Difference between revisions of "Master Level Course Projects"

This project consists in develop one of the solutions proposed in literature and extended the LURCH software to include this kind of interface.

• C/C++ and OpenCV library
• Matlab (optionally)
• Linux

References:

• Phinx project [1]

Tools and instruments

C++, C, BCI2000

Linux

Bibliography

R. Blatt et al. Brain Control of a Smart Wheelchair [2]

The project aims at reproducing algorithms for ErrP detection from the literature.

Tools and instruments

Matlab

Bibliography

P.W. Ferrez, J. Millán. You Are Wrong! Automatic Detection of Interaction Errors from Brain Waves [3]

G. Schalk et al. EEG-based communication: presence of an error potential [4]

The idea is to use one or more calibrated camera to estimate the position and the trajectories of the moving objects in the scene. The skills required for this project are:

• C/C++ and OpenCV library
• Linux o.s.
• Geometry/Image processing
• Probabilistic robotics/IMAD

The project can be turned into a thesis extending the algorithm for a generic outdoor environment.

The idea is to use a calibrated camera to recognize the shelves, estimate the scale and improve the image quality. The skills required for this project are:

• C/C++ and OpenCV library
• Matlab (optionally)
• Linux
• Geometry/Image processing

It can be used in algorithms for tasks like matching different views of an object or scene (e.g. for stereo vision) and object recognition. The aim of this work is to integrate in an existent framework the existing solution proposed in literature.

Skills

• C/C++ and OpenCV library
• Matlab (optionally)
• Linux
• Geometry/Image processing

References:

• Oxford website [5]
• Hess website [6]
• Feature FAST [7]

Improvements could be the basis for a tesi.

Skills

• C/C++ and OpenCV library
• Matlab (optionally)
• Linux
• Geometry/Image processing

References:

• Visual SLAM by Single Catadioptric Stereo [8]
• Catadioptric reconstruction [9]

The aim of this tesina/project is to implement a trinocular algorithm based on SUGR, a library for Uncertain Projective Geometry.

Skills

• C/C++ and OpenCV library
• Matlab (optionally)
• Linux
• Geometry/Image processing

Skills

• C/C++ and OpenCV library
• Matlab (optionally)
• Linux
• Geometry/Image processing

The idea is create an images database indexer that allows to make query using images or strings.

Skills

• C/C++ and OpenCV library
• Matlab (optionally)
• Linux

References:

• CBIR system definition [10]
• Image database [11]

• NEURAL-BASED VISION SYSTEM - In this project we experiment different models of the neural structure of the visual cortex. In particular we collect couples of images with different fixation point and we reconstruct disparity, depth, shape recognition. A stereo disparity map is already available.
• RECOGNITION OF HAND MOTIONS - Develop a system able to recognize the hand gestures from images of people. Use this for robot interfacing.
• SEMANTIC MODELLING OF ACTIONS - Construct a semantic representation of actions to be matched against the data obtained from video/movements analysis.

• DATA MINING FOR RULE INDUCTION - The project is about analysing large data sets to discover correlations between substructures and properties. CDK and WEKA are possible software to use.
• IReNNS: LEARNING FROM STRUCTURES - This project will develop IReNNS, a recurrent neural net package, and develop new applications.
• REC analysis in WEKA - The project is about theory and the implementation in Java of REC diagrams. A Matlab tool is available at http://www.cs.rpi.edu/~bij2/rec.html.
• HOW TO TEACH SCIENCE - Propose a web site to teach University students a scientific topic using active participation and virtual experimentation.
• CAUSAL REASONING - In this project we will develop the theory of causality proposed by Pearl and apply it to examples in biological sciences.

Other topics coming soon.

analysis of different algorithms for the optimization of pseudo-Boolean functions, i.e., functions defined over binary variables with values in R. These functions have been studied a lot in the mathematical programming literature, and different algorithms have been proposed (1). More recently, the same problems have been faced in evolutionary computations, with the use of genetic algorithms, and in particular estimation of distribution algorithms (2,3). Estimation of distribution algorithms are a recent meta-heuristic, where classical crossover and mutation operators used in genetic algorithms are replaced with operators that come from statistics, such as sampling and estimation.

The focus will be on the implementation of a new algorithm able to combine different approaches (estimation and sampling, from one side, and exploitation of prior knowledge about the structure of problem, on the other), together with the comparison of the results with existing techniques that historically appear in different (and often separated) communities. Good coding (C/C++) abilities are required. Since the approach will be based on statistical models, the student is supposed to be comfortable with notions that come from probability and statistics courses. The project could require some extra effort in order to build and consolidate some background in math, especially in Bayesian statistics and MCMC techniques, such as Gibbs and Metropolis samplers (4).

The project can be extended to master thesis, according to interesting and novel directions of research that will emerge in the first part of the work. Possible ideas may concern the proposal of new algorithms able to learn existing dependencies among the variables in the function to be optimized, and exploit them in order to increase the probability to converge to the global optimum.

Picture taken from http://www.ra.cs.uni-tuebingen.de/

Bibliography

1. Boros, Endre and Boros, Endre and Hammer, Peter L. (2002) Pseudo-boolean optimization. Discrete Applied Mathematics.
2. Pelikan, Martin; Goldberg, David; Lobo, Fernando (1999), A Survey of Optimization by Building and Using Probabilistic Models, Illinois: Illinois Genetic Algorithms Laboratory (IlliGAL), University of Illinois at Urbana-Champaign.
3. Larrañga, Pedro; & Lozano, Jose A. (Eds.). Estimation of distribution algorithms: A new tool for evolutionary computation. Kluwer Academic Publishers, Boston, 2002.
4. Image Analysis, Random Fields Markov Chain Monte Carlo Methods

estimation.

The focus will be on the implementation of new hybrid algorithms able to combine estimation of distribution algorithms with different approaches available in the evolutionary computation literature, such as genetic algorithms and evolutionary strategies, together with other local search techniques. Good coding (C/C++) abilities are required. Some background in combinatorial optimization form the "Fondamenti di Ricerca Operativa" is desirable. The project could require some effort in order to build and consolidate some background in MCMC techniques, such as Gibbs and Metropolis samplers (4). The project could be extended to master thesis, according to interesting and novel directions of research that will emerge in the first part of the work.

Computer vision provides a large number of optimization problems, such as new-view synthesis, image segmentation, panorama stitching and texture restoration, among the others, (6). One common approach in this context is based on the use of binary Markov Random Fields and on the formalization of the optimization problem as the minimum of an energy function expressed as a square-free polynomial, (5). We are interested in the proposal, comparison and evaluation of different Estimation of Distribution Algorithms for solving real world problems that appear in computer vision.

Pictures taken from http://www.genetic-programming.org and (6)

Bibliography

1. Pelikan, Martin; Goldberg, David; Lobo, Fernando (1999), A Survey of Optimization by Building and Using Probabilistic Models, Illinois: Illinois Genetic Algorithms Laboratory (IlliGAL), University of Illinois at Urbana-Champaign.
2. Larrañga, Pedro; & Lozano, Jose A. (Eds.). Estimation of distribution algorithms: A new tool for evolutionary computation. Kluwer Academic Publishers, Boston, 2002.
3. Lozano, J. A.; Larrañga, P.; Inza, I.; & Bengoetxea, E. (Eds.). Towards a new evolutionary computation. Advances in estimation of distribution algorithms. Springer, 2006.
4. Pelikan, Martin; Sastry, Kumara; & Cantu-Paz, Erick (Eds.). Scalable optimization via probabilistic modeling: From algorithms to applications. Springer, 2006.
5. Image Analysis, Random Fields Markov Chain Monte Carlo Methods
6. Carsten Rother, Vladimir Kolmogorov, Victor Lempitsky, Martin Szummer. Optimizing Binary MRFs via Extended Roof Duality, CVPR 2007

Information Geometry has been recently applied in different fields, both to provide a geometrical interpretation of existing algorithms, and more recently, in some contexts, to propose new techniques to generalize or improve existing approaches. Once the student is familiar with the theory of Information Geometry, the aim of the project is to apply these notions to existing machine learning algorithms.

Possible ideas are the study of a particular model from the point of view of Information Geometry, for example as Hidden Markov Models, Dynamic Bayesian Networks, or Gaussian Processes, to understand if Information Geometry can give useful insights with such models. Other possible direction of research include the use of notions and ideas from Information Geometry, such as the mixed parametrization based on natural and expectation parameters (3) and/or families of divergence functions (2), in order to study model selection from a geometric perspective. For example by exploiting projections and other geometric quantities with "statistical meaning" in a statistical manifold in order to chose/build the model to use for inference purposes.

Since the project has a theoretical flavor, mathematical inclined students are encouraged to apply. The project requires some extra effort in order to build and consolidate some background in math, partially in differential geometry, and especially in probability and statistics.

Bibliography

1. Shun-ichi Amari, Hiroshi Nagaoka, Methods of Information Geometry, 2000
2. Shun-ichi Amari, Information geometry of its applications: Convex function and dually flat manifold, Emerging Trends in Visual Computing (Frank Nielsen, ed.), Lecture Notes in Computer Science, vol. 5416, Springer, 2009, pp. 75–102
3. Shun-ichi Amari, Information geometry on hierarchy of probability distributions, IEEE Transactions on Information Theory 47 (2001), no. 5, 1701–1711.

One of the common hypothesis in regression analysis is that the noise introduced in order to model the linear relationship between regressors and dependent variable has a Gaussian distribution. A generalization of this hypothesis leads to a more general framework, where the geometry of the regression task is no more Euclidean. In this context different estimation criteria, such as maximum likelihood estimation and other canonical divergence functions do not coincide anymore. The target of the project is to compare the different solutions associated to different criteria, for example in terms of robustness, and propose generalization of LASSO and LARS in non Euclidean contexts.

The project will focus on the understanding of the problem and on the implementation of different algorithms, so (C/C++ or Matlab or R) coding will be required. Since the project has also a theoretical flavor, mathematical inclined students are encouraged to apply. The project may require some extra effort in order to build and consolidate some background in math, especially in probability and statistics.

Picture taken from (2)

Bibliography

1. Tibshirani, R. (1996), Regression shrinkage and selection via the lasso. J. Royal. Statist. Soc B., Vol. 58, No. 1, pages 267-288
2. Bradley Efron, Trevor Hastie, Iain Johnstone and Robert Tibshirani, Least Angle Regression, 2003

variables associated to both internal nodes and leaves of the tree.

The project will focus on the understanding of the problem and on the implementation of different algorithms, so (C/C++ or Matlab or R) coding will be required. Since the approach will be based on statistical models, the student is supposed to be comfortable with notions that come from probability and statistics courses.

The project is thought to be extended to master thesis, according to interesting and novel directions of research that will emerge in the first part of the work. Possible ideas may concern the proposal and implementation of new algorithms, based on recent approaches to phylogenetic inference available in the literature, as in (3) and (4). In this case the thesis requires some extra effort in order to build and consolidate some background in math in oder to understand some recent literature, especially in (mathematical) statistics and, for example, in the emerging field of algebraic statistics (5).

Other possible novel applications of phylogenetic trees have been proposed in contexts different from biology, as in (6). Malware (malicious software) is software designed to infiltrate a computer without the owner's informed consent. Often malwares are related to previous programs thought evolutionary relationships, i.e., new malwares appear as small mutations of previous softwares. We are interested in the use of techniques from phylogenetic trees to create a taxonomy of real world malwares.

Picture taken from http://www.tolweb.org/tree/ and http://www.blogscienze.com

Bibliography

1. Felsenstein 2003: Inferring Phylogenies
2. Semple and Steel 2003: Phylogenetics: The mathematics of phylogenetics
3. Louis J. Billera, Susan P. Holmes and and Karen Vogtmann Geometry of the space of phylogenetic trees. Advances in Applied Math 27, 733-767 (2001)
4. Evans, S.N. and Speed, T.P. (1993). Invariants of some probability models used in phylogenetic inference. Annals of Statistics 21, 355-377.
5. Lior Pachter, Bernd Sturmfels 2005, Algebraic Statistics for Computational Biology.
6. A. Walenstein, E-Md. Karim, A. Lakhotia, and L. Parida. Malware Phylogeny Generation Using Permutations of Code, Journal in Computer Virology, v1.1, 2005.

The problems that will be proposed are still unknown. As soon as the domains will be published, the work will start by analyzing their main characteristics and, then we will identify which RL algorithms are most suited for solving such problems. After an implementation phase, the project will required a long experimental period to tune the parameters of the learning algorithms in order to improve the performance as much as possible.

References:

• Drivatar in Forza Motorsport (http://gar.eecs.ucf.edu/)
• TORCS (http://torcs.sourceforge.net/)
• CIG@PoliMI (http://cig.dei.polimi.it/)
• Daniele Loiacono's AI&R Lab Projects (http://home.dei.polimi.it/loiacono/uploads/Teaching/HomePage/AIRLab_AA20092010.pdf)

Picture taken from http://gar.eecs.ucf.edu/

References

• Galactic Arm Race (http://gar.eecs.ucf.edu/)
• TORCS (http://torcs.sourceforge.net/)
• CIG@PoliMI (http://cig.dei.polimi.it/)
• J. Togelius, R. De Nardi and S. Lucas. Towards automatic personalised content creation for racing games (http://julian.togelius.com/Togelius2007Towards.pdf)
• Daniele Loiacono's AI&R Lab Projects (http://home.dei.polimi.it/loiacono/uploads/Teaching/HomePage/AIRLab_AA20092010.pdf)

Several theses and projects are available on this topic and involve different games: TORCS, Unreal Tournament and Quake Live. Please contact us for additional information.

References

• TORCS (http://torcs.sourceforge.net/)
• QuakeLive (http://www.quakelive.com)
• Daniele Loiacono's AI&R Lab Projects (http://home.dei.polimi.it/loiacono/uploads/Teaching/HomePage/AIRLab_AA20092010.pdf)

Our projects and theses on this topic involve two different games: Unreal Tournament 2004 and TORCS. Please contact us for additional information.

References

• Drivatar in Forza Motorsport (http://research.microsoft.com/en-us/projects/drivatar/default.aspx)
• TORCS (http://torcs.sourceforge.net/ TORCS)
• Botprize (http://botprize.org/)
• Daniele Loiacono's AI&R Lab Projects (http://home.dei.polimi.it/loiacono/uploads/Teaching/HomePage/AIRLab_AA20092010.pdf)

While browsing the Web, a user should be able to visualize the annotations relative to the page they are visiting, and to add new annotations as well.

The purpose of this project is to develop translation tools for different social annotation systems, collect their data in a common format (expressed using an ontology), and show them through a unique user interface, able to display different annotations (i.e. geo coordinates, dates, tags, etc.) in different ways. Moreover, tests and evaluations should be performed on this aggregator to show how efficient the queries are when performed on-the-fly or from an intermediate knowledge base.

Aim of this project is to expand a traditional search engine with semantic information, so that also documents containing words related to the ones in the query can be retrieved. Existing thesauri and ontologies can be used, as well as more dynamic and collaborative sources of knowledge such as user tags and wikipedia pages and categories.

Starting points for this work can be the projects "SeQuEx - Semantic Query Expansion" and "Enriching search results with semantic metadata".

Goal of this project is to develop a Facebook application to analyze a user's network of friends and automatically detect groups to suggest lists.

In the project "Wikipedia Category Map" a tool has been developed to extract the graph of Wikipedia categories, to store it in RDF format and to interactively visualize and explore it. Aim of this project is to analyze the resulting graph for the extraction of semantic relationships; for example it is possible to define metrics of distance between topics in the graph, which can be useful for various purposes in information retrieval.

Aim of this project is to design and develop a software tool to extract this kind of information from a single social network platform (decided by the student). It may be required also some kind of analysis or visual representation of the collected data.

One possible example is the annotation of mp3 files available on the Web. They can be classified in genres or associated to datatype properties, such as rating, title, length and release date... also exploiting data already available in http://musicbrainz.org/

In order to speed up the development and the maintenance of embedded applications, a way to update the firmware on a microcontroller without the need of connecting cables or programmers can be very handy. We are developing a framework for rapid prototyping of low-cost robots, with smart devices that exchange data on a CAN bus network. The CAN bus bootloader is one of the components we need for this project, enabling remote firmware upgrades of all the devices connected to the CAN network.

This project aims to develop a CAN bus bootloader for STM32 ARM Cortex-M3 microcontrollers, and eventually for other architectures.

No special skills are required, except basic c and object oriented programming.

The aim of the project is to design, implement, test, and evaluate a new version of the device. While the charge phase, implemented by an ad-hoc IC controlled by the PIC, is quite efficient, some work has to be done in order to design a new solenoid and improve the way the PIC implements a sort of modulation of strength of the shoot, in order to implement small passages between robots. The final purpose is to improve the efficiency of the system, with the aim of minimize the energy consumed and maximize the energy transmitted to the ball, in order to obtain more powerful shots.

The first part of the project is focused on some theoretical aspects in order to understand the current design and evaluate how to improve it, while the second phase will be focused on the implementation and test of a new prototype of the kicking device. Experience with PIC-based systems is a plus, but not required, while some experience with electronics circuits is highly recommended. Students are supposed to work in the lab following a set of safety guidelines and rules with circuits at 400V, with 10A pick current during charge phase, and even more during shots. Student from electronics engineering are really welcomed to choose this project.

1. http://www.er.ams.eng.osaka-u.ac.jp/robocup-mid/index.cgi

When developing embedded applications it is frequently needed to look at *hardware register content* in order to *debug the code*. All commercial development suites offer register views that show their contents as well as the meaning of each bit. Open source development solutions currently lack this feature, meaning that you have to look to the correct memory location and map the content to the corresponding register bits manually. This seems to be one of the most limiting issues when developing embedded application using open source solutions.

This project aims to fill this gap, developing an Eclipse plug-in that shows the register contents in a tree viewer, like most commercial suites do.

The aim of the project is first to improve the current design of the PIC-based board, and realize a new working prototype, and then to implement and evaluate different algorithms able to estimate more precisely the x,y and theta odometric data from the mice readings. Experience with PIC-based systems and some experience with electronics circuits is a plus. Students are supposed to redesign the electronic board, improve the firmware of the PIC, and work on the algorithm that estimates the robot position on the PC. It would be also interesting to evaluate the possibility to embed the optimization and estimation algorithms in the firmware of the PIC in order to produce a stand-alone device.

Ask the tutors of the project for extra material, such as data-sheets and other documentation.

• implementing embedded algorithms for the estimation of the IMU attitude to be compared with the actual one (e.g., Kalman filter, DCM, Madgwick, etc.)
• developing a, easy to use, procedure for the calibration of IMU parameters
• making a comparison with commercial units using a robot arm as testbed
• validate the accuracy of the IMU on a flying platform
• integrate the measurements from a GPS to reduce drift and provide accurate positiong (this will make it definitely a MS thesis)

Material

• electronic board and eclipse based C development toolkit for ARM processors
• papers describing the algorithms we are interested in implementing

Expected outcome:

• few different AHRS algorithms with comparative results
• user-friendly procedure to calibrate the IMU
• a sistem which integrated IMU and GPS to provide accurate positioning

Required skills or skills to be acquired:

• C programming on ARM microcontroller
• background on kalman filtering and attitude estimation

When developing embedded applications it is common the need to test some algorithm in some fast way, without to re-program the whole firmware every time. PAWN (http://www.compuphase.com/pawn/) is a *simple and lightweight scripting language with a C-like syntax*. Execution speed, stability, simplicity and a small footprint were essential design criteria for both the language and the abstract machine, making PAWN suitable for embedded applications.

This project aims to port the abstract machine to ARM Cortex-M3 microcontrollers, add a set of functions to interface with the underlying hardware peripherals and then to embed it as ChibiOS/RT (http://www.chibios.org) thread.

• Implementation of mechanical and electronical parts of the robots for the management of the ball and kicking
• Design of robot behaviors (fuzzy systems)
• Coordination of robots
• New sensors

These projects allow to experiment with real mobile robots. Participation to the championships is a unique experience (2000 people, with 800 robots playing all sort of games...)

The project can be turned into a thesis by facing different problems in depth.

TiltOne is a robot with only two wheels that can stand in vertical position following an unstable equilibrium point. The control is applied by commanding an amount of torque to the wheels, allowing the robot to mantain it's gravity center vertical aligned to the wheel axis.

The aim of the project proposal is to implement and compare different control solutions, based on classical approach (as PID and LQR control) and Machine Learning approach (as Reinforcement Learning control policies), that allow the robot to move following a given trajectory at a given speed.

Inertial Measurement Unit and a Fire-i Camera. The pro ject will be focus on the problem to estimate both unknown rotation between the two devices and the extrinsic/intrinsic parameters of the camera. This algorithm allows to use the system for SLAM or robotics applications, like a wereable device for autonomous navigation or augmented reality.

Tools and instruments

Matlab/C++

Links

Matlab Toolbox for mutual calibration [12]

List of pubblications[13]

After a deepen bibliography research, the work will be focused on developing one of these algorithms into an existing framework and, only for tesi option, investigate possible improvements.

The algorithms interested are based on [14]:

• Extended Kalman Filter [15]
• Unscented Kalman Filter [16]
• FastSLAM [17]
• GraphSLAM [18]

Tools and instruments

Matlab/C++

• BIOINSPIRED ROBOT HEAD FOR VISION - design and build a robot head able to host 2 cameras with 2dof of freedom each to create a human-like vision system. The movements can be obtained using 4 McKibben actuators for each camera, or electric actuators.
• NEW HARDWARE FOR MAXIMUMOne - the humanoid robot is moved by more than 20 actuators and needs input from all of them. The new architecture FPGA based will move the arm and the head.
• SIMULATOR OF HUMANOID ROBOT - complete the simulator of MaximumOne with all the dof. The simulator will use the same Matlab algorithms of the controller.
• INTEGRATING MANIPULATION AND VISION ON MAXIMUMOne - develop a natural vision system that uses the neck and the eyes movements to follow objects and to concentrate on grasping targets. The integration can be done in matlab/Simulink and integrated in the MaximumOne model.
• MANIPULATION ONTOLOGIES - develop an ontology approach to find the right way to grasp an object , considering both the object and the hand characteristics.
• PATH PLANNING AND COLLISION AVOIDANCE IN OOPS - Randomized path planning is a strategy to produce paths for complex devices. An open source project (OOPS) is available; the project is about integrating path planning with a robot simulator.

All the projects can be turned into a thesis.

• KINEMATIC/DYNAMIC MODEL OF WARUGADAR - develop a complete kinematic analysis of a quadruped robot, useful for planning the foot position on uneven terrains. The dynamic model will be useful for learning different gaits.
• GAIT GENERATION AND CONTROL FOR WARUGADAR - Study Central Pattern Generation, develop a CPG implementation in Matlab or Python. Adapt the method to a quadruped robot (Warugadar).
• ROBO FISH - Continue the development of hardware and software for the robotic fish Zoidberg2, and study a fish colony.
• EMBOT WALKING - complete the robot with 4 wheels used as feet. Control it and experiment.
• ROBOTIC EXPERIMENTS WITH BIOLOID - using Bioloid experiments hw and gaits, develop software for the humanoid challenges at ICRA2010.

All the projects can be turned into a thesis.

• ADVANCED EMG ANALYSIS - To develop prosthesis and therapeutic tools we need to analyze EMG signals according to position, velocity, force. Starting from signal acquisition develop in Matlab a model for force and velocity control.
• HAND PROSTHESIS DESIGN - Define a virtual model of the hand prosthesis (actuation and sensors) and develop a controller based on multi-class classification of electro-myographic signals.

All the projects can be turned into a thesis.