This paper introduces ANYmal, a quadrupedal robot that features outstanding mobility and dynamic motion capability. Thanks to novel, compliant joint modules with integrated electronics, the 30 kg, 0.5 m tall robotic dog is torque controllable very robust against impulsive loads during running or jumping. The presented machine was designed focus on outdoor suitability, simple maintenance, user-friendly handling enable future operation in real world scenarios. Performance tests actuators...

With the introduction of Microsoft Kinect for Windows v2 (Kinect v2), an exciting new sensor is available to robotics and computer vision researchers. Similar original Kinect, capable acquiring accurate depth images at high rates. This useful robot navigation as dense robust maps environment can be created. Opposed working with structured light technology, based on time-of-flight measurement principle might also used outdoors in sunlight. In this paper, we evaluate application mobile...

This paper introduces ANYmal, a quadrupedal robot that features outstanding mobility and dynamic motion capability. Thanks to novel, compliant joint modules with integrated electronics, the 30 kg, 0.5m tall robotic dog is torque controllable very robust against impulsive loads during running or jumping. The presented machine was designed focus on outdoor suitability, simple maintenance, user-friendly handling enable future operation in real world scenarios. Performance tests actuators...

Mobile robots build on accurate, real-time mapping with onboard range sensors to achieve autonomous navigation over rough terrain. Existing approaches often rely absolute localization based tracking of external geometric or visual features. To circumvent the reliability issues these approaches, we propose a novel terrain method, which bases proprioceptive from kinematic and inertial measurements only. The proposed method incorporates drift uncertainties state estimation noise model distance...

Robots working in natural, urban, and industrial settings need to be able navigate challenging environments. In this paper, we present a motion planner for the perceptive rough-terrain locomotion with quadrupedal robots. The finds safe footholds along collision-free swing-leg motions by leveraging an acquired terrain map. To end, novel pose optimization approach that enables robot climb over significant obstacles. We experimentally validate our ANYmal autonomously traversing obstacles such...

This paper presents a framework for planning safe and efficient paths legged robot in rough unstructured terrain. The proposed approach allows to exploit the distinctive obstacle negotiation capabilities of robots, while keeping complexity low enough enable over considerable distances short time. We compute typical terrain characteristics such as slope, roughness, steps build traversability map. map is used assess costs individual footprints function robot-specific negotiating steps, gaps...

Abstract This paper provides insight into the application of quadrupedal robot ANYmal in outdoor missions industrial inspection (autonomous for gas and oil sites [ARGOS] challenge) search rescue (European Robotics League (ERL) Emergency Robots). In both competitions, legged had to autonomously semiautonomously navigate real‐world scenarios complete high‐level tasks such as payload delivery. ARGOS competition, used a rotating light detection ranging sensor localize on site map terrain...

This paper presents a control approach based on whole body framework combined with hierarchical optimization. Locomotion is formulated as multiple tasks (e.g. maintaining balance or tracking desired motion of one the limbs) which are solved in prioritized way using QP solvers. It shown how complex locomotion behaviors can purely emerge from robot-specific inequality (i.e. torque reaching limits) together optimization and system manipulability. Without any specific planning, this task leads...

This paper presents a state estimation approach for legged robots based on stochastic filtering. The key idea is to extract information from the kinematic constraints given through intermittent contacts with ground and fuse this inertial measurements. To end, we design an unscented Kalman filter consistent formulation of underlying model. increase robustness filter, outliers rejection methodology included into update step. Furthermore, present nonlinear observability analysis system, where,...

This work approaches the problem of controlling quadrupedal running and jumping motions with a parametrized, model-based, state-feedback controller.Inspired by motor learning principles observed in nature, our method automatically fine tunes parameters controller repeatedly executing slight variations same motion task.This learn-through-practice process is performed simulation order to best exploit computational resources prevent robot from damaging itself.In ensure that results match...

This paper presents a framework which allows quadrupedal robot to execute dynamic gaits including trot, pace and lateral walk, as well smooth transition between them. Our method relies on an online ZMP based motion planner continuously updates the reference trajectory function of contact schedule state robot. The is coupled with hierarchical whole-body controller optimizes forces by solving cascade prioritized tasks. We tested our ANYmal, fully torque controllable actuated series-elastic actuators.

This paper presents a framework for planning safe and efficient paths legged robot in rough unstructured terrain. The proposed approach allows to exploit the distinctive obstacle negotiation capabilities of robots, while keeping complexity low enough enable over considerable distances short time. We compute typical terrain characteristics such as slope, roughness, steps build traversability map. map is used assess costs individual footprints function robot-specific negotiating steps, gaps...

Quadrupedal locomotion on sloped terrains poses different challenges than walking in a mostly flat environment. The robot's configuration needs to be explicitly controlled order avoid slipping and kinematic limits. To this end, information about the terrain's inclination is required for carefully planning footholds, pose of main body, modulation ground reaction forces. This even more important dynamic trotting, as only two support legs are available compensate gravity drive desired motion....

Legged robots require a robust and fast responding feet contact detection strategy. Common force sensors are often too heavy can be easily damaged during impacts with the terrain. Therefore, it is desirable to detect without sensor. This paper introduces probabilistic strategy which considers full dynamics differential/forward kinematics maximize use of available information for estimation. papers shows that such much more accurate than state-of-the-art only take one measure into account,...

This paper introduces Free Gait, a software framework for the control of robust, versatile, and task-oriented legged robots. In contrast to common hardware abstraction layers, this work focusses on description execution generic whole-body motions (whole-body layer). The motion generation algorithms are connected through Gait API. facilitates development higher level behaviors planning algorithms. API is structured accommodate variety task-space commands. With these, applicable intuitive...

Quadrupedal locomotion on sloped terrains poses different challenges than walking in a mostly flat environment. The robot's configuration needs to be explicitly controlled order avoid slipping and kinematic limits. To this end, information about the terrain's inclination is required for carefully planning footholds, pose of main body, modulation ground reaction forces. This even more important dynamic trotting, as only two support legs are available compensate gravity drive desired motion....

Abstract This paper addresses the issues of unstructured terrain modeling for purpose navigation with legged robots. We present an improved elevation grid concept adopted to specific requirements a small robot limited perceptual capabilities. propose extension update mechanism by incorporating formal treatment spatial uncertainty. Moreover, this presents uncertainty models structured light RGB-D sensor and stereo vision camera used produce dense depth map. The model is based on propagation...

This paper presents the application of reinforcement learning to improve performance highly dynamic single legged locomotion with compliant series elastic actuators. The goal is optimally exploit capabilities hardware in terms maximum jump height, distance, and energy efficiency periodic hopping. These challenges are tackled method Policy Improvement Path Integrals (PI <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ) a model-free...

Collaboration between aerial and ground robots can benefit from exploiting the complementary capabilities of each system, thereby improving situational awareness environment interaction. For this purpose, we present a localization method that allows robot to determine track its position within map acquired by flying robot. To maintain invariance with respect differing sensor choices viewpoints, utilizes elevation maps built independently robot's onboard sensors. The are then used for global...

In this paper we present a method for fusing optical flow and inertial measurements. To end, derive novel visual error term which is better suited than the standard continuous epipolar constraint extracting information contained in By means of an unscented Kalman filter (UKF), then tightly coupled with measurements order to estimate egomotion sensor setup. The individual landmark positions are not part state anymore. Thus, dimensionality space significantly reduced, allowing fast online...

Flying and walking robots can use their complementary features in terms of viewpoint payload capability to the best a heterogeneous team. To this end, we present our online collaborative navigation framework for unknown challenging terrain. The method leverages flying robot's onboard monocular camera create both map visual simultaneous localization mapping dense representation environment as an elevation map. This shared knowledge from platform enables robot localize itself against global...

The dynamics of micro coaxial helicopters are coupled, especially in the presence a stabilizer bar and dynamic maneuvers. This paper presents model-based approach for active decoupling helicopter. allows easier more accurate operation system. nonlinear model covers all degrees freedom attitude altitude. It accounts hover cruise flight situations explicitly captures off-axis bar. A six-axis force/torque sensor an RPM measurement system used custom built test bench. is applied analysis forces...

Collaboration between aerial and ground robots can benefit from exploiting the complementary capabilities of each system, thereby improving situational awareness environment interaction. For this purpose, we present a localization method that allows robot to determine track its position within map acquired by flying robot. To maintain invariance with respect differing sensor choices viewpoints, utilizes elevation maps built independently robot's onboard sensors. The are then used for global...

The development of robots that can dance has received considerable attention. However, they are often either limited to a pre-defined set movements and music or demonstrate little variance when reacting external stimuli, such as microphone camera input. In this paper, we contribute with novel approach allowing legged robot listen live while dancing in synchronization the diverse fashion. This is achieved by extracting beat from an onboard real-time, subsequently creating choreography picking...