A5015540714- Robotic Locomotion and Control
- Robotics and Sensor-Based Localization
- Robotic Path Planning Algorithms
- Robotics and Automated Systems
- Modular Robots and Swarm Intelligence
- Anomaly Detection Techniques and Applications
- Advanced Neural Network Applications
- Robot Manipulation and Learning
- Planetary Science and Exploration
- Wildlife-Road Interactions and Conservation
- Video Surveillance and Tracking Methods
- Reinforcement Learning in Robotics
- 3D Surveying and Cultural Heritage
- Autonomous Vehicle Technology and Safety
- Soil Mechanics and Vehicle Dynamics
- Remote Sensing and LiDAR Applications
- Biomimetic flight and propulsion mechanisms
- Hand Gesture Recognition Systems
- Adversarial Robustness in Machine Learning
- Smart Agriculture and AI
- Intelligent Tutoring Systems and Adaptive Learning
- Data Management and Algorithms
- Advanced Manufacturing and Logistics Optimization
- Geophysical Methods and Applications
- Bayesian Modeling and Causal Inference
ETH Zurich
2016-2024
Robotic Research (United States)
2022-2024
École Polytechnique Fédérale de Lausanne
2022
Robotics Research (United States)
2019
Some of the most challenging environments on our planet are accessible to quadrupedal animals but remain out reach for autonomous machines. Legged locomotion can dramatically expand operational domains robotics. However, conventional controllers legged based elaborate state machines that explicitly trigger execution motion primitives and reflexes. These designs have escalated in complexity while falling short generality robustness animal locomotion. Here we present a radically robust...
Legged robots that can operate autonomously in remote and hazardous environments will greatly increase opportunities for exploration into under-explored areas. Exteroceptive perception is crucial fast energy-efficient locomotion: perceiving the terrain before making contact with it enables planning adaptation of gait ahead time to maintain speed stability. However, utilizing exteroceptive robustly locomotion has remained a grand challenge robotics. Snow, vegetation, water visually appear as...
Legged robots have the potential to traverse diverse and rugged terrain. To find a safe efficient navigation path carefully select individual footholds, it is useful be able predict properties of terrain ahead robot. In this letter, we propose method collect data from robot-terrain interaction associate images. Using sparse acquired in teleoperation experiments with quadrupedal robot, train neural network generate dense prediction front training data, project foothold positions robot...
Autonomous exploration of subterranean environments constitutes a major frontier for robotic systems, as underground settings present key challenges that can render robot autonomy hard to achieve. This problem has motivated the DARPA Subterranean Challenge, where teams robots search objects interest in various environments. In response, we CERBERUS system-of-systems, unified strategy using legged and flying robots. Our proposed approach relies on ANYmal quadraped primary robots, exploiting...
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...
Perceiving the surrounding environment is crucial for autonomous mobile robots. An elevation map provides a memory-efficient and simple yet powerful geometric represen-tation of terrain ground The robots can use this information navigation in an unknown or perceptive locomotion control over rough terrain. Depending on application, various post processing steps may be incorpo-rated, such as smoothing, inpainting plane segmentation. In work, we present mapping pipeline leveraging GPU fast...
Autonomous wheeled-legged robots have the potential to transform logistics systems, improving operational efficiency and adaptability in urban environments. Navigating environments, however, poses unique challenges for robots, necessitating innovative solutions locomotion navigation. These include need adaptive across varied terrains ability navigate efficiently around complex dynamic obstacles. This work introduces a fully integrated system comprising control, mobility-aware local...
Planetary exploration robots encounter challenging terrain during operation. Vision-based approaches have failed to reliably predict soil characteristics in the past, making it necessary probe tactilely. We present a robust, haptic inspection approach for variety of fine, granular media, which are representative Martian soil. In our approach, robot uses one limb perform an impact trajectory, while supporting main body with remaining three legs. The resulting vibration, is recorded by sensors...
Abstract Legged robots are exceedingly versatile and have the potential to navigate complex, confined spaces due their many degrees of freedom. As a result computational complexity, there exist no online planners for perceptive whole‐body locomotion in tight spaces. In this paper, we present new method planning multilegged robots, which generates body poses, footholds, swing trajectories collision avoidance. Measurements from an onboard depth camera used create three‐dimensional map terrain...
In this work, we present a learning-based pipeline to realise local navigation with quadrupedal robot in cluttered environments static and dynamic obstacles.Given high-level commands, the is able safely locomote target location based on frames from depth camera without any explicit mapping of environment.First, sequence images current trajectory are fused form model world using state representation learning.The output lightweight module then directly fed into target-reaching...
Legged robots have the ability to adapt their walking posture navigate confined spaces due high degrees of freedom. However, this has not been exploited in most common multilegged platforms. This paper presents a deformable bounding box abstraction robot model, with accompanying mapping and planning strategies, that enable legged autonomously change its body shape spaces. The is achieved using robot-centric multi-elevation maps generated distance sensors carried by robot. path based on...
Navigation planning for legged robots has distinct challenges compared to wheeled and tracked systems due the ability lift legs off ground step over obstacles. While most navigation planners assume a fixed traversability value single terrain patch, we overcome this limitation by proposing reachability-based planner robots. We approximate robot morphology set of reachability body volumes, assuming that volumes need always be in contact with environment, while should contact-free. train...
Mobile robots are becoming very popular in real-world outdoors applications, where there many challenges robot control and perception. One of the most critical problems is to characterise terrain traversed by robot. This knowledge indispensable for optimal negotiation. Currently, approaches performing classification from vision, but not enough research on identification a direct interaction with environment. In our work, we proposed new methods force/torque data an legged foot ground,...
Navigating off-road with a fast autonomous vehicle depends on robust perception system that differentiates traversable from non-traversable terrain. Typically, this semantic understanding which is based supervised learning images annotated by human expert. This requires significant investment in time, assumes correct expert classification, and small details can lead to misclassification. To address these challenges, we propose method for predicting high- low-risk terrains only past...
This article presents the CERBERUS robotic system-of-systems, which won DARPA Subterranean Challenge Final Event in 2021. The was organized by with vision to facilitate novel technologies necessary reliably explore diverse underground environments despite grueling challenges they present for autonomy. Due their geometric complexity, degraded perceptual conditions combined lack of GNSS support, austere navigation conditions, and denied communications, subterranean settings render autonomous...
Navigation on challenging terrain topographies requires the understanding of robots' locomotion capabilities to produce optimal solutions. We present an integrated framework for real-time autonomous navigation mobile robots based elevation maps. The performs rapid global path planning and optimization that is aware robot. A GPU-aided, sampling-based planner combined with a gradient-based optimizer provides paths by using neural network-based cost predictor which trained in simulation. show...
Due to the highly complex environment present during DARPA Subterranean Challenge, all six funded teams relied on legged robots as part of their robotic team. Their unique locomotion skills being able step over obstacles require special consideration for navigation planning. In this work, we and examine ArtPlanner, planner used by team CERBERUS Finals. It is based a sampling-based method that determines valid poses with reachability abstraction uses learned foothold scores restrict areas...
This article presents the CERBERUS robotic system-of-systems, which won DARPA Subterranean Challenge Final Event in 2021. The was organized by with vision to facilitate novel technologies necessary reliably explore diverse underground environments despite grueling challenges they present for autonomy. Due their geometric complexity, degraded perceptual conditions combined lack of GPS support, austere navigation conditions, and denied communications, subterranean settings render autonomous...
Navigation in natural outdoor environments requires a robust and reliable traversability classification method to handle the plethora of situations robot can encounter. Binary algorithms perform well their native domain but tend provide overconfident predictions when presented with out-of-distribution samples, which lead catastrophic failure navigating unknown environments. We propose overcome this issue by using anomaly detection on multi-modal images for classification, is easily scalable...
A common scenario in Search and Rescue robotics is to map patrol a disaster site assess the situation plan potential missions of rescue teams. Particular importance has be given changes environment as these may correspond critical events like building collapses, movement objects, etc. This paper presents change detection pipeline for LiDAR-equipped robots assist humans detecting those changes. The local 3D point cloud data compared an octree-based occupancy representation by computing...
The high agility of legged systems allows them to operate in rugged outdoor environments. In these situations, knowledge about the terrain geometry is key for foothold planning enable safe locomotion. However, on penetrable or highly compliant (e.g. grass) visibility supporting ground surface obstructed, i.e. it cannot directly be perceived by depth sensors. We present a method estimate underlying topography fusing haptic information foot contact closure locations with exteroceptive sensing....
A common scenario in Search and Rescue robotics is to map patrol a disaster site assess the situation plan potential missions of rescue teams. Particular importance has be given changes environment as these may correspond critical events like building collapses, movement objects, etc. This paper presents change detection pipeline for LiDAR-equipped robots assist humans detecting those changes. The local 3D point cloud data compared an octree-based occupancy representation by computing...