Abstract--We present a hierarchical reinforcement learning (RL) pipeline for training one-armed legged robots to perform pick-and-place (P&P) tasks end-to-end--from approaching the payload to releasing it at a target area--in both single-robot and cooperative dual-robot settings. We introduce a novel dynamic reward curriculum that enables a single policy to efficiently learn long-horizon P&P operations by progressively guiding the agents through payload-centered sub-objectives. Compared to state-of-the-art approaches for long-horizon RL tasks, our method improves training efficiency by 55% and reduces execution time by 18.6 % in simulation experiments. In the dual-robot case, we show that our policy enables each robot to attend to different components of its observation space at distinct task stages, promoting effective coordination via autonomous attention shifts. T o our knowledge, this is the first RL pipeline that tackles the full scope of collaborative P&P with two legged manipulators. ANY industries rely on human workers to perform physically demanding tasks such as lifting and transporting heavy loads through cluttered environments. Logistics and construction are among the most prominent examples, where workers engage in monotonous, repetitive pick-and-place (P&P) operations with significant risks of physical injury. While advancements in hardware technology have made it possible for robots to traverse complex environments [19, 25] and dynamically manipulate rigid objects [4], the necessary levels of autonomy and coordination for long-horizon mobile manipulation have yet to be effectively demonstrated in real-world scenarios.

RSL-RL is an open-source Reinforcement Learning library tailored to the specific needs of the robotics community. Unlike broad general-purpose frameworks, its design philosophy prioritizes a compact and easily modifiable codebase, allowing researchers to adapt and extend algorithms with minimal overhead. The library focuses on algorithms most widely adopted in robotics, together with auxiliary techniques that address robotics-specific challenges. Optimized for GPU-only training, RSL-RL achieves high-throughput performance in large-scale simulation environments. Its effectiveness has been validated in both simulation benchmarks and in real-world robotic experiments, demonstrating its utility as a lightweight, extensible, and practical framework to develop learning-based robotic controllers.

Achieving robust locomotion on complex terrains remains a challenge due to high dimensional control and environmental uncertainties. This paper introduces a teacher prior framework based on the teacher student paradigm, integrating imitation and auxiliary task learning to improve learning efficiency and generalization. Unlike traditional paradigms that strongly rely on encoder-based state embeddings, our framework decouples the network design, simplifying the policy network and deployment. A high performance teacher policy is first trained using privileged information to acquire generalizable motion skills. The teacher's motion distribution is transferred to the student policy, which relies only on noisy proprioceptive data, via a generative adversarial mechanism to mitigate performance degradation caused by distributional shifts. Additionally, auxiliary task learning enhances the student policy's feature representation, speeding up convergence and improving adaptability to varying terrains. The framework is validated on a humanoid robot, showing a great improvement in locomotion stability on dynamic terrains and significant reductions in development costs. This work provides a practical solution for deploying robust locomotion strategies in humanoid robots.

Ensuring safe navigation in complex environments requires accurate real-time traversability assessment and understanding of environmental interactions relative to the robot`s capabilities. Traditional methods, which assume simplified dynamics, often require designing and tuning cost functions to safely guide paths or actions toward the goal. This process is tedious, environment-dependent, and not generalizable. To overcome these issues, we propose a novel learned perceptive Forward Dynamics Model (FDM) that predicts the robot`s future state conditioned on the surrounding geometry and history of proprioceptive measurements, proposing a more scalable, safer, and heuristic-free solution. The FDM is trained on multiple years of simulated navigation experience, including high-risk maneuvers, and real-world interactions to incorporate the full system dynamics beyond rigid body simulation. We integrate our perceptive FDM into a zero-shot Model Predictive Path Integral (MPPI) planning framework, leveraging the learned mapping between actions, future states, and failure probability. This allows for optimizing a simplified cost function, eliminating the need for extensive cost-tuning to ensure safety. On the legged robot ANYmal, the proposed perceptive FDM improves the position estimation by on average 41% over competitive baselines, which translates into a 27% higher navigation success rate in rough simulation environments. Moreover, we demonstrate effective sim-to-real transfer and showcase the benefit of training on synthetic and real data. Code and models are made publicly available under https://github.com/leggedrobotics/fdm.

Quadrupedal animals have the ability to perform agile while accurate tasks: a trained dog can chase and catch a flying frisbee before it touches the ground; a cat alone at home can jump and grab the door handle accurately. However, agility and precision are usually a trade-off in robotics problems. Recent works in quadruped robots either focus on agile but not-so-accurate tasks, such as locomotion in challenging terrain, or accurate but not-so-fast tasks, such as using an additional manipulator to interact with objects. In this work, we aim at an accurate and agile task, catching a small object hanging above the robot. We mount a passive gripper in front of the robot chassis, so that the robot has to jump and catch the object with extreme precision. Our experiment shows that our system is able to jump and successfully catch the ball at 1.05m high in simulation and 0.8m high in the real world, while the robot is 0.3m high when standing.

Construction sites are challenging environments for autonomous systems due to their unstructured nature and the presence of dynamic actors, such as workers and machinery. This work presents a comprehensive panoptic scene understanding solution designed to handle the complexities of such environments by integrating 2D panoptic segmentation with 3D LiDAR mapping. Our system generates detailed environmental representations in real-time by combining semantic and geometric data, supported by Kalman Filter-based tracking for dynamic object detection. We introduce a fine-tuning method that adapts large pre-trained panoptic segmentation models for construction site applications using a limited number of domain-specific samples. For this use case, we release a first-of-its-kind dataset of 502 hand-labeled sample images with panoptic annotations from construction sites. In addition, we propose a dynamic panoptic mapping technique that enhances scene understanding in unstructured environments. As a case study, we demonstrate the system's application for autonomous navigation, utilizing real-time RRT* for reactive path planning in dynamic scenarios. The dataset (https://leggedrobotics.github.io/panoptic-scene-understanding.github.io/) and code (https://github.com/leggedrobotics/rsl_panoptic_mapping) for training and deployment are publicly available to support future research.

Abstract--Legged locomotion has recently achieved remarkable success with the progress of machine learning techniques, especially deep reinforcement learning (RL). Controllers employing neural networks have demonstrated empirical and qualitative robustness against real-world uncertainties, including sensor noise and external perturbations. However, formally investigating the vulnerabilities of these locomotion controllers remains a challenge. This difficulty arises from the requirement to pinpoint vulnerabilities across a long-tailed distribution within a highdimensional, temporally sequential space. As a first step towards quantitative verification, we propose a computational method that leverages sequential adversarial attacks to identify weaknesses in learned locomotion controllers. Our research demonstrates that, even state-of-the-art robust controllers can fail significantly under well-designed, low-magnitude adversarial sequence. Through experiments in simulation and on the real robot, we validate our approach's effectiveness, and we illustrate how the results it generates can be used to robustify the original policy and offer valuable insights into the safety of these black-box policies.

We study the problem of mobile manipulation using legged robots equipped with an arm, namely legged loco-manipulation. The robot legs, while usually utilized for mobility, offer an opportunity to amplify the manipulation capabilities by conducting whole-body control. That is, the robot can control the legs and the arm at the same time to extend its workspace. We propose a framework that can conduct the whole-body control autonomously with visual observations. Our approach, namely Visual Whole-Body Control(VBC), is composed of a low-level policy using all degrees of freedom to track the body velocities along with the end-effector position, and a high-level policy proposing the velocities and end-effector position based on visual inputs. We train both levels of policies in simulation and perform Sim2Real transfer for real robot deployment. We perform extensive experiments and show significant improvements over baselines in picking up diverse objects in different configurations (heights, locations, orientations) and environments.

Abstract-- Quadrupedal robots have played a crucial role in various environments, from structured environments to complex harsh terrains, thanks to their agile locomotion ability. However, these robots can easily lose their locomotion functionality if damaged by external accidents or internal malfunctions. The proposed framework consists of three components: 1) a random joint masking strategy for simulating impaired joint scenarios, 2) a joint state estimator to predict an implicit status of current joint condition based on past observation history, and 3) progressive curriculum learning to allow a single network to conduct both normal gait and various joint-impaired gaits. We verify that our framework enables the Unitree's Go1 robot to walk under various impaired joint conditions in realworld indoor and outdoor environments. Recently, quadrupedal robots have played a crucial role in various applications, such as human rescue, disaster response, and harsh terrain exploration [1], [2], [3].

Legged robots navigating cluttered environments must be jointly agile for efficient task execution and safe to avoid collisions with obstacles or humans. Existing studies either develop conservative controllers (< 1.0 m/s) to ensure safety, or focus on agility without considering potentially fatal collisions. This paper introduces Agile But Safe (ABS), a learning-based control framework that enables agile and collision-free locomotion for quadrupedal robots. ABS involves an agile policy to execute agile motor skills amidst obstacles and a recovery policy to prevent failures, collaboratively achieving high-speed and collision-free navigation. The policy switch in ABS is governed by a learned control-theoretic reach-avoid value network, which also guides the recovery policy as an objective function, thereby safeguarding the robot in a closed loop. The training process involves the learning of the agile policy, the reach-avoid value network, the recovery policy, and an exteroception representation network, all in simulation. These trained modules can be directly deployed in the real world with onboard sensing and computation, leading to high-speed and collision-free navigation in confined indoor and outdoor spaces with both static and dynamic obstacles.