Acknowledgements
|
|
|
Image-based rendering and modeling
The goal
of this project is to use the constraints of (man-made) environments
(i.e. parallel and orthogonal lines and planes) in order to develop
sef of efficient and completely automated techniques for aquisition
of a variety of models from a sparse set of uncalibrated views. The
techniques apply to full 3D models as well as image mosaic representations.
|
|
|
|
Geometry and Dynamics of Multiple Views
All possible
constraints among multiple views of different image features (points, lines,
planes, curves) can be characterized as rank conditions of the so-called
multiple view matrix. This characterization gives rise to a new set of algorithms
from structure and motion recovery, image matching and image transfer from
multiple views of the static as well as dynamic scenes.
|
|
|
|
Optimization and Sensitivity Analysis in Structure and Motion Recovery Problems
The estimation
of scene structure and camera motion from image sequence requires
optimization techniques whose performance depends of measurement noise,
type of scene and camera motion. Carefull study of the behavior of the
objective function as these parameters vary sheds some light on the
difficulties with the traditional estimation techniques and in certain
cases reveals ambiguities present in human visual system.
|
|
|
|
Collision Avoidance and Small Aircraft Transportation Systems
Conflict
resolution architecture for multi-agent hybrid systems with emphasis
on Air Traffic Management Systems (ATMS) are developed. In order to allow
optimization of agents' objectives both cooperative and noncooperative
inter-agent coordination is proposed. The merging of inter-agent coordination,
which is modeled by discrete event systems, and agent dynamics, which
are modeled by differential equations, results in hybrid system.
|
|
|
|
Stereo Drive
This project
explored the feasibility of the use of visual sensing as a part of the
Advanced Vehicle Control System. The use of visual sensing has been demostrated
in the context of longitudinal and lateral control. The integration of
the vision subsystem with the vehicle control subsystem enabled us to
perform real-world experiments with vision as an integral component of
the vehicles control system.
|
|
|
Composition of Reactive Behaviors
Interaction between simplest reactive
behaviors is being explored in the context of Lego robotic platforms,
engaged in simple navigation , map building and pursuit and evasion
tasks involving multiple robots. The platforms are also being used
for class projects and masters thesis project on the development of
tools to programming embedded systems. |
|
|
| Discrete Event Modeling
of (Cooperative) Robotic Behaviors
As a part of my thesis work I designed
an architecture for a team of mobile agents involved in a variety of
navigation tasks. Agent's tasks were specified as networks of processes.
Modeling the individual processes in terms of finite state machines
enabled formal analysis of such task descriptions. As a result of this
analysis, a discrete event controller for the entire team was
synthesized. Safety and liveness properties of the system have been
verified, guaranteeing correct use of resources and proper communication
protocols between various components of the system.
|
|
|
|
Visually Guided Agents
The complexity of the tasks robotic agents can achieve
is directly linked to the representation of the environment they reside
in. While the visual sensing opens an avenue for estimating various properties
from single or multiple images; capturing the geometry and dynamics of the
environment is crucial for successful interaction of the robotic system with
the unknown and dynamically changing environment.
Vision and Control
I am particularly interested in the role and representation
of visual information for navigation. We are currently investigating both
theoretical and technical issues regarding the use of the vision in the
feedback loop of an automomous control system or a system with the human
in the loop.
From Images to 3D models
The focus of this work is on the development of algorithms
estimation of geometric and photometric properties of the environment.
Increasing the level of automation as well photorealism
of the obtained models has important implications in the areas of
image-based rendering, virtual reality, robotics and computer
graphics. The computational models and algorithms both draw
some inspiration as well as provide some insights into human
visual perception.
Hybrid and Discrete Event Systems
For more information go to
