Some Early Boom Shaping Developments (Ferri, 1969)

This project will use simple aerodynamic and sonic boom models to examine modeling of the design problem itself. We will attempt to establish a meaningful direct functional dependence between the shape of the aircraft and aerodynamic and noise quantities of interest by studying the sensitivity of these quantities to changes in shape. We will experiment with several direct multiobjective optimization problem formulations. References:

1. Seebass, R., Argrow, B.; "Sonic Boom Minimization Revisited", AIAA Paper 98-2956
2. Shepherd, K.P., Sullivan, B.M.; "A Loudness Calculation Procedure Applied to Shaped Sonic Booms", NASA Technical Paper 3134, 1991
3. Carlson, H.W., Maglieri, D.J.; "Review of Sonic Boom Generation Theory and Prediction Methods", J. Acoust. Soc. Amer., 51, pp. 675-685 (1972)
4. Alonso, J.J., Kroo, I.M., Jameson, A. "Advanced Algorithms for Design and Optimization of Quiet Supersonic Platform", 40th AIAA Aerospace Sciences Meeting and Exhibit, AIAA Paper 2002-0144, Reno, NV, January 2002
5. Raymer, D.P.; "Aircraft Design: a Conceptual Approach", Third Edition, AIAA, 1999

Prerequisites:

Required: Scientific computing skills (Matlab or Fortran 90/95 or C), 1 semester in nonlinear optimization

Desired: Some background in statistical modeling, numerical analysis, multiobjective optimization

Keywords: multidisciplinary optimization, supersonic design, low boom, aerodynamic optimization

Figure 1: A typical Basic Service Set (BSS), with one AP, and several mobile stations.

The IEEE Standard governing WLANs describes two modes of operation: Distributed Coordination Function (DCF), and Point Coordination Function (PCF). By and large, chipset manufacturers implement only the DCF mode, and compatibility testing is done for this mode exclusively. The DCF is a contention-based mechanism where each wireless device (AP, or STA) competes for air time. More specifically, the 802.11 standard is implemented as follows:

1. At regular intervals (typically hundreds of ms) the AP broadcasts a beacon signal, which resets all devices internal clocks;
2. Assume a transmission opportunity ended at time t0. Depending on the status of the internal Backoff counter (BCK) of the device, the following actions can take place:

• If BCKr=0, and there is no activity on air for DIFS (Distributed Inter Frame Spacing = 50us in 802.11b) time, then station starts transmitting its data packet;
• If BCK=0 and during the DIFS period there is activity on air then device generates a random BCK between 0 and CW-1 (initially CW=CWmin = 16, in 802.11b); Then the following rules apply:

• For each slot time (Ts) of medium inactivity, the BCK decrements;
• The countdown is stopped whenever medium is busy, and the the countdown is resumed only after a supplemental AIFS (Arbitration Inter Frame Spacing, =DIFS in 802.11b) wait;
• When BCK reaches 0, the device transmits its packet data;
• If receiver (AP, or STA) receives successfully the packet, then it sends back on the air an Acknowledgement (ACK) frame, after a SIFT (Short Inter Frame Spacing = 10us) period after transmitter finishes its transmission;
• If transmitter receives the ACK correctly, then it assumes data was received correctly, and transmission ends; On the other hand, if ACK is not broadcast, or the transmitter does not receive correctly the ACK, then it assumes the transmission was not successful, and the following rules apply:

1. If current number of retransmissions has not reached a max threshold, then increment the Number of Retransmissions counter
2. If CW<= CWmax, then CW doubles;
3. A new random BCK is generated between 0 and CW-1;
4. Transmission process is restarted from step b. above

• When transmission ends, a post-backoff mechanism is implemented, by which a random BCK between 0 and CWmin-1 is generated, and a virtual countdown process is started obeying b.i and b.ii above.

Description of the problem

Basically there are two distinct regimes, completely opposite from one another:

1. Deterministic Regime: when no collision happen, and the initial MAC instance time are sufficiently far apart, then transmission happens in a deterministic mode. Such a case may happen when only voice data (such as VoIP stations are connected to the AP), or periodic transmitting stations are present. The deterministic regime analysis gives an upper bound on system performance;
2. Stochastic Regime: once collisions happen, or medium is detected busy during a packet arrival, the random generation of a Backoff counter happens, and the contention-based mechanism kicks in.

1. It does not take into account the deterministic regime, nor does the performance converge to that upper bound;
2. The Markovianity assumption is not always justifiable; is it possible to introduce a deterministic-stochastic hybrid model?
3. Subsequent improvements of the standard are not yet captured by the current model; in particular the 802.11n draft introduces new MAC mechanisms.

The goal of this research group is to address one or more of the issues above. Ideally, students should have:

- familiarity with basic stochastic modeling concepts (such as Markov chains)
- familiarity with use of network simulation software (such as ns2);
- familiarity with time-series data analysis software (such as perl, Matlab);

Figure 2: A Markov Chain Model for a WLAN device.

Bibliography

[1] G.Bianchi, Performance Analysis of the IEEE 802.11 Distributed Coordination Function, IEEE Journal on Selected Areas of Communications, 18 (3), 2000, 535-547.

[2] P.E.Engelstad and O.N.Osterbo, Non-Saturation and Saturation Analysis of IEEE 802.11e EDCA with Starvation Prediction, MSWiM.05: Proceedings of the 8th ACM international symposium on Modeling, analysis and simulation of wireless and mobile systems, Montreal, Canada, 2005.

If the telescope sensor is properly instrumented, it is possible to obtain time-tagged pairs of angles that relate the space object position to the sensor. With enough angle pairs, it is possible to estimate the position and velocity (the state) of the object, along with estimates of the uncertainties of these parameters. The workshop problem is to develop techniques to identify all the streaks made by each object. Streaks created by an object must somehow be associated with one another and disassociated from those made by other objects. One solution approach treats the state data as vectors in R6 and uses statistical clustering techniques for the association. A variation on this approach addresses physical properties of the orbits, sorting according to those least likely to change with small state variations.

Regardless of the approach, there are several interesting aspects to the problem. Automatic streak detection is required, with transform techniques of interest. Orbit mechanics are essential to effective state estimation as well as clustering techniques. In addition, traditional clustering techniques are computationally taxing. A related problem is identification of asteroids that might pose a hazard to planet earth. References:

Vallado, David A., Fundamentals of Astrodynamics and Applications, Edition 2, Microsoft Press, 2004; Milani, Andrea, "Three Short Lectures on Identifications and Orbit Determination," http://copernico.dm.unipi.it/~milani/preprints/preprint.html, 2006; Kaufman, L. and Rousseeuw, P., Finding Groups in Data - An Introduction to Cluster Analysis. Wiley Interscience 2005

Prerequisites:

Required: computing proficiency demonstrated by knowledge of at least one compiler, one semester differential equations, one semester statistics

Desired: one semester numerical analysis, familiarity with orbit mechanics and estimation theory.

Keywords: orbit mechanics, astronomical telescopes, statistical clustering

The Pan-starrs telescope on Mount Haleakela in Hawaii will be used, among other tasks, to search for asteroids. However, using its 1.4 billion pixel sensor, it will also detect earth-orbiting objects.

I am interested in examining the following problem: Given a set of performance objectives, how does one determine the space of designs available to the designer and find the optimal designs? How does the designer best visualize this space of options? Because he doesn't want just "the" answer, he wants to understand many aspects of the answer. While I'm interested in the general vehicle design problem, we will focus on aircraft design using a baseline tool that is to be determined as a concrete example with which we can test our ideas.

As more three-dimensional data becomes available, the value of automatic tools for utilizing such data increases. Several desired applications need methods by which to automate the finding of correspondences between three-dimensional data sets. These three-dimensional data sets frequently share many geometric characteristics, but also have significant differences, due to differences in data collection geometries, changes in sensor capabilities, temporal changes in the object of interest, and noise in the data.

One approach to finding unknown coordinate transformations, which are needed to align multi-dimensional data sets, is to require an expert to examine each set and label certain common landmarks. If sufficient landmarks, having the same unique labels can be found in both sets, the three-dimensional coordinates of the landmarks enable the coordinate transformation to be estimated. This is like aligning images of faces, by first extracting the coordinates the tips of the noses, the left corners of the mouths, the bases of the right earlobes, etc.

But when no prior expertise is available, we need methods of estimating the transformation from set of automatically generated coordinates of 'interesting' locations (unlabeled landmarks). We expect that a significant subset of corresponding unlabeled landmarks may exist somewhere in the data set to which we need to compare. To solve our alignment problems, we need to devise automated methods to robustly find a pair of large subsets from a pair of sets of unlabeled landmarks, such that the subsets have similar geometric characteristics.

Does there exist a rigid motion mapping the configuration of red points onto a subset of the blue points? If so, what is the blue subset, and what is the rigid motion? If not, how much deformation of the red configuration is needed to make it so?

In principle, these problems can be solved by exhaustively comparing every possibility, but the level of effort grows exponentially fast with the number of landmarks. Our goal will be to find and test new approaches to this problem, seeking to devise algorithms which are robust and far more efficient.

References:

1. Oliver Faugeraus, Three-Dimensional Computer Vision, MIT Press, 2001
2. Ian L. Dyrden, Kanti V. Mardia, {Statistical Shape Analysis}, Wiley, 1998
3. D. G. Kendall, D. Barden, T. K. Carne, H. Le, Shape and Shape Theory, Wiley Series in Probability and Statistics, 1999
4. Gene H. Golub, Charles Van Loan, Matrix Computations, Johns Hopkins University Press, 1996

Prerequisites:

Basics of linear algebra and matrix theory, basic computer programming skills, elementary Euclidean geometry

Desired: Ability to bring relevant ideas from one or more of geometry, invariant theory, optimization theory, graph theory, combinatorics, or something else.

Our design problems are motivated by a popular optical technology called Wavelength Division Multiplexing (WDM). In this setting each fiber is partitioned into a fixed number of wavelengths and demands sharing a common fiber must be transported on distinct wavelengths. A demand stays on the same wavelength along its routing path as much as possible. When this is infeasible, we can either deploy an extra fiber for the demand to continue on the same wavelength; or place a wavelength converter for the demand to continue on a different wavelength. Both options incur cost. One objective is to assign wavelengths and place converters in an advantageous way so as to minimize the total cost.

In this project we explore algorithms and heuristics for assigning wavelengths and placing converters. The goals include studying the tradeoff between optimality and complexity and understanding the gap between theoretical bounds and practical performance.

References:

[1] Matthew Andrews and Lisa Zhang, Complexity of Wavelength Assignment in Optical Network Optimization. (Please see Section VI.) Proceedings of IEEE INFOCOM 2006. Barcelona, Spain, April 2006. http://cm.bell-labs.com/~ylz/2006.coloring4.pdf

[2] C. Chekuri, et al. Design Tools for Transparent Optical Networks. Bell Labs Technical Journal. Vol. 11, No. 2, pp. 129-143, 2006.

Prerequisite:

Required: One semester of algorithms; One semester of theory of computing; One semester of programming.

Desired: Knowledge of Python and CPLEX.

Keywords: Analysis of algorithms, combinatorial optimization, implementation of heuristics