with limited exploration for further finetuning the policy on more challenging terrains. By conducting comprehensive simulation and hardware experiments, we demonstrate that the proposed IFM framework can significantly improve the performance of the given MPC controller on rough, slippery, and conveyor terrains that require careful coordination of footsteps. We also showcase that IFM can efficiently produce more symmetric, periodic, and energy-efficient gaits compared to Vanilla RL with a minimal burden of reward shaping.

Simulation-based reinforcement learning approaches are leading the next innovations in legged robot control. However, the resulting control policies are still not applicable on soft and deformable terrains, especially at high speed. The primary reason is that reinforcement learning approaches, in general, are not effective beyond the data distribution: The agent cannot perform well in environments that it has not experienced. To this end, we introduce a versatile and computationally efficient granular media model for reinforcement learning. Our model can be parameterized to represent diverse types of terrain from very soft beach sand to hard asphalt. In addition, we introduce an adaptive control architecture that can implicitly identify the terrain properties as the robot feels the terrain. The identified parameters are then used to boost the locomotion performance of the legged robot. We applied our techniques to the Raibo robot, a dynamic quadrupedal robot developed in-house. The trained networks demonstrated high-speed locomotion capabilities on deformable terrains: The robot was able to run on soft beach sand at 3.03 meters per second although the feet were completely buried in the sand during the stance phase. We also demonstrate its ability to generalize to different terrains by presenting running experiments on vinyl tile flooring, athletic track, grass, and a soft air mattress.

Many studies on quadrupedal manipulators have been conducted for extending the workspace of the end-effector. Many of these studies, especially the recent ones, use model-based control for the arm and learning-based control for the leg. Some studies solely focused on model-based control for controlling both the base and arm. However, model-based controllers such as MPC can be computationally inefficient when there are many contacts between the end-effector and the object. The dynamics of the interactions between a quadrupedal manipulator and the object in contact are complex and often unpredictable without high-resolution contact sensors on the end-effector. In this study, we investigate the possibility of using a reinforcement learning strategy to control an end-effector of a legged mobile manipulator. The proposed framework is verified for a walking and tracking task of the end-effector in a simulation environment.

The energy efficiency theory of human bipedal locomotion has been widely accepted as a neuro-musculoskeletal control method. However, coactivation of agonist and antagonist muscles in the lower limb has been observed during various limb movements, including walking. The emergence of this coactivation cannot be explained solely by the energy efficiency theory and remains a subject of debate. To shed light on this, we investigated the role of muscle coactivations in walking stability using a forward dynamics musculoskeletal simulation combined with neural-network-based gait controllers. Our study revealed that a gait controller with minimal muscle activations had a high probability of falls under challenging gait conditions such as slippery ground and uneven terrain. Lower limb muscle coactivations emerged in the process of gait controller training on slippery ground. Controllers with physiological coactivation levels demonstrated a significantly reduced probability of falls. Our results suggest that achieving stable walking requires muscle coactivations beyond the minimal level of muscle energy. This study implies that coactivations likely emerge to maintain gait stability under challenging conditions, and both coactivation and energy optimization of lower limb muscles should be considered when exploring the foundational control mechanisms of human walking.

Legged robots that can operate autonomously in remote and hazardous environments will greatly increase opportunities for exploration into underexplored areas. Exteroceptive perception is crucial for fast and energy-efficient locomotion: Perceiving the terrain before making contact with it enables planning and adaptation of the gait ahead of time to maintain speed and stability. However, using exteroceptive perception robustly for locomotion has remained a grand challenge in robotics. Snow, vegetation, and water visually appear as obstacles on which the robot cannot step or are missing altogether due to high reflectance. In addition, depth perception can degrade due to difficult lighting, dust, fog, reflective or transparent surfaces, sensor occlusion, and more. For this reason, the most robust and general solutions to legged locomotion to date rely solely on proprioception. This severely limits locomotion speed because the robot has to physically feel out the terrain before adapting its gait accordingly. Here, we present a robust and general solution to integrating exteroceptive and proprioceptive perception for legged locomotion. We leverage an attention-based recurrent encoder that integrates proprioceptive and exteroceptive input. The encoder is trained end to end and learns to seamlessly combine the different perception modalities without resorting to heuristics. The result is a legged locomotion controller with high robustness and speed. The controller was tested in a variety of challenging natural and urban environments over multiple seasons and completed an hour-long hike in the Alps in the time recommended for human hikers.