Summer School - Advanced
Intelligent Systems
Synopsis
- Exposure to research topics at
cutting edge of Intelligent Systems.
- Global faculty.
- Concept based practical and
theoretical learning.
- Opportunity to conduct original
research and publish a conference paper.
- Cost of entire training program
including taxes.
- Study material and software on
CD.
- Certificate.
In
keeping with the pioneering spirit at IURS, we have worked very hard
over the last few months in order to design an Advanced Intelligent
Systems summer school that would not only expose participants to higher
level concepts in Robotics, but other areas of smart computing as well
and prepare them to step in to a world of true thinking-adapting
machines.
From concepts such as
advanced computer vision that make locating patterns and objects easy,
to building maps from Laser scans or deriving robot odometry from
infra-red scans this course will cover it all. Not only this, but
concepts from sensor networks and machine learning will be covered as
well! Since every concept covered in the workshop requires a good grasp
in order to proceed forward, practical exercises will be interwoven in
to each step in order to provide the best learning opporunities.
For this course as well IURS has
engaged not only its own team members, educated in India and abroad, as
faculty but also talented
researchers who hail from or studied in USA, Canada, Germany,
Romania and Hungary, not as guest lecturers but as faculty for
the course.
At
the end of the course participants will be able to choose a faculty
member to work on a project over the course of a month and then publish
the obtained results in a suitable conference or journal to enhance
their own academic experience. The course will not only teach students
concepts but will dip their hands deep in to intelligent systems
reasearch by giving them an opportunity to conduct original scholarly
work.
Prerequisites: Familiarity
with programming in C/C++ and MATLAB for analysis work.
Registration
Deadline : 30 Jun 2009 (On spot may also be accepted for
higher charges)
Dates:
1 Jul 2009 - 10
Jul 2009
Timings:
4.00PM to 6.00PM.
Duration:
20 hrs initial program + project work in contact and guidance of
researcher.
Venue
Indian Retail School
New Delhi
Accommodation:
Assitance in securing a place will be provided - please ensure you
inform us that you will need accommodation.
Course Contents
- Sensor Networks
- Hardware architecture.
- Topologies.
- Synchronization
mechanisms.
- Scheduling algorithms.
- Advanced Sensing &
Computer Vision
- Pattern matching .
- Object location.
- Scene matching.
- Scan matching.
- 2D & 3D map construction.
- Feature extraction.
- 3D Point Cloud Processing.
- Scale Invariant Feature Transforms.
- Plane matching & fitting.
- Region growing.
- Pose estimation
- Visual SLAM.
- Expedited registration mechanisms.
- Iterative closest point matching
(ICP).
- Human body recognition.
- Odometry derivation.
- Machine
Learning
- Artificial neural
networks.
- Recurrent neural
networks.
- Deep belief networks.
- Simulated annealing.
- Graphical
& Visualization Methods
- Graphics algorithms.
- Computer visualization.
- Information
visualization.
- Scientific
visualization.
- Uncertainty
visualization.
|
|
Hands on
Guided Tasks
|
Project
|
Description
|
|
Sensor Networks
|
Students
will be introduced to common problems faced and tasks performed by
researchers and engineers in the sensor networks field.
- Introduction to TinyOS .
- Using radio connected timer operated
blink controls.
- Remote data collection & logging.
- Using the ActiveMessage components.
- Biologically inspired synchronization .
The
goal of the exercise will be to familiarize students with the structure
of TinyOS, a commonly used OS in the sensor networks field, and have
them write applications that deal with common problems .
|
|
Plane Fitting
|
This combination of
exercises will try to motivate the need of 3d sensors like
stereo cameras and the need of plane fitting on the data provided by
this sensors. It will introduce two intersting algorithms for plane
fitting that are suitable for robotics, namely region growing plane
fitting algorithm and plane fitting based on mixture of gaussians.
During the course, students
will get to implement plane fitting based on mixture of gaussian and
use it for tracking the translation of the robot.
|
|
Machine Learning
|
This set of exercises will start with a
broad introduction to machine learning
techniques. It will go through mixture of gaussian models, parzen
windows, simulated anealing, fuzzy logics, baysian networks, and end
with neuronal networks.
A new intersting type of recurrent neural
networks will be presented, namely echo state networks. After each
chapter the techniques learned will be used in short practical
experiments that will include pattern recognition or decision taking.
|
|
Visual SLAM
|
Increase
in the usage of 3D sensing systems means that sensors like IR Swiss
Rangers, stereo cameras and other ToF systems provide not only
intensity images but 3D point clouds that are noisy in nature, however,
extremely useful if this noisy data can be accurately processed.
This
exercise in the course will deal with processing such 3D point clouds
in order to find persistent features in them that are invariant to
scale and rotation and also to some degree of illumination changes.
Furthermore, these features will then be matched between 3D point cloud
sets to perform scene matching, object detection, pattern detection,
map building and a quick form of known-correspondence registration
introduced in order to derive robot odometry for performing Visual SLAM
with the aid of these recognized features.
|
|
Grand Symphony
|
The
last project of the course has been termed the grand symphony, because
having learnt concepts of Machine Learning and Advanced Sensory data
processing and also having performed practical exercises all through
related to each concept, the students will work towards the common goal
of building a project that combines knowledge learnt from all the
topics in the class.
At the moment the final guided research
project topics are shortlisted as:
- 3D Map building using visual markers.
- Real-time robot odometry derivation from
3D point clouds.
- Uncertainty calculation and visualization
in 3D maps.
The students will be free to propose topics and ideas for a final
project themselves as well. The students will work under supervision
and mentorship of the faculty to complete the projects in a few weeks
after the course.
Results of the projects will be translated in to research papers that
will be submitted to conferences/journals as per their research merit.
|
Cost:
|
Registration
Payment Received By
|
Amount
|
|
Jun 25, 2009
|
|
Course Description
Having
had an introduction to basic concepts of robotics and intelligent
systems, this course is designed to blow the doors guarding the steps
of intelligent systems research wide open.
Students
will not only be introduced to cutting edge research topics, but will
also have an opportunity to work upon these brand new ideas and
concepts that are barely being introduced in to the research community
itself as well. Working with sensor data from Stereo Cameras, Swiss IR
Rangers and Laser Range Finders, while applying advanced Machine
Learning concepts the students will have an opportunity to experience
the cutting-edge techniques behind pattern matching, scene matching,
visual SLAM and map building. Not only this but students will also be
exposed to other intelligent systems concepts like sensor networks,
which will prepare them to deal with not only robotics as an
application field, but any field of intelligent systems.
A
lot of the course work is based upon concepts that are brand new in
research as well; for example, the SIFT is as recent as 2004. Working
on these cutting edge technologies, the students of this course will
truly be prepared to handle the upcoming developments in intelligent
systems and contribute to those as well. After all, intelligent systems
is not only restricted to robotics.
The Faculty
The
faculty of this course has been educated, and even are from, the USA,
Canada, India, Germany, Romania, Bulgaria and other countries. Having
won many international meritorious awards and conducted research in the
fields of intelligent systems and robotics, this diverse group of
individuals is highly qualified to train students in this field.
Moreover, experience in conducting various workshops in USA and India
has given the faculty world-class experience in ensuring the highest
levels of academic excellence throughout the course. Furthermore,
students stand to benefit from this dynamic and highly talented faculty
through the active research programs of IURS and their own home
institutions, in which they are involved, as well. This faculty has won
numerous international awards, published many papers and holds various
patents in their relevant fields.
Click here for more
information on the faculty.
|
