Publications‎ > ‎

Conference Papers

Towards Robust Skill Generalization: Unifying LfD and Motion Planning

posted Aug 30, 2017, 7:40 PM by Reza A

Muhammad Asif Rana, Mustafa Mukadam, S. Reza Ahmadzadeh, Sonia Chernova, Byron Boots

Reference:
Muhammad Asif Rana, Mustafa Mukadam, S. Reza Ahmadzadeh, Sonia Chernova and Byron Boots, "Towards
Robust Skill Generalization: Unifying LfD and Motion Planning", In Proc. Robotics: Science and
Systems (RSS), Workshop on (Empirically) Data-Driven Manipulation, Boston, MA, USA, pp. 1--3,
12th-16th Jul. 2017.
Bibtex Entry:
@INPROCEEDINGS{rana2017towards, TITLE={Towards Robust Skill Generalization: Unifying LfD and Motion Planning}, AUTHOR={Rana, Muhammad Asif and Mukadam, Mustafa and Ahmadzadeh, Seyed Reza and Chernova, Sonia and Boots, Byron}, BOOKTITLE={Robotics: Science and Systems ({RSS}), Workshop on (Empirically) Data-Driven Manipulation}, PAGES={1--3}, YEAR={2017}, MONTH={July}, ADDRESS={Boston, MA, USA}, }
Abstract:
We present a novel unifying approach to conventional learning from demonstration (LfD) and motion
planning using probabilistic inference for skill reproduction. We also provide a new probabilistic
skill model that requires minimal parameter tuning, and is more suited for encoding skill
constraints and performing inference in an efficient manner. Preliminary experimental results using
real-world data are presented.

PDF Preview:



Skill Generalization via Inference-Based Planning

posted Jul 4, 2017, 4:35 PM by Reza A

Muhammad Asif Rana, Mustafa Mukadam, S. Reza Ahmadzadeh, Sonia Chernova, Byron Boots

Reference:
Muhammad Asif Rana, Mustafa Mukadam, S. Reza Ahmadzadeh, Sonia Chernova and Byron Boots, "Skill
Generalization via Inference-Based Planning", In Proc. Robotics: Science and Systems (RSS),
Workshop on Mathematical Models, Algorithms, and Human-Robot Interaction, Boston, MA, USA, pp. 1--3,
12th-16th Jul. 2017.
Bibtex Entry:
@INPROCEEDINGS{rana2017skill, TITLE={Skill Generalization via Inference-Based Planning}, AUTHOR={Rana, Muhammad Asif and Mukadam, Mustafa and Ahmadzadeh, Seyed Reza and Chernova, Sonia and Boots, Byron}, BOOKTITLE={Robotics: Science and Systems ({RSS}), Workshop on Mathematical Models, Algorithms, and Human-Robot Interaction}, PAGES={1--3}, YEAR={2017}, MONTH={July}, ADDRESS={Boston, MA, USA}, }
Abstract:
We present a novel approach which unifies conventional learning from demonstration (LfD) and motion
planning using probabilistic inference, for generalizable skill reproduction. We also provide a new
probabilistic skill model that requires minimal parameter tuning, and is more suited for encoding
skill constraints and performing inference in an efficient manner. Experimental validation on a
manipulation skill is presented.

PDF Preview:



Generalized Cylinders for Learning, Reproduction, Generalization, and Refinement of Robot Skills

posted Jun 5, 2017, 10:26 AM by Reza A   [ updated Jul 4, 2017, 4:30 PM ]

S. Reza Ahmadzadeh, Muhammad Asif Rana, Sonia Chernova

Reference:
S. Reza Ahmadzadeh, Muhammad Asif Rana, Sonia Chernova, "Generalized Cyliners for Learning,
Reproduction, Generalization, and Refinement of Robot Skills", In Proc. Robotics: Science and
Systems (RSS 2017), Massachusetts Institute of Technology in Cambridge, Massachusetts, USA,
12th-16th Jul. 2017.
Bibtex Entry:
@INPROCEEDINGS{ahmadzadeh2017generalized, TITLE={Generalized Cylinders for Learning, Reproduction,Generalization, and Refinement of Robot Skills}, AUTHOR={Ahmadzadeh, S. Reza and Rana Muhammad Asif and Chernova, Sonia}, BOOKTITLE={Robotics: Science and Systems ({RSS})}, YEAR={2017}, MONTH={July}, PAGES={1--10}, ADDRESS={Boston, MA, USA} }
Abstract:
This paper presents a novel geometric approach for learning and reproducing trajectory-based skills
from human demonstrations. Our approach models a skill as a Generalized Cylinder, a geometric
representation composed of an arbitrary space curve called spine and a smoothly varying
cross-section. While this model has been utilized to solve other robotics problems, this is the
first application of Generalized Cylinders to manipulation. The strengths of our approach are the
model’s ability to identify and extract the implicit characteristics of the demonstrated skill,
support for reproduction of multiple trajectories that maintain those characteristics,
generalization to new situations through nonrigid registration, and interactive human refinement of
the resulting model through kinesthetic teaching. We validate our approach through several
real-world experiments with a Jaco 6-DOF robotic arm.

PDF Preview:



Social Gaze Behavior for Face to Face Human-Robot Interaction

posted May 5, 2017, 7:40 AM by Reza A   [ updated Jun 23, 2017, 9:43 PM ]

Hagen Lehmann, Frank Broz, S. Reza Ahmadzadeh, Alessio Del Bue, Lorenzo Natale, Giorgio Metta

Reference:
Hagen Lehmann, Frank Broz, S. Reza Ahmadzadeh, Alessio Del Bue, Lorenzo Natale, Giorgio Metta,
"Social Gaze Behavior for Face to Face Human-Robot Interaction", In Proc. 9th International
Workshop on Human-Friendly Robotics (HFR 2016), Genoa, Italy, pp. 112-115, 29-30 September 2016.
Bibtex Entry:
@INPROCEEDINGS{lehmann2016social, TITLE={Social Behavior for Face to Face Human-Robot Interaction}, AUTHOR={Lehmann, Hagen and Broz, Frank and Ahmadzadeh, Seyed Reza and Del Bue, Alessio and Natale, Lorenzo and Metta, Giorgio}, BOOKTITLE={9th International Workshop on Human-Friendly Robotics ({HFR})}, PAGES={112--115}, YEAR={2016}, MONTH={September}, ADDRESS={Genoa, Italy} }
Abstract:
This short paper discusses the importance of human-like gaze behaviors for humanoid robots with
physical eyes. It gives a brief overview of the functions gaze fulfills in human-human interactions
from a social evolutionary perspective. In the second part we describe how human-like gaze has been
implemented in robot in the research field of social robotics in the last years. The last part of
the paper briefly introduces an architecture for a conversational gaze controller (CGC). The
parameters of this gaze controller are based on the analysis of a large corpus of gaze tracking
data collected during human-human conversations. We describe our experimental approach for
obtaining this data and discuss the implications of endowing robots with human-like behaviors and
enabling them to engage in face to face social interactions.

PDF Preview:



Trajectory Learning from Demonstration with Canal Surfaces: A Parameter-Free Approach

posted Dec 8, 2016, 8:15 AM by Reza A   [ updated Jun 26, 2017, 1:50 PM ]

Seyed Reza Ahmadzadeh, Roshni Kaushik, Sonia Chernova

Reference:
Seyed Reza Ahmadzadeh, Roshni Kaushik, Sonia Chernova, "Trajectory Learning from Demonstration with
Canal Surfaces: A Parameter-free Approach", In Proc. 16th IEEE-RAS Intl Conf. on Humanoid Robots
(Humanoids 2016), Cancun, Mexico, pp. 544--549, 15-17 Non. 2016.
Bibtex Entry:
@INPROCEEDINGS{ahmadzadeh2016trajectory, TITLE={Trajectory Learning from Demonstration with Canal Surfaces: A Parameter-free Approach}, AUTHOR={Ahmadzadeh, Seyed Reza and Kaushik, Roshni and Chernova, Sonia}, BOOKTITLE={Humanoid Robots ({H}umanoids), {IEEE-RAS} International Conference on}, YEAR={2016}, MONTH={November}, ORGANIZATION={IEEE}, PAGES={544--549}, ADDRESS={Cancun, Mexico}, DOI={10.1109/HUMANOIDS.2016.7803328} }
DOI:
10.1109/HUMANOIDS.2016.7803328
Abstract:
We present a novel geometric framework for intuitively encoding and learning a wide range of
trajectory-based skills from human demonstrations. Our approach identifies and extracts the main
characteristics of the demonstrated skill, which are spatial correlations across different
demonstrations. Using the extracted characteristics, the proposed approach generates a continuous
representation of the skill based on the concept of canal surfaces. Canal surfaces are Euclidean
surfaces formed as the envelope of a family of regular surfaces (e.g. spheres) whose centers lie
on a space curve. The learned skill can be reproduced, as a time-independent trajectory, and
generalized to unforeseen situations inside the canal while its main characteristics are preserved.
The main advantages of the proposed approach include: (a) requiring no parameter tuning,
(b) maintaining the main characteristics and implicit boundaries of the skill, and (c) generalizing
the learned skill over the initial condition of the movement, while exploiting the whole
demonstration space to reproduce a variety of successful movements. Evaluations using simulated and
real-world data exemplify the feasibility and robustness of our approach.

PDF Preview:



Encoding Demonstrations and Learning New Trajectories using Canal Surfaces

posted Jul 1, 2016, 8:08 AM by Reza A   [ updated Jun 24, 2017, 9:40 AM ]

Seyed Reza Ahmadzadeh, Sonia Chernova

Reference:
Seyed Reza Ahmadzadeh, Sonia Chernova, "Encoding Demonstrations and Learning New Trajectories using
Canal Surfaces", In Proc. 25th Int. joint Conf. on Artificial Intelligence (IJCAI 2016), Workshop
on Interactive Machine Learning: Connecting Humans and Machines, New York City, NY, USA, 9th-15th
July 2016.
Bibtex Entry:
@INPROCEEDINGS{ahmadzadeh2016encoding, TITLE={Encoding Demonstrations and Learning New Trajectories using Canal Surfaces}, AUTHOR={Ahmadzadeh, Seyed Reza and Chernova, Sonia}, BOOKTITLE={25th Inernational joint Conference on Artificial Intelligence ({IJCAI}), Workshop on Interactive Machine Learning: Connecting Humans and Machines}, PAGES={1--7}, YEAR={2016}, MONTH={July}, ADDRESS={New York City, NY, USA}, ORGANIZATION={{IEEE}} }
Abstract:
We propose a novel learning approach based on differential geometry to extract and encode important
characteristics of a set of trajectories captured through demonstrations. The proposed approach
represents the trajectories using a surface in Euclidean space. The surface, which is called Canal
Surface, is formed as the envelope of a family of regular implicit surfaces (e.g. spheres) whose
centers lie on a space curve. Canal surfaces extract the essential aspects of the demonstrations
and retrieve a generalized form of the trajectories while maintaining the extracted constraints.
Given an initial pose in task space, a new trajectory is reproduced by considering the relative
ratio of the initial point with respect to the corresponding cross-section of the obtained canal
surface. Our approach produces a continuous representation of the set of demonstrated trajectories
which is visually perceivable and easily understandable even by non-expert users. Preliminary
experimental results using simulated and real-world data are presented.

PDF Preview:



A Geometric Approach for Encoding Demonstrations and Learning New Trajectories

posted Jun 19, 2016, 9:01 AM by Reza A   [ updated Jun 24, 2017, 9:43 AM ]

Seyed Reza Ahmadzadeh, Sonia Chernova

Reference:
Seyed Reza Ahmadzadeh, Sonia Chernova, "A Geometric Approach for Encoding Demonstrations and
Learning New Trajectories", In Proc. Robotics: Science and Systems (RSS 2016), Workshop on Planning
for Human-Robot Interaction: Shared Autonomy and Collaborative Robotics, Ann Arbor, Michigan, USA,
18-22 June 2016.
Bibtex Entry:
@INPROCEEDINGS{ahmadzadeh2016geometric, TITLE={A Geometric Approach for Encoding Demonstrations and Learning New Trajectories}, AUTHOR={Ahmadzadeh, Seyed Reza and Chernova, Sonia}, BOOKTITLE={Robotics: Science and Systems ({RSS}), Workshop on Planning for Human-Robot Interaction: Shared Autonomy and Collaborative Robotics}, PAGES={1--3}, YEAR={2016}, MONTH={June}, ADDRESS={Ann Arbor, Michigan, USA}, ORGANIZATION={{IEEE}} }
Abstract:
We propose a novel learning approach based on differential geometry to extract and encode important
characteristics of a set of trajectories captured through demonstrations. The proposed approach
represents the trajectories using a surface in Euclidean space called Canal Surface. The surface is
formed as the envelope of a family of regular implicit surfaces (e.g. spheres) whose centers lie on
a space curve. Canal surfaces extract the essential aspects of the demonstrations and retrieve a
generalized form of the trajectories while maintaining the extracted constraints. Given a random
initial pose in task space, a new trajectory is reproduced by considering the relative ratio of the
initial point with respect to the corresponding cross-section of the obtained canal surface. Our
approach produces a continuous representation which is visually perceivable and easily
understandable even by non-expert users. Preliminary experimental results using real-world data are
presented.

PDF Preview:



Learning Symbolic Representations of Actions from Human Demonstrations

posted May 30, 2016, 3:37 PM by Reza A   [ updated Jun 24, 2017, 9:47 AM ]

Seyed Reza Ahmadzadeh, Ali Paikan, Fulvio Mastrogiovanni, Lorenzo Natale, Petar Kormushev, Darwin G. Caldwell

Reference:
Seyed Reza Ahmadzadeh, Ali Paikan, Fulvio Mastrogiovanni, Lorenzo Natale, Petar Kormushev,
Darwin G. Caldwell, "Learning Symbolic Representations of Actions from Human Demonstrations",
In Proc. IEEE Intl Conf. on Robotics and Automation, (ICRA 2015), Seattle, WA, USA, pp.
3801--3808, 26-30 May 2015.
Bibtex Entry:
@INPROCEEDINGS{ahmadzadeh2015learning, TITLE={Learning Symbolic Representations of Actions from Human Demonstrations}, AUTHOR={Ahmadzadeh, Seyed Reza and Paikan, Ali and Mastrogiovanni, Fulvio and Natale, Lorenzo and Kormushev, Petar and Caldwell, Darwin G.}, BOOKTITLE={Robotics and Automation ({ICRA}), {IEEE} International Conference on}, PAGES={3801--3808}, YEAR={2015}, MONTH={May}, ADDRESS={Seattle, Washington, USA}, ORGANIZATION={{IEEE}}, DOI={10.1109/ICRA.2015.7139728} }
DOI:
10.1109/ICRA.2015.7139728
Abstract:
In this paper, a robot learning approach is proposed which integrates Visuospatial Skill Learning,
Imitation Learning, and conventional planning methods. In our approach, the sensorimotor skills
(i.e., actions) are learned through a learning from demonstration strategy. The sequence of
performed actions is learned through demonstrations using Visuospatial Skill Learning. A standard
action-level planner is used to represent a symbolic description of the skill, which allows the
system to represent the skill in a discrete, symbolic form. The Visuospatial Skill Learning module
identifies the underlying constraints of the task and extracts symbolic predicates (i.e., action
preconditions and effects), thereby updating the planner representation while the skills are being
learned. Therefore the planner maintains a generalized representation of each skill as a reusable
action, which can be planned and performed independently during the learning phase. Preliminary
experimental results on the iCub robot are presented.

PDF Preview:



Learning Reactive Robot Behavior for Autonomous Valve Turning

posted May 30, 2016, 3:27 PM by Reza A   [ updated Jun 24, 2017, 9:53 AM ]

Seyed Reza Ahmadzadeh, Petar Kormushev, Rodrigo S. Jamisola Jr., Darwin G. Caldwell

Reference:
Seyed Reza Ahmadzadeh, Petar Kormushev, Rodrigo S. Jamisola Jr., Darwin G. Caldwell, "Learning
Reactive Robot Behavior for Autonomous Valve Turning", In Proc. 14th IEEE-RAS Intl Conf. on
Humanoid Robots (Humanoids 2014), Madrid, Spain, pp. 366-373, 18-20 Nov. 2014.
Bibtex Entry:
@INPROCEEDINGS{ahmadzadeh2014learning, TITLE={Learning Reactive Robot Behavior for Autonomous Valve Turning}, AUTHOR={Ahmadzadeh, Seyed Reza and Kormushev, Petar and Jamisola, Rodrigo S. and Caldwell, Darwin G.}, BOOKTITLE={Humanoid Robots ({H}umanoids), {IEEE-RAS} International Conference on}, PAGES={685--691}, YEAR={2014}, MONTH={November}, ORGANIZATION={IEEE}, ADDRESS={Madrid,Spain}, DOI={10.1109/HUMANOIDS.2014.7041386} }
DOI:
10.1109/HUMANOIDS.2014.7041386
Abstract:
A learning approach is proposed for the challenging task of autonomous robotic valve turning in the
presence of active disturbances and uncertainties. The valve turning task comprises two phases:
reaching and turning. For the reaching phase the manipulator learns how to generate trajectories to
reach or retract from the target. The learning is based on a set of trajectories demonstrated in
advance by the operator. The turning phase is accomplished using a hybrid force/motion control
strategy. Furthermore, a reactive decision making system is devised to react to the disturbances
and uncertainties arising during the valve turning process. The reactive controller monitors the
changes in force, movement of the arm with respect to the valve, and changes in the distance to the
target. Observing the uncertainties, the reactive system modulates the valve turning task by
changing the direction and rate of the movement. A real-world experiment with a robot manipulator
mounted on a movable base is conducted to show the efficiency and validity of the proposed approach.

PDF Preview:



Multi-Objective Reinforcement Learning for AUV Thruster Failure Recovery

posted May 30, 2016, 3:17 PM by Reza A   [ updated Jun 24, 2017, 9:57 AM ]

Seyed Reza Ahmadzadeh, Petar Kormushev, Darwin G. Caldwell

Reference:
Seyed Reza Ahmadzadeh, Petar Kormushev, Darwin G. Caldwell, "Multi-Objective Reinforcement Learning
for AUV Thruster Failure Recovery", In Proc. IEEE Symp. Series on Adaptive Dynamic Programming and
Reinforcement Learning (ADPRL 2014), IEEE Symp. Series on Computational Intelligence (SSCI 2014),
Orlando, FL, USA, 8-12 Dec. 2014.
Bibtex Entry:
@INPROCEEDINGS{ahmadzadeh2014multi, TITLE={Multi-Objective Reinforcement Learning for {AUV} Thruster Failure Recovery}, AUTHOR={Ahmadzadeh, Seyed Reza and Kormushev, Petar and Caldwell, Darwin G.}, BOOKTITLE={{IEEE} Symposium on Adaptive Dynamic Programming and Reinforcement Learning ({ADPRL}), Proc. {IEEE} Symposium Series on Computational Intelligence ({SSCI})}, PAGES={1--8}, YEAR={2014}, MONTH={December}, ORGANIZATION={IEEE}, ADDRESS={Florida, USA}, DOI={10.1109/ADPRL.2014.7010621} }
DOI:
10.1109/ADPRL.2014.7010621
Abstract:
This paper investigates learning approaches for discovering fault-tolerant control policies to
overcome thruster failures in Autonomous Underwater Vehicles (AUV). The proposed approach is a
model-based direct policy search that learns on an on-board simulated model of the vehicle. When
a fault is detected and isolated the model of the AUV is reconfigured according to the new
condition. To discover a set of optimal solutions a multi-objective reinforcement learning approach
is employed which can deal with multiple conflicting objectives. Each optimal solution can be used
to generate a trajectory that is able to navigate the AUV towards a specified target while
satisfying multiple objectives. The discovered policies are executed on the robot in a closed-loop
using AUV's state feedback. Unlike most existing methods which disregard the faulty thruster, our
approach can also deal with partially broken thrusters to increase the persistent autonomy of the
AUV. In addition, the proposed approach is applicable when the AUV either becomes under-actuated
or remains redundant in the presence of a fault. We validate the proposed approach on the model of
the Girona500 AUV.

PDF Preview:



1-10 of 20