Journal Articles
Nisticò, Ylenia; Fahmi, Shamel; Pallottino, Lucia; Semini, Claudio; Fink, Geoff
On Slip Detection for Quadruped Robots Journal Article
In: 2022.
@article{NisFahPal22,
title = {On Slip Detection for Quadruped Robots},
author = {Ylenia Nistic\`{o} and Shamel Fahmi and Lucia Pallottino and Claudio Semini and Geoff Fink},
editor = {Salvatore Pirozzi},
url = {https://iit-dlslab.github.io/papers/nistico22sensors.pdf
https://perro.tru.ca/wp-content/uploads/2024/01/nistico22sensors.pdf
},
year = {2022},
date = {2022-04-13},
urldate = {2022-04-13},
abstract = {Legged robots are meant to autonomously navigate unstructured environments for applications like search and rescue, inspection, or maintenance. In autonomous navigation, a close relationship between locomotion and perception is crucial; the robot has to perceive the environment and detect any change in order to autonomously make decisions based on what it perceived. One main challenge in autonomous navigation for legged robots is locomotion over unstructured terrains. In particular, when the ground is slippery, common control techniques and state estimation algorithms may not be effective, because the ground is commonly assumed to be non-slippery. This paper addresses the problem of slip detection, a first fundamental step to implement appropriate control strategies and perform dynamic whole-body locomotion. We propose a slip detection approach, which is independent of the gait type and the estimation of the position and velocity of the robot in an inertial frame, that is usually prone to drift problems. To the best of our knowledge, this is the first approach of a quadruped robot slip detector that can detect more than one foot slippage at the same time, relying on the estimation of measurements expressed in a non-inertial frame. We validate the approach on the 90 kg Hydraulically actuated Quadruped robot (HyQ) from the Istituto Italiano di Tecnologia (IIT), and we compare it against a state-of-the-art slip detection algorithm.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Legged robots are meant to autonomously navigate unstructured environments for applications like search and rescue, inspection, or maintenance. In autonomous navigation, a close relationship between locomotion and perception is crucial; the robot has to perceive the environment and detect any change in order to autonomously make decisions based on what it perceived. One main challenge in autonomous navigation for legged robots is locomotion over unstructured terrains. In particular, when the ground is slippery, common control techniques and state estimation algorithms may not be effective, because the ground is commonly assumed to be non-slippery. This paper addresses the problem of slip detection, a first fundamental step to implement appropriate control strategies and perform dynamic whole-body locomotion. We propose a slip detection approach, which is independent of the gait type and the estimation of the position and velocity of the robot in an inertial frame, that is usually prone to drift problems. To the best of our knowledge, this is the first approach of a quadruped robot slip detector that can detect more than one foot slippage at the same time, relying on the estimation of measurements expressed in a non-inertial frame. We validate the approach on the 90 kg Hydraulically actuated Quadruped robot (HyQ) from the Istituto Italiano di Tecnologia (IIT), and we compare it against a state-of-the-art slip detection algorithm.
Fahmi, Shamel; Fink, Geoff; and Claudio Semini,
On State Estimation for Legged Locomotion over Soft Terrain Journal Article
In: IEEE Sensors Letters, 2021.
@article{nokey,
title = {On State Estimation for Legged Locomotion over Soft Terrain},
author = {Shamel Fahmi and Geoff Fink and and Claudio Semini},
url = {https://iit-dlslab.github.io/papers/fahmi21lsens.pdf},
year = {2021},
date = {2021-01-00},
urldate = {2021-01-00},
journal = {IEEE Sensors Letters},
abstract = {Locomotion over soft terrain remains a challenging problem for legged robots. Most of the work done on state
estimation for legged robots is designed for rigid contacts, and does not take into account the physical parameters of the
terrain. That said, this letter answers the following questions: how and why does soft terrain affect state estimation for
legged robots? To do so, we utilized a state estimator that fuses IMU measurements with leg odometry that is designed
with rigid contact assumptions. We experimentally validated the state estimator with the HyQ robot trotting over both soft
and rigid terrain. We demonstrate that soft terrain negatively affects state estimation for legged robots, and that the state
estimates have a noticeable drift over soft terrain compared to rigid terrain.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Locomotion over soft terrain remains a challenging problem for legged robots. Most of the work done on state
estimation for legged robots is designed for rigid contacts, and does not take into account the physical parameters of the
terrain. That said, this letter answers the following questions: how and why does soft terrain affect state estimation for
legged robots? To do so, we utilized a state estimator that fuses IMU measurements with leg odometry that is designed
with rigid contact assumptions. We experimentally validated the state estimator with the HyQ robot trotting over both soft
and rigid terrain. We demonstrate that soft terrain negatively affects state estimation for legged robots, and that the state
estimates have a noticeable drift over soft terrain compared to rigid terrain.
estimation for legged robots is designed for rigid contacts, and does not take into account the physical parameters of the
terrain. That said, this letter answers the following questions: how and why does soft terrain affect state estimation for
legged robots? To do so, we utilized a state estimator that fuses IMU measurements with leg odometry that is designed
with rigid contact assumptions. We experimentally validated the state estimator with the HyQ robot trotting over both soft
and rigid terrain. We demonstrate that soft terrain negatively affects state estimation for legged robots, and that the state
estimates have a noticeable drift over soft terrain compared to rigid terrain.
Fahmi, Shamel; Focchi, Michele; Radulescu, Andreea; Fink, Geoff; Barasuol, Victor; Semini, Claudio
Stance: Locomotion adaptation over soft terrain Journal Article
In: IEEE Transactions on Robotics, vol. 36, no. 2, pp. 443–457, 2020.
@article{fahmi2020stance,
title = {Stance: Locomotion adaptation over soft terrain},
author = {Shamel Fahmi and Michele Focchi and Andreea Radulescu and Geoff Fink and Victor Barasuol and Claudio Semini},
url = {https://iit-dlslab.github.io/papers/fahmi19tro.pdf},
year = {2020},
date = {2020-01-01},
urldate = {2020-01-01},
journal = {IEEE Transactions on Robotics},
volume = {36},
number = {2},
pages = {443\textendash457},
publisher = {IEEE},
abstract = {Whole-Body Control (WBC) has emerged as an
important framework in locomotion control for legged robots.
However, most WBC frameworks fail to generalize beyond rigid
terrains. Legged locomotion over soft terrain is difficult due to
the presence of unmodeled contact dynamics that WBCs do not
account for. This introduces uncertainty in locomotion and affects
the stability and performance of the system. In this paper, we propose a novel soft terrain adaptation algorithm called STANCE:
Soft Terrain Adaptation and Compliance Estimation. STANCE
consists of a WBC that exploits the knowledge of the terrain to
generate an optimal solution that is contact consistent and an
online terrain compliance estimator that provides the WBC with
terrain knowledge. We validated STANCE both in simulation
and experiment on the Hydraulically actuated Quadruped (HyQ)
robot, and we compared it against the state of the art WBC. We
demonstrated the capabilities of STANCE with multiple terrains
of different compliances, aggressive maneuvers, different forward
velocities, and external disturbances. STANCE allowed HyQ to
adapt online to terrains with different compliances (rigid and
soft) without pre-tuning. HyQ was able to successfully deal with
the transition between different terrains and showed the ability
to differentiate between compliances under each foo},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Whole-Body Control (WBC) has emerged as an
important framework in locomotion control for legged robots.
However, most WBC frameworks fail to generalize beyond rigid
terrains. Legged locomotion over soft terrain is difficult due to
the presence of unmodeled contact dynamics that WBCs do not
account for. This introduces uncertainty in locomotion and affects
the stability and performance of the system. In this paper, we propose a novel soft terrain adaptation algorithm called STANCE:
Soft Terrain Adaptation and Compliance Estimation. STANCE
consists of a WBC that exploits the knowledge of the terrain to
generate an optimal solution that is contact consistent and an
online terrain compliance estimator that provides the WBC with
terrain knowledge. We validated STANCE both in simulation
and experiment on the Hydraulically actuated Quadruped (HyQ)
robot, and we compared it against the state of the art WBC. We
demonstrated the capabilities of STANCE with multiple terrains
of different compliances, aggressive maneuvers, different forward
velocities, and external disturbances. STANCE allowed HyQ to
adapt online to terrains with different compliances (rigid and
soft) without pre-tuning. HyQ was able to successfully deal with
the transition between different terrains and showed the ability
to differentiate between compliances under each foo
important framework in locomotion control for legged robots.
However, most WBC frameworks fail to generalize beyond rigid
terrains. Legged locomotion over soft terrain is difficult due to
the presence of unmodeled contact dynamics that WBCs do not
account for. This introduces uncertainty in locomotion and affects
the stability and performance of the system. In this paper, we propose a novel soft terrain adaptation algorithm called STANCE:
Soft Terrain Adaptation and Compliance Estimation. STANCE
consists of a WBC that exploits the knowledge of the terrain to
generate an optimal solution that is contact consistent and an
online terrain compliance estimator that provides the WBC with
terrain knowledge. We validated STANCE both in simulation
and experiment on the Hydraulically actuated Quadruped (HyQ)
robot, and we compared it against the state of the art WBC. We
demonstrated the capabilities of STANCE with multiple terrains
of different compliances, aggressive maneuvers, different forward
velocities, and external disturbances. STANCE allowed HyQ to
adapt online to terrains with different compliances (rigid and
soft) without pre-tuning. HyQ was able to successfully deal with
the transition between different terrains and showed the ability
to differentiate between compliances under each foo
Proceedings Articles
Dettmann, Alexander; Planthaber, Steffen; Bargsten, Vinzenz; Dominguez, Raul; Cerilli, Gianluca; Marchitto, Marco; Fink, Geoff; Focchi, Michele; Barasuol, Victor; Semini, Claudio; others,
Towards a Generic Navigation and Locomotion Control System for Legged Space Exploration Proceedings Article
In: 16th Symposium on Advanced Space Technologies in Robotics and Automation, 2022.
@inproceedings{dettmann2022towards,
title = {Towards a Generic Navigation and Locomotion Control System for Legged Space Exploration},
author = {Alexander Dettmann and Steffen Planthaber and Vinzenz Bargsten and Raul Dominguez and Gianluca Cerilli and Marco Marchitto and Geoff Fink and Michele Focchi and Victor Barasuol and Claudio Semini and others},
url = {https://www.dfki.de/fileadmin/user_upload/import/12388_ant_astra22.pdf},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
booktitle = {16th Symposium on Advanced Space Technologies in Robotics and Automation},
abstract = {The exploration of lunar craters is of high interest, but
their rugged and inclined terrain also exceeds the mobility capabilities of current rovers, opening up a field of
application for legged exploration systems. This paper
presents a navigation and locomotion control system that
enables legged robots to be able to perceive the terrain,
to plan a path to a desired goal, and to control the path
execution while traversing unconsolidated, inclined, and
rugged terrain. The navigation system is closely coupled with the robotic motion control to be able to exploit the full potential of the flexible locomotion system.
In addition, the followed approach introduces a suitable
level of abstraction to achieve a modular and generic software that can support different types of walking robots.
This will allow, depending on specific requirements of
future space missions, to use energy-efficient four-legged
as well as more stable six-legged robots. The paper describes the guidance, navigation, and control approach
and shows first experimental results.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The exploration of lunar craters is of high interest, but
their rugged and inclined terrain also exceeds the mobility capabilities of current rovers, opening up a field of
application for legged exploration systems. This paper
presents a navigation and locomotion control system that
enables legged robots to be able to perceive the terrain,
to plan a path to a desired goal, and to control the path
execution while traversing unconsolidated, inclined, and
rugged terrain. The navigation system is closely coupled with the robotic motion control to be able to exploit the full potential of the flexible locomotion system.
In addition, the followed approach introduces a suitable
level of abstraction to achieve a modular and generic software that can support different types of walking robots.
This will allow, depending on specific requirements of
future space missions, to use energy-efficient four-legged
as well as more stable six-legged robots. The paper describes the guidance, navigation, and control approach
and shows first experimental results.
their rugged and inclined terrain also exceeds the mobility capabilities of current rovers, opening up a field of
application for legged exploration systems. This paper
presents a navigation and locomotion control system that
enables legged robots to be able to perceive the terrain,
to plan a path to a desired goal, and to control the path
execution while traversing unconsolidated, inclined, and
rugged terrain. The navigation system is closely coupled with the robotic motion control to be able to exploit the full potential of the flexible locomotion system.
In addition, the followed approach introduces a suitable
level of abstraction to achieve a modular and generic software that can support different types of walking robots.
This will allow, depending on specific requirements of
future space missions, to use energy-efficient four-legged
as well as more stable six-legged robots. The paper describes the guidance, navigation, and control approach
and shows first experimental results.
Fink, Geoff; Semini, Claudio
Proprioceptive sensor fusion for quadruped robot state estimation Proceedings Article
In: 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 10914–10920, IEEE 2020.
@inproceedings{fink2020proprioceptive,
title = {Proprioceptive sensor fusion for quadruped robot state estimation},
author = {Geoff Fink and Claudio Semini},
url = {https://iit-dlslab.github.io/papers/fink20iros.pdf},
year = {2020},
date = {2020-01-01},
urldate = {2020-01-01},
booktitle = {2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
pages = {10914\textendash10920},
organization = {IEEE},
abstract = {Estimation of a quadruped’s state is fundamentally important to its operation. In this paper we develop a lowlevel state estimator for quadrupedal robots that includes attitude, odometry, ground reaction forces, and contact detection.
The state estimator is divided into three parts. First, a nonlinear
observer estimates attitude by fusing inertial measurements.
The attitude estimator is globally exponentially stable and is
able to initialize with large errors in the initial state estimates
whereas a state-of-the-art EKF would diverge. This is practical
for situations when the robot has fallen over and needs to start
from its side. Second, leg odometry is calculated with encoders,
force sensors, and torque sensors in the robot’s joints. Lastly,
the leg odometry and inertial measurements are fused to obtain
linear position and velocity. We experimentally validate the
state estimator using a novel dataset from the HyQ robot. For
the entirety of the experiment the estimated attitude matched
the ground truth data and had a root mean square error
(RMSE) of [2 1 5] deg, the velocity estimates has a RMSE
of [0.11 0.15 0.04] m/s, and the position estimates, which are
unobservable, drifted on average [2 1 8] mm/s.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Estimation of a quadruped’s state is fundamentally important to its operation. In this paper we develop a lowlevel state estimator for quadrupedal robots that includes attitude, odometry, ground reaction forces, and contact detection.
The state estimator is divided into three parts. First, a nonlinear
observer estimates attitude by fusing inertial measurements.
The attitude estimator is globally exponentially stable and is
able to initialize with large errors in the initial state estimates
whereas a state-of-the-art EKF would diverge. This is practical
for situations when the robot has fallen over and needs to start
from its side. Second, leg odometry is calculated with encoders,
force sensors, and torque sensors in the robot’s joints. Lastly,
the leg odometry and inertial measurements are fused to obtain
linear position and velocity. We experimentally validate the
state estimator using a novel dataset from the HyQ robot. For
the entirety of the experiment the estimated attitude matched
the ground truth data and had a root mean square error
(RMSE) of [2 1 5] deg, the velocity estimates has a RMSE
of [0.11 0.15 0.04] m/s, and the position estimates, which are
unobservable, drifted on average [2 1 8] mm/s.
The state estimator is divided into three parts. First, a nonlinear
observer estimates attitude by fusing inertial measurements.
The attitude estimator is globally exponentially stable and is
able to initialize with large errors in the initial state estimates
whereas a state-of-the-art EKF would diverge. This is practical
for situations when the robot has fallen over and needs to start
from its side. Second, leg odometry is calculated with encoders,
force sensors, and torque sensors in the robot’s joints. Lastly,
the leg odometry and inertial measurements are fused to obtain
linear position and velocity. We experimentally validate the
state estimator using a novel dataset from the HyQ robot. For
the entirety of the experiment the estimated attitude matched
the ground truth data and had a root mean square error
(RMSE) of [2 1 5] deg, the velocity estimates has a RMSE
of [0.11 0.15 0.04] m/s, and the position estimates, which are
unobservable, drifted on average [2 1 8] mm/s.
Fink, Geoff; Semini, Claudio
The DLS quadruped proprioceptive sensor dataset Proceedings Article
In: Proceedings of the 2020 International Conference on Climbing and Walking Robots (CLAWAR), Moscow, Russia, pp. 24–26, 2020.
@inproceedings{fink2020dls,
title = {The DLS quadruped proprioceptive sensor dataset},
author = {Geoff Fink and Claudio Semini},
url = {https://iit-dlslab.github.io/papers/fink20clawar.pdf
},
year = {2020},
date = {2020-01-01},
urldate = {2020-01-01},
booktitle = {Proceedings of the 2020 International Conference on Climbing and Walking Robots (CLAWAR), Moscow, Russia},
pages = {24\textendash26},
abstract = {This paper presents novel datasets of the hydraulically actuated robot HyQ’s proprioceptive sensors. All of the datasets include absolute and relative joint encoders, joint force and torque sensors,
and MEMS-based and fibre optic-based inertial measurement units (IMUs). Additionally, a motion capture system recorded the ground truth data with millimetre accuracy. In the datasets HyQ
was manually controlled to trot in place or move around the laboratory. The sequence includes:
forward and backwards motion, side-to-side motion, zig-zags, yaw motion, and a mix of linear and
yaw motion. Additionally, there is motion on both rigid and soft terrain. All of the datasets are at
least five minutes long and one dataset is thirty minutes long. The aim of these datasets is to test,
evaluate, and compare different algorithms for state estimation using only proprioceptive sensors.
To aid in the development of new state estimation algorithms for soft terrain there are datasets
comparing rigid to soft terrain. Lastly, the extra long endurance trot dataset is for examining the
problem of long-term attitude estimation drift.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This paper presents novel datasets of the hydraulically actuated robot HyQ’s proprioceptive sensors. All of the datasets include absolute and relative joint encoders, joint force and torque sensors,
and MEMS-based and fibre optic-based inertial measurement units (IMUs). Additionally, a motion capture system recorded the ground truth data with millimetre accuracy. In the datasets HyQ
was manually controlled to trot in place or move around the laboratory. The sequence includes:
forward and backwards motion, side-to-side motion, zig-zags, yaw motion, and a mix of linear and
yaw motion. Additionally, there is motion on both rigid and soft terrain. All of the datasets are at
least five minutes long and one dataset is thirty minutes long. The aim of these datasets is to test,
evaluate, and compare different algorithms for state estimation using only proprioceptive sensors.
To aid in the development of new state estimation algorithms for soft terrain there are datasets
comparing rigid to soft terrain. Lastly, the extra long endurance trot dataset is for examining the
problem of long-term attitude estimation drift.
and MEMS-based and fibre optic-based inertial measurement units (IMUs). Additionally, a motion capture system recorded the ground truth data with millimetre accuracy. In the datasets HyQ
was manually controlled to trot in place or move around the laboratory. The sequence includes:
forward and backwards motion, side-to-side motion, zig-zags, yaw motion, and a mix of linear and
yaw motion. Additionally, there is motion on both rigid and soft terrain. All of the datasets are at
least five minutes long and one dataset is thirty minutes long. The aim of these datasets is to test,
evaluate, and compare different algorithms for state estimation using only proprioceptive sensors.
To aid in the development of new state estimation algorithms for soft terrain there are datasets
comparing rigid to soft terrain. Lastly, the extra long endurance trot dataset is for examining the
problem of long-term attitude estimation drift.
Semini, Claudio; Barasuol, Victor; Focchi, Michele; Boelens, Chundri; Emara, Mohamed; Casella, Salvatore; Villarreal, Octavio; Orsolino, Romeo; Fink, Geoff; Fahmi, Shamel; others,
Brief introduction to the quadruped robot HyQReal Proceedings Article
In: International Conference on Robotics and Intelligent Machines (I-RIM), IRIM 2019.
@inproceedings{semini2019brief,
title = {Brief introduction to the quadruped robot HyQReal},
author = {Claudio Semini and Victor Barasuol and Michele Focchi and Chundri Boelens and Mohamed Emara and Salvatore Casella and Octavio Villarreal and Romeo Orsolino and Geoff Fink and Shamel Fahmi and others},
url = {https://iit-dlslab.github.io/papers/semini19irim.pdf},
year = {2019},
date = {2019-01-01},
urldate = {2019-01-01},
booktitle = {International Conference on Robotics and Intelligent Machines (I-RIM)},
organization = {IRIM},
abstract = {This extended abstract provides a short introduction
to the latest hydraulic quadruped robot (HyQReal) that has
been developed at IIT. Compared to its predecessors HyQ and
HyQ2Max, this third generation machine is completely power
autonomous and features a smart hydraulic actuation system
developed by Moog Inc. The hydraulic actuators and the first
outdoor trials with the new robot are briefly introduced.
Index Terms\textemdashquadruped robot, hydraulics, power-autonomy},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This extended abstract provides a short introduction
to the latest hydraulic quadruped robot (HyQReal) that has
been developed at IIT. Compared to its predecessors HyQ and
HyQ2Max, this third generation machine is completely power
autonomous and features a smart hydraulic actuation system
developed by Moog Inc. The hydraulic actuators and the first
outdoor trials with the new robot are briefly introduced.
Index Terms—quadruped robot, hydraulics, power-autonomy
to the latest hydraulic quadruped robot (HyQReal) that has
been developed at IIT. Compared to its predecessors HyQ and
HyQ2Max, this third generation machine is completely power
autonomous and features a smart hydraulic actuation system
developed by Moog Inc. The hydraulic actuators and the first
outdoor trials with the new robot are briefly introduced.
Index Terms—quadruped robot, hydraulics, power-autonomy
Barasuol, Victor; Fink, Geoff; Focchi, Michelle; Caldwell, Darwin; Semini, Claudio
On the detection and localization of shin collisions and reactive actions in quadruped robots Proceedings Article
In: International Conference on Climbing and Walking Robots, pp. 51, 2019.
@inproceedings{barasuol2019detection,
title = {On the detection and localization of shin collisions and reactive actions in quadruped robots},
author = {Victor Barasuol and Geoff Fink and Michelle Focchi and Darwin Caldwell and Claudio Semini
},
url = {https://iit-dlslab.github.io/papers/barasuol19clawar.pdf
},
year = {2019},
date = {2019-01-01},
urldate = {2019-01-01},
booktitle = {International Conference on Climbing and Walking Robots},
volume = {49},
pages = {51},
abstract = {In this work, we investigate the impact of shin (distal limb) collisions on the performance of
quadruped locomotion. In particular, we carry out a detailed study on the locomotion performance
sensitivity to systematic errors and delays in the estimation of the shin contact location. We
propose a sensor-less model based on kinematics to estimate the location of a single contact
point at the shin level. In order to improve the robustness of robot locomotion, we use insights
from our sensitivity study and our proposed model to develop a reactive strategy to detect and
feedback information about the contact into the trunk controller. The effectiveness of the proposed
approaches is experimentally demonstrated on the HyQ robot.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In this work, we investigate the impact of shin (distal limb) collisions on the performance of
quadruped locomotion. In particular, we carry out a detailed study on the locomotion performance
sensitivity to systematic errors and delays in the estimation of the shin contact location. We
propose a sensor-less model based on kinematics to estimate the location of a single contact
point at the shin level. In order to improve the robustness of robot locomotion, we use insights
from our sensitivity study and our proposed model to develop a reactive strategy to detect and
feedback information about the contact into the trunk controller. The effectiveness of the proposed
approaches is experimentally demonstrated on the HyQ robot.
quadruped locomotion. In particular, we carry out a detailed study on the locomotion performance
sensitivity to systematic errors and delays in the estimation of the shin contact location. We
propose a sensor-less model based on kinematics to estimate the location of a single contact
point at the shin level. In order to improve the robustness of robot locomotion, we use insights
from our sensitivity study and our proposed model to develop a reactive strategy to detect and
feedback information about the contact into the trunk controller. The effectiveness of the proposed
approaches is experimentally demonstrated on the HyQ robot.