a brief overview of projects

Unfortunately I don't have enough time to update this page often.

A probabilistic analysis of JPEG image compression

A probabilistic analysis of JPEG compression standard shows the limitations of compression via small DCT blocks. Still the DCT transform make it a lot easier to independentize the coeffecients to be saved. Combination of Huffman and RLE used in JPEG also improve the compression ratio.

GFpy, a tool for matrix manipulation in python

An alpha version is now available. I dont gurantee that it works, but it will be helpful. take a look at the package homepage

Implementation of very long BCH codes

BCH codes for purpose of error floor control of LDPCs is implemented.

Hybrid Fuzzy-A* for mobilie robots

A combination of Fuzzy modelling with A* search, for smooth robot planning: project homepage

Function Approximation Using Probabilitic Methods

• In Collaboration with Daniel Khashabi.

The main goal of this project was to find a model to consider correlation and dependency between multiple output of estimator (in a multi-output system). Through this project all related method, naming RVMs, GPs, Time Series, PGMs, and their extentions were studied and some of the proposed algorithms were examined. Majority of the proposed algorithms were supervised in case that correlation between outputs was determined by user.

On the Sufficiency of Matched Filters

In this document, sufficiency of matched filters are proven in the probability sense. the document is available here: PDF

Adaptive Tiled Neural Networks

abstract:

In this paper, a novel function approximation approach based on a combination of conventional Neural Networks and tile coding approximators is proposed. The proposed approach can maintain the desired features of both approaches whiles eliminates the deficiencies of each method. The combination will reduce the sharpness of tile coding. It will also provide an easy way to adjust the accuracycomplexity of the approximation according to the function being approximated (adaptive tiling) and the subspace used on. In this algorithm, it is possible to construct the approximator with specified and various approximation accuracies in different subspaces. This feature enables us to allocate an arbitrary accuracycomplexity wherever a more accurate approximation is needed. Finally simulation studies are presented to show the efficiency of and applicability of the proposed approach.

relatated page: homepage

more details are available in IEEE xplore.

Experiences with Non-linear Modelling and Control of Ball and Plate Control System.

• In Collaboration with Daniel Khashabi. Under supervision of Dr. H. A. Talebi

The project started as a part of the ‘‘linear control“ course, instructed by Dr. H. A. Talebi. The goal of the project was mathematical modeling, linearization and controller design for movement of a ball on a plate, “ball-on-plate”. We started the project, reading the older papers which were devoted mostly to implementation of new controlling algorithms and methods. We derived the nonlinear mathematical modeling of the system, implementing the non-linear implementation in the Simulink. The linearization was done by Simulink’s tools and finally we designed the controller with the use of MATLAB’s tools for design and optimization of controllers. We got the ‘‘best project” of the course award plus $100 by Dr. Talebi !

More details, sources, explanations about the project: English report, Project homepage

Continuous-space Reinforcement Learning and Its Application on Double inverted Pendulum

• In Collaboration with Danial khashabi and Mohsen Fallahi Under supervision of Dr. B. Samadi

Being interested in main idea of Reinforcement Learning, control with just a reward signal. We started the project, with the goal of studying the cutting-edge ideas, papers and trends in the RL, contributing to the current trends by their practical implementations. We chose ‘‘double inverted pendulum“ as a challenging benchmark for our project and started its construction in the ‘‘industrial control laboratory” which was under the supervision of Dr. Behzad Samadi. The simulation was based on Webot’s 3d models which provide a very realistic simulation of 3d environment, working in parallel with MATLAB implementation of the RL. Finally, we ended up, with several successful simulations; but the utterly successful practical implementation of the model remained unreachable, due to unavailability of high-quality equipments.

More results and reports of the project: PDF-Persian

Linear Control of Flexible Joint

• In collaboration with danial khashabi and faramarz jabar vaziri

In continuing experience with linear control, a flexible joint was modeled and controlled. Several methods were used to design a PID controller and best were chosen. One of the methods (the best one) was also applied to a real flexible joint, the performance was so good that the instructor was kind of surprised.

Image of the real flexible joint.

For more details see full English report

Design and Simulation of an Operational Amplifier with HSPICE

In studying electronics one of the most important parts is an operational amplifier or OPAMP. In this project an OPAMP were designed using MOS transistors and simulated with HSPICE in nonlinear model (real mode) and its performance was tested.

Designed circuit.

For more details see full Farsi report

Generating Motion Patterns Using Evolutionary Computation in Digital Soccer

Abstract— Dribbling an opponent player in digital soccer environment is an important practical problem in motion planning. It has special complexities which can be generalized to most important problems in other similar Multi Agent Systems. In this paper, we propose a hybrid computational geometry and evolutionary computation approach for generating motion trajectories to avoid a mobile obstacle. In this case an opponent agent is not only an obstacle but also one who tries to harden dribbling procedure. One characteristic of this approach is reducing process cost of online stage by transferring it to offline stage which causes increment in agents’ performance. This approach breaks the problem into two offline and online stages. During offline stage the goal is to find desired trajectory using evolutionary computation and saving it as a trajectory plan. A trajectory plan consists of nodes which approximate information of each trajectory plan. In online stage, a linear interpolation along with Delaunay triangulation in xy-plan is applied to trajectory plan to retrieve desired action.

Interpretation of Script into LaTeX Code

Most of times working with latex code is not as easy as working with a simple script, so it might be easier for operator to write its contends in some defined script and then it can be interpreted or translated into latex code. In other hand the script can be output of another program who don’t even know what and how the output is used, this example has several features like thumb index and is multilingual and bi-directional. Unfortunately this project is for a Company and I can’t give the code out but an example can be useful.

example of input script (contends are copy-write protected): TXT

example of output file (contends are copy-write protected): PDF

Modelling of Non-ideal Rope and the End of Rope

Primary aim of this project was to find an analytic explanation of the behavior of the end of a non-ideal rope. But after several discussions with my friends, it was clear that the plot was depended to many parameters that couldn’t be all considered.

animation: GIF image