הפקולטה להנדסה ע"ש איבי ואלדר פליישמן

~~Nimrod Shacham,
M.Sc. student under the supervision of Dr. Ofer Amrani

Wednesday, June 17, 2015 at 15:00
Room 011, Kitot Bldg., Faculty of Engineering

Diversified Decoding in HDPC Message Passing Algorithms

Abstract

For decades, Error Correcting Codes have been a fundamental building block of any communications system. The subject of ensuring good communication rate over unreliable channels is of critical importance and is becoming even more so for modern codes which push performance towards the theoretical limits of the communication channel. Low Density Parity Check Codes are one such family of codes. Despite the fact that they are lacking sound mathematical analysis, they have had tremendous success through the multitude of variations on the basic decoding algorithms and constructions. This thesis is concerned with employing the ideas behind iterative decoding of LDPC for other families of codes, particularly short codes with high density matrices (HDPC). There is a need for reliable graph-based iterative decoding of short codes; for instance this is beneficial in the framework of a newly introduced random version of OFDMA (R-OFDMA).
We first investigate how the structure of the parity check matrix can alter an iterative decoder's performance. Specifically we focus on the allowable redundancies in the matrices, where it is possible to benefit from inserting redundant parity checks without altering the code properties. We then investigate how altering the input LLRs to the decoder can affect the decoder's results and performance. By allowing multiple decoding passes, each with a slightly altered input matrix, one can positively affect error performance, keep the decoding latency unaltered and only increase the decoding complexity. Several algorithms for achieving this are offered and studied. We employ computation tree analysis and density evolution which support the simulation results.

17 ביוני 2015, 15:00

חדר 011, בניין כיתות-חשמל

כנס 40 שנה למחלקה להנדסת תעשיה

קרא עוד אודות כנס 40 שנה למחלקה להנדסת תעשיה

EE Seminar: Coherent Clustering and its Use for Image Denoising

~~
Moran Vatelmacher,
M.Sc. student under the supervision of Prof. Shai Avidan

Monday, June 22, 2015 at 15:00
Room 011, Kitot Bldg., Faculty of Engineering

Coherent Clustering and its Use for Image Denoising

Abstract

Cluster analysis or clustering is the task of grouping a set of objects in such a way that objects in the same group (called a cluster) are more similar (in some sense or another) to each other than to those in other groups (clusters). Today, the classical clustering methods are based on appearance information only. This information is sometimes insufficient, especially in the case of noisy data where there is a need to differentiate inliers from outliers. The human eye can sometimes see that a certain area belongs to a certain cluster even if the appearance distance from another cluster is shorter. This is partially done using context analysis. When using context information it is easier to identify outliers. This kind of analysis can help in image analysis applications such as image denoising - discussed here - for which segmentation of the image into clusters is relevant. Our results show that better clustering improves denoising.

22 ביוני 2015, 15:00

חדר 011, בניין כיתות-חשמל

סמינר מחלקתי

11 ביוני 2015, 13:00

חדר 206 בניין וולפסון

The Impact of Low Workload on Safety Drivers in TaxiBot Robotic Tugs

Inbal Levi - Department of Industrial Engineering

Abstract

The TaxiBot is a semi-robotic pilot-controlled towing vehicle designed to taxi airplanes to and from the gate and the runway, without using the airplane engines.

Airplane engines are inefficient when operated on the ground, creating air pollution and noise. The TaxiBot towing system provides an economic and effective solution that can help overcome this by reducing fuel cost, noise, vulnerability to Foreign Object Damage (FOD) and CO2 emissions.

The towing system allows airplanes to move on the ground while leaving the airplane pilot in control. After the airplane is pushed back from passenger gates all taxiing processes are controlled by the pilot from the cockpit. When towing the airplane the pilot steers the TaxiBot by using the tiller and braking pedals. In the TaxiBot rides a safety driver. The driver's primary role is to push back the airplane from the gate and to take the vehicle back from the runway. However, in emergency situations, the safety driver is supposed to take over the steering and breaking action.

The pilot's control at all times during the taxiing process causes prolonged low workload for safety drivers. The boredom due to low workload causes drops in driver performance, attention lapses, tiredness and out-of-the-loop-unfamiliarity (OOTLUF).

This study looks at ways to keep the driver involved in the taxiing mission, to improve his performance in emergency situations. During the first stage of the study, I interviewed 32 pilots about their experience and views regarding taxiing and help from a system like the TaxiBot. Based on these interviews and the relevant literature we developed an experimental system that simulates the taxiing. The experimental conditions differed in the task the driver performed while the plane is taxiing under the control of the pilot (no task , Dead Man Switch, hazard detection, navigation) and the frequency of adverse events (low vs. high). Twenty participants performed the task in each of the eight experimental conditions. Results showed that involving the operator in the taxiing mission improves the detection probability and thus helps dealing with the boredom. user behavior.

This work was performed under the supervision of Prof. Joachim Meyer.

ההרצאה תתקיים ביום חמישי 11.06.15, בשעה 13:00 בחדר 206, בנין וולפסון הנדסה, הפקולטה להנדסה, אוניברסיטת תל-אביב.

EE Seminar: Polar Codes for Noisy Write Once Memories

~~
Speaker: Adi Guzi Yitzhak
M.Sc. student under the supervision of Prof. David Burshtein

Wednesday, June 10th, 2015 at 15:30
Room 011, Kitot Bldg., Faculty of Engineering
Polar Codes for Noisy Write Once Memories
Abstract

Write once memories (WOM) are storage devices in which it is impossible or costly
to erase some memory cells. Flash memory is a one important example.
The problem of rewriting on a noisy WOM is considered.
We present a Blahut-Arimoto algorithm for computing the max sum rate of the noisy WOM, and show some numerical results.
A coding scheme for the noisy WOM using polar codes is also presented.
It is shown empirically that the scheme is close to the max sum-rate of noisy
WOMs when an arbitrary number of multiple writes is permitted.
The encoding and decoding complexities scale as O(N logN) where N is the blocklength.

10 ביוני 2015, 15:30

חדר 011, בניין כיתות-חשמל