|
Research
Areas Multi-agent Systems |
|
![MCj04326800000[1]](Supervision_files/image025.gif)
|
Title: “Collision avoidance algorithms and application in Air Traffic
Management” |
|
|
Description:
The need for automated collision avoidance (Conflict Detection &
Resolution – CDR) arises in Air Traffic Management (ATM) as the increasing
airspace density challenges the capacity of human-based ATM. The sought ATM
CDR algorithms need to ensure flight safety as well as respect the limited
maneuverability of aircraft. Objectives: Our research focuses on the
following directions: ·
3D Extension of our Multi-agent
Navigation Function framework. ·
Development of Conflict Detection
and Resolution algorithms for ATM that comply with the aircraft's control
capabilities Achievements:
Our results so far include: ·
A Navigation Functions based
decentralized control algorithm for multiple 3-D nonholonomic vehicles,
providing guaranteed collision avoidance and convergence to the aircraft
destinations. ·
Integration of Navigation Functions
and Model Predictive Control (MPC) in a multi-aircraft control scheme to
avoid collisions without violating aircraft constraints. (Collaborative work
with the John Lygeros’ group@ETH) Researcher:
Giannis
Roussos |
|
|
Title: “ Development of distributed decision-making algorithms for flight
safety ” |
|
|
Description:
Non-holonomic systems with input constraints are becoming of increasing
interest along with the development of autonomous aircraft. The need for
multiple agent control laws that respect the agents' input constraints is
becoming necessary. These control laws must be safe, computationally feasible
and fully exploit the agents' dynamics. Objectives: Our
research currently includes: ·
Hybrid control laws for
non-holonomic systems with constrained inputs ·
Fusion of motion planning and
formal verification techniques for trajectory generation Achievements: Our results so far include: ·
Development of a hybrid control law
for steering non-holonomic agents with input constraints in unknown
environments, populated with obstacles, using only local sensing ·
Development of a multi-agent
control law using model checking techniques for a group of non-holonomic
agents with input constraints, guaranteeing convergence to a goal and
collision avoidance. Researcher:
Apollon
Oikonomopoulos |
|
|
Title: “Algorithms for cooperative multi-robotic
coverage” |
|
|
Description:
We consider 2D multi-agent
coverage. Particularly, we focus on determining algorithms for covering a
constrained planar space using trajectories on specific geometrical curves
projected on the particular space. The coverage procedure is combined with
suitably formed navigation functions controlling the agents’ motions among
the considered obstacles. Completeness, effectiveness, non-backtracking,
time-efficiency, robustness are important optimization components. Objectives: The research directions are the following: ·
Specifying alternative geometrical
ways of covering 2D constrained spaces ·
Motion control of the agents while
executing coverage tasks. ·
Nonholonomic multi-agent coverage of
unknown, bounded, complex and static environments. Agents have limited
sensing radius. ·
Optimization of the coverage method Achievements:
Our current achievements include: ·
Use of Space-Filling curves , a class of geometric
curves that have the self-repeating property ·
Treatment of arbitrary number of
convex and static obstacles. The workspace outer envelope is assumed convex. ·
Navigation Functions based control
for holonomic agents. ·
Completeness and non-backtracking
have been addressed. Researcher:
Kyriaki Makri |
|
|
Title: “Modeling and Hybrid Control of a swarm of multiple multilinked micro
robots” |
|
|
Description:
Micro Robots will play an
important role in the future in a variety of applications, including medical
applications and micro assembly applications. The inherent limitations of
these robots, including special physics, low computational and memory
capabilities and low power make them a challenging system to control, a
challenge expanded by the fact that a swarm of micro robots is necessary to
circumvent via cooperation the low individual capabilities. In our research
we try to focus on controlling and modeling issues regarding micro robots, and
especially we are trying to use the same powerful control primitives
available in the literature for macro robots, in the realm of micro robots. Objectives: We
focus on the following directions : ·
Single micro robot motion planning
and servo control ·
Multiple micro robot formation
control ·
Cooperative actuation Achievements:
Some of our main achievements include:
·
Multi level control of a single
robot masking the micro character of the micro robot motion to a unicycle. ·
Algorithms for cooperative transportation
& manipulation of rigid bodies ·
Algorithms for cooperative motion
of multiple robots under the presence of noise ·
Distributed formation of gird like
swarm structures Researcher: Grigoris Lionis |
|
|
Title: “Control methodologies for neuro-robotic systems” |
|
|
Description: Neuro-robotics is an emerging
field, derived from neuroscience and robotics. Research on models of human
motor control is conducted aiming at using them for controlling robots. Surface
Electromyographic (EMG) signals are used as the main control interface
between the user and a robot arm. Mathematical models are used and systems
are trained in order to decode in real-time the muscle activity to the
desired motion. The estimated motion is used to control a robot arm in 3D
space, thus incorporating the human arm redundancy. Objectives: Our main efforts are targeted
towards: ·
Analyzing human upper limb motion ·
Decoding EMG signals to human arm
motion ·
Controlling a robot arm using EMG-based
arm motion estimates Achievements:
So far we have achieved: ·
the derivation & Identification
of arm motion primitives and models describing joint angle dependencies
during the 3D motion of the arm ·
the development of decoding models
providing a continuous representation of the arm motion in 3D space using EMG
signals from a large population of muscles of the upper limb ·
Real-Time Control of a redundant
robot arm based on the EMG-based motion estimates Researcher: Panagiotis Artemiadis |
|
Title: “Motion Tasks for Neuro-Robotic Systems” |
|
|
Description: In realizing an EMG driven
neuro-robotic system, challenges include: (i) using the noisy EMG signals as
reference inputs to the feedback loop and (ii) mimicking human behavior
during obstacle avoidance and interaction with non-planar surfaces. It is
difficult to only use human neural signals to control an artifact. A human
arm cannot use the sensory feedback from the robot in order to directly
control it and the artifact does not have exactly the same morphology as the
human arm to allow perfect interpretation of the neural signals. Objectives: Introduction and implementation
of a strategy combining compliant behavior of the robot with respect to its
environment and obstacle avoidance with the use of an sensory (artificial)
skin. Achievements: Given a revolute joint robot
manipulator, with kinematic input constraints and operating in a known static
and bounded environment, we introduced a feedback dynamic control law that
allows the end-effector ·
converging to any feasible surface
point, ·
tracking a predefined trajectory
across the surface, and ·
being compliant with the surface. The methodology guarantees global convergence
and collision avoidance properties. Researcher:
Xanthi
Papageorgiou |
|
|
Title: “ Robotic Tele-operation Driven By
Electro-myo-graphic Signals” |
|
|
Description: By recording surface, non-invasive,
Electromyographic (EMG) signals from human forearms the grasping force, or/and
the force exerted from each human digit can be predicted. The
predicted force is the input for controlling a robot manipulator in order to
complete a compliance task. Objectives: Initial objectives include: ·
Developing advanced control methodologies, such as model
predictive controllers (MPC), for tele-operation of a remote robotic unit. ·
Finding
models accurately correlating EMG signals with forces exerted by human hand. ·
Applying
the proposed control framework to the Mitsubishi PA-10 robotic manipulator. Achievements:
Early results include: ·
An MPC
strategy has been formulated for generic 6 DOF robotic manipulators
incorporating actuator and state constraints. Conducted relevant computer
simulations. Researcher:
Alina
Eqtami |
|
|
|
|
|
Title: “Teleoperation of Underwater Robotic Vehicles Using Visual Servoing
Techniques” |
|
|
Description:
We address the issue of semi-autonomous ROV based underwater ship-hull or
structures inspection. Thus, issues related to Sensor Fusion, Localization, Target
Tracking in Image Space and Visual Servoing Control Schemes for nonholonomic
and holonomic vehicles are considered. Also, the development of a robust
shared-autonomy teleoperation scheme based on visual servo control is
pertinent in relevant applications. Objectives: ·
Optimal vehicle state estimation sing optical and
inertial sensors. ·
Development of
control schemes considering nonholonomic issues (due to ROV under-actuation)
and keeping the inspection target inside the field of vision ·
A comprehensive and user friendly GUI, providing
stabilized video as well as the most important information about the vehicle
state. Achievements:
So far we have accomplished the following: ·
Robust target tracking and target- referenced
localization of the vehicle using
optical sensors (onboard camera and laser pointers)
·
Development of an on-line path planning technique
for keeping the target inside the field of vision during point to point
motions ·
Development of a vision-based switching controller for effective
teleoperation for nonholonomic underwater vehicles. Researcher:
George Karras |
|
|
Title: “Compensation of External Disturbances for Underwater Robotic Vehicles”” |
|
|
Description:
The use of underwater robotic vehicles (AUVs, ROVs) has been extended during
the past few decades motivating the need for autonomous and semi-autonomous
systems performing tasks such as inspections, surveys, scientific
explorations, etc. Underwater vehicles operate in complex and uncertain
environments partially ought to the highly nonlinear hydrodynamics and
external disturbances (waves, currents and cable effects). Thus, the
development of feedback control schemes, accompanied with online estimation
algorithms of the dynamic environment is required for efficient and reliable
performance in real-time applications. Objectives: According to the above our
research is mainly focused on: ·
The identification of coupled
dynamic models for underwater vehicles ·
The online estimation of
environmental disturbances ·
The development of robust control
schemes, with emphasis to the viability of the system due to state
constraints, for the stabilization of an underactuated vehicle with respect
to specific targets, in the presence of environmental disturbances Achievements:
So far, we have achieved: ·
The identification of the 3D
decoupled dynamic model of an underactuated ROV, using an offline
identification algorithm. ·
The implementation of a Sensor
Fusion process, using measurements from an IMU and a Laser Vision System
(LVS). ·
The development of a dynamic
model-based algorithm providing an estimation of the vector of external
disturbances. Researcher:
Dimitra Panagou |
|
|
Title: “Mobile
robots cooperation methods for dynamic environment perception” |
|
|
Description:
A prerequisite for robotic collision-free
navigation in populated places is detection of walking by people. In contrast
to generic targets, possibly moving stochastically, human walking exhibits
recurrent patterns both in the behavioral level, as people tend to follow
specific paths, and in the motion level, as human body makes specific
stereotyped movements during walking. Cooperation between robots can help to
achieve higher overall efficiency due to occlusion accommodation. Objectives: The focus of our work is on the
following objectives: ·
Endow a moving mobile robot with
the capability of distinguishing between static background and dynamic
foreground without the need for accurate localization. ·
Endow a moving mobile robot with
the capability of tracking either walking people, or other moving objects,
within its surrounding area. ·
Developing methods augmenting the
overall environmental perception through cooperation between multiple robots. Achievements: Our achievements so far include
the following: ·
Static and dynamic background
identification, localization and mapping for a laser range-finder equipped
mobile robot ·
Tracking of multiple walking people
by a single mobile robot equipped with a laser range-finder. The proposed
algorithm is based on Multiple Hypothesis Tracking and Kalman Filtering. ·
Extension of the walking people
tracking algorithm to a centrally controlled dual-robot system. Researcher:
Nicolas
Tsokas |
|
|
Title: “Design & Development of a Modular Robotic
Structure” |
|
|
Description: The concept of a novel robotic module, the “R-Cell”, is
analyzed and developed. R-Cell can be utilized in building distributed, homogeneous
robotic systems. Each R-Cell is a rectangle endowed with: (a) motion
capabilities provided by four revolutionary joints, each equipped with a
clamping mechanism, and (b) deformation capabilities realized by four
prismatic joints. Our concept could be useful for a variety of applications
encompassing modular robotics like self assembly, self repair and reactive
shape optimization that conventional robots cannot accommodate. Objectives: Immediate future
objectives include: ·
Developing robotics structures able to reconfigure
themselves by rearranging their modules. ·
Developing modular non-rigid robotic structures with dexterous
deformation and force creating capabilities. ·
Building and integrating a moderate
(~10) number of modules. Achievements: Early efforts has lead to : ·
The Design of R-Cell, a
Lattice-type of module combining self-reconfiguring with dexterous motion and deformation
capabilities. ·
Analyzing “Deformation” and the force creating mechanism of the resulting robotic structure consisting of
R-Cell modules and via computer simulations demonstrating how: (a) a modular robotic
structure composed of R-Cells
reacts to external forces by optimizing its shape to better accommodate them
and (b) a requested force can be delivered at a given point of the modular
robot. Researcher: Giannis
Sissakis
|
|
