You are here

Open Master projects offered by SPECS on: Robots, Insects, Rodents, Humans and Synthetic Agents

PROJECTS 2014-2015

Coming Soon!!

 

 

_____________________________________________________________________________________

PROJECTS 2008-2013

 

ROBOTS

Learning affordances: from sensory motor coordination to learning how to interact with the world

Keywords: Robotics, Affordances, Sensory Motor Contingencies

Description: Affordances define the relation between an agent and its environment through its motor and sensing capabilities (e.g. graspable, movable or eatable). In this project we propose to investigate a unified framework to be able to learn the affordance capabilities of a robotic device together with the affordances of an external scenario, i.e. a table with square or circular objects. For the investigation we will use simulation environments: the "icub simulator" and "graspit!" that require c++ programming skills.  The research is part of a new european project called "extending Sensory Motor Contingencies (eSMC)" starting soon in SPECS in which the candidate can also be involved. The project has connections to rehabilitation purposes, learning the affordances of an impaired arm.

Type: Research

Knowledge: basic programming skills

Supervisor: Marti Sanchez Fibla

Coordinator: Paul F.M.J. Verschure

Contact (email): marti.sanchez(at)upf.edu

 

Real Robot Foraging Experimentation

Keywords: Cognitive Systems, Robotics, Foraging, Outdoor Navigation, Benchmarking. 

Description: The synthetic forager (SF) project aims to understand the basis of foraging behavior by constructing an autonomous agent that can act in the real-world. One of the main steps of this project is the validation of the neuromorphic control model: the robot has to perform in ways that are consistent with the behavior of live foraging animals. Foraging essentially means that the agent explores its environment, identifies regions of interest and successfully extracts/collects resources.

The objective of this project is to design real-world experiments (i.e. in the street) that could validate robot foraging behavior. For such it will be necessary to review the literature about foraging and the behavioural experiments with live animals. The student will have to: identify relevant features of foraging and the experiments used to account for them; understand the principles of our model and the particularities of the SF robot; develop and perform the experiments with the SF robot; analyze the data. The student will be able to design and suggest new experiments to be realized also on live animals by SPECS's partners.


Tasks:

  1. Outdoor Robot Experiments.

  2. Benchmarking the DAC cognitive architecture.

Knowledge: . C++ programing, Linux and Matlab skills are welcomed.

Supervisor: Cesar Costa

Coordinator: Paul F.M.J. Verschure

Contact (email): cesarcosta(at)gmail.com

 

Learning how to solve the Morris Water Maze using the Distributed Adaptive Control architecture.

Keywords: Cognitive Systems, Robotics, Robot Experiments, Maze Navigation, Sequence Learning

Description: The main features to be able to solve the Morris Water maze are learning how to recognize where you are according to some perceptual cues and then finding a hidden goal according to your current position. Starting from the DAC cognitive architecture that is being developed at SPECS the aim of this project will be to integrate these first capabilities and test them out in a real robot immersed in a virtual mixed reality environment. The ability to navigate in mazes will also depends on the agent capabilities of performing sequence learning, that is learning the correct sequence of visual cues that can lead to the goal. The project will include real robot experiments and the possibility  of using a mixed reality robot arena for the final demonstrations.
 The research is linked to a new European project starting soon in SPECS called GoalLeaders in which the candidate could be involved.

Type: Research

Knowledge: basic programming skills

Supervisor: Marti Sanchez Fibla

Coordinator: Paul F.M.J. Verschure

Contact (email): marti.sanchez(at)upf.edu

 

Robot Maze Navigation and Exploration

Keywords: Cognitive Systems, Robotics, Robot Experiments, Maze Navigation, Sequence Learning

Description: Maze navigation and exploration is a controlled setting for robot and real biological system experimentation. For navigating through a maze one must be able in a first stage to follow corridors and recognize crossings. Then one should be able to recognize whenever we are encountering the same crossing again. A reactive navigation system that could identify crossings would be an important building block for a cognitive system that could negotiate a maze. Starting from the DAC cognitive architecture that is being developed at SPECS the aim of this project will be to integrate these first capabilities and test them out in a real robot immersed in a virtual mixed reality maze environment. The ability to navigate in mazes will also depends on the agent capabilities of performing sequence learning, that is learning the correct sequence of visual cues that can lead to the exit. The project will include real robot experiments and the possibility of using a mixed reality robot arena for the final demonstrations.


Tasks:  

  1. Model a cognitive task,

  2. implement it in a real robot.

  3. Performe experiments in the mixed reality robot arena.  

Knowledge:  C++ programing, Linux and Matlab skills are welcomed.

Supervisor: Encarni Marcos

Coordinator: Paul F.M.J. Verschure

Contact (email): encarni.marcos(at)gmail.com

 

Cognitive robotics: Application of a Cognitive Model on Robots

Keywords:

Description: In XIM a cognitive model based on the distributed adaptive control architecture (DAC) is used to maintain a world model consisting of the knowledge of visitors in the space, using not only in position but also information about color of clothes and weight of people. This model is well-suited for application on a behaving robot, which also has to update its world-model constantly using incoming partial sensory input. In this project the student will explore the existing model and study its application on a real-world robot. The final goal is to make the robot identify different objects in its environment while it is freely behaving, and performing simple tasks based on the internal goals of the robot, which are modulated at the same time by the consequences of their own actions.

Tasks: 

  1. Study the existing model,

  2. propose the application on a robotic platform and implementation,

  3. documentation and presentation of the work

Knowledge: Programming knowledge in C++, mathematical background and a sound interest in robotics

Supervisor: Andre Luvizotto

Coordinator: Paul F.M.J. Verschure

Contact (email): luvizotto(at)gmail.com

 

Visual navigation using a blimp/quad copter based flying robot

Keywords: cognition, navigation, robotics, biologically inspired

Description: Integration of existing biologically based insect neural models for visual navigation on a flying robot. Current models include distance estimation, speed estimation, course control and collision avoidance. The integrated system has to be able to perform autonomous take off, landing and basic navigation. The robot platform to use is a blimp robot, filled with helium and/or a fast moving quad copter. The former is an ideal platform for indoor experimentation, that has autonomy of about an hour. The latter is fast and furious. Both systems are interfaced to the iqr neural simulator by means of a bluetooth link and a wireless video analogue link

Tasks: The integration of the current navigation capabilities and development of landing and taking off control system. The student will work with the neural simulator IQR and the goal is to achieve a fully autonomous flying system.

Knowledge: Interest in biology, basic programming 

Supervisor: Melle Hofman

Coordinator: Paul F.M.J. Verschure

Contact (email): mel(at)capone.nl

 

Studying Artificial Olfaction and Vision with Autonomous Robots

Description: Insects, and in particular the moth, are amazingly good at detecting minute concentrations of pheromone signals and can locate them and distances greater than 1 Km. The most impressive thing is that the brain of an insect consists only of about 200.000 neurons, whereas a human has > 1 10^12, and are extremely good at navigating using multimodal sensory input (olfaction, vision, proprio-receptive). The overall goal of this project is to study the insect strategies and neural substrate for autonomous navigation. This will require the modeling of neural structures of insects by means of the neural simulator iqr, and their link to a robot to evaluate its performance under controlled conditions. In this project the student will explore existing models and study its application on a real-world robot.

Type: Research

Tasks: Study existing models, propose the application on a robotic platform and implementation, documentation and presentation of the work

url: http://www.specs.upf.edu/?q=projects/446

Knowledge: Programming knowledge in C++, mathematical background and a sound interest in robotics

Supervisor: Melle Hofman

Coordinator: Paul F.M.J. Verschure

Contact (email): mel(at)capone.nl

 

Swarm behavior in robots

Description: Social insects such as bees or ants do display optimized swarm behaviors despite their limited computational capabilities (with brains of about 200.000 neurons). We propose to study some of those behaviors such as foraging, pheromone signaling, etc and implement them in a neural simulator and several robots. In the flying variation of this system (simulated or physical) we will explore the idea of reconstructing surface contours using the images gathered from the swarm.

Type: Research

Tasks: The implementation of the models will be done using the iqr neural simulator, on which the student will be trained. The student will work on the Distributed Adaptive Control framework for behavior regulation. The models will be tested first on simulations and later using a set of wireless robots (epuck's) and/or flying robots.

Knowledge: Interest in biology and robotics

Keywords: cognition, navigation, robotics, biologically inspired

Supervisor: Zenon Mathews / Paul Verschure

Coordinator: Paul F.M.J. Verschure

Contact (email): paul.verschure(at)gmail.com

 

A novel automated experimental arena for insects: Construction of an interactive arena for ants experiments using tracking data

Keywords: navigation, interaction, tracking

Description: We want to study rule learning in walking animals, and in particular the ant. The idea is to develop and arena that has some passages and signs that can be used to give cues to ants about the path to take to reach the food source. These paths and signs will indicate the only path to find the food and they will change every run so the ant does not follow odor traces or Path integration signals. Its trajectory will be tracked by an existing tracking system (AnTS) and a system will have to be created to decide on how to change the environment accordingly.

Tasks: 

  1. Development of the hardware of the interactive arena, and its control software.

  2. use the AnTS tracking system to track the animal behavior and decide what  action to take depending on the animal's behavior.

Knowledge: Interest in biology or robotics, basic electronics and/or programming

Supervisor: Zenon Mathews

Coordinator: Paul F.M.J. Verschure

Contact (email): zenon.mathews(at)gmail.com

 

Insect landmark navigation model

Keywords:

Description: What does the world look like if you are an ant? When studying the behavior of small animals such as insects, we normally are tied to the human perspective. Yet, the world as an ant sees it look very different. Not only because of the perspective, but also because the insect eye, and the mammalian eye are fundamentally different. In collaboration with biologists, we have collected data of ants navigating an arena which contains visual landmarks. The question we are asking is: What is the minimal model that could explain the navigation behavior of the ants?

Type: Research

Tasks: Reconstruction of ant arena in Unity, implementation of a faceted eye, collect data from ant trajectory, realization of a minimal landmark navigation model, evaluation of the model

Knowledge: 3D programming, neuronal modeling

Keywords: Insect navigation, 3D

Supervisor: Zenon Mathews

Coordinator: Paul F.M.J. Verschure

Contact (email): zenon.mathews(at)gmail.com

 

Discovering the life-like properties of future robotics: A world of autonomous behaving Robots: their environment and ecology

Keywords: Robotics, biologically inspired navigation, interaction, tracking.

Description:  This project wants to create and deploy a collection of walking, crawling, slithering and jumping robots that will interact with the users to allow the experience hands-on on what the future holds. Visitors and especially children can begin to better comprehend how application of “natural” behavior to machines, as it is done in our miniature bio-mimetic systems, can help to study and find better solution for interactive intelligent machine that in turn can help to explore which ecological laws can be exploited in a given task (or to solve a certain problem). Especially insect based robots are believed to be of great practical use in many applications. Constructing such robots will also provide new insights in natural behaving systems. In this project, bio-mimetic-inspired robots will be situated in a simulated ecological-like environment that will incorporate a few natural scenarios (i.e. desert- and meadow-like terrain). Possibly, these scenarios will be interchanged during a day/night cycle. The inhabitants of this environment will be behaving according to the behavior of typical key players of any given ecological system: food collectors, predators, scavengers.

Tasks: Installation Realization: construction of the Robots arena, implementation of relevant behavior in the robots, (visions, collision detection, swarm behavior).
Knowledge: Interest in Multi-robot exploration and coordination, basic electronics and

Supervisor: Anna Mura

Coordinator: Paul F.M.J. Verschure

Contact (email): a.mura3000(at)gmail.com

Mixed and Virtual Reality

Social behavior for Avatars based on psychological model

Keywords: Distance regulation, modeling

Description: The “Zurich Model of Social Motivation” is a bio-cybernetic model that integrates motivational systems of fundamental importance to the control of human social behavior, into a theory of social motivation. The models main predictions are related to distance regulation between two or more social agents. The regulation of distance is conveyed to the subject by means of motives. In its basic form, the model postulates three interacting motivational subsystems: the security, the arousal and the autonomy/sex system. In this project the aim is to deploy the ZHM for the modeling of the behavior of users in a virtual environment.

Tasks:

  1. Update the existing implementation of the ZHM, and integrate it with the control of Avatars

  2. Real-time parameter estimation of ZHM parameters based on user behavior, and deduction of predictions.

Type: Research

Knowledge: Modeling, programming, interest in psychological models

Supervisor: Ulysses Bernardet

Coordinator: Paul F.M.J. Verschure

Contact (email): ulysses.bernardet(at)upf.edu

 

Rich expression in mixed reality: Usage of Moiré patterns for luminous floor animations in XIM

Description: One of the key challenges in a rich mixed-reality installation is the non-symbolic expression of states. When using computer graphics, the number of possible animations is nearly infinite. This flexibility has the shortcoming that there is no straight forward way to go from a given (internal) state to its expression. Yet this is exactly what an installation such as the eXperience Induction Machine needs. The goal of this projects is to explore simple animation algorithms for this task.
Glass patterns, i.e. Moiré patterns that are generated by the superposition of correlated random layers, should provide a good starting point for this. The ideas is that with relatively few parameters (color, density, movement direction), a large number of different animations with different emotional connotations can be drawn. In a fist step, these patterns will be drawn on the floor in XIM. 
To test the effectiveness of the approach, the different patterns will be validated using questionnaires and physiological measurements.


Tasks: Prototype implementation of Glass patterns (in any framework known), deployment of the pattern in XIM, validation of the pattern algorithm by means of questionnaires and physiology

url: http://diwww.epfl.ch/w3lsp/books/moire/kit1.html

Type: Research

Knowledge: Knowledge of an animation framework, basic programming skills, experimental research

Keywords: Expression of emotions, HCI

Supervisor: Alberto Betella, Ulysses Bernardet

Coordinator: Paul F.M.J. Verschure

Contact (email): ulysses.bernardet(at)upf.edu

 

Composition engine for real-time, multi-level, multi-modal effector control

Description: The definition of virtual and mixed-reality environments comprises a large number of controllable parameters. These parameters include environmental factors, such as the illumination, and terrain, but also parameters of virtual characters, acting in the virtual environment, sonification, and abstract effector displays. The “composition engine” will provide a high level interface to control signals of effectors. The need for the composition engine stems from the problem of orchestrating atomic elements into a coherent, effective whole, a fundamental problem met in any multi-level control system.

A comparable problem can be found in the domain of music, where the issue is how to come from high-level compositional parameters to the specific parameters for sound synthesis. In previous work we haveaddressed this problem by proposing the, so called, RoBoser paradigm. RoBoser is a real-world composition system built on the interactive compositional system “Curvasom“ (Jônatas Manzolli & Verschure, 2005; Wassermann, Manzolli, Eng, & Verschure, 2003). Curvasom operates in real time using a selection of predefined sets of sonic parameters, taking advantage of the MIDI protocol. Curvasom is based on a number of internal heuristics for transforming input states into timed MIDI events, and a set of parametric sound specifications with which the system is seeded (Jonatas Manzolli & Maia, 1998). To generalize the concept of Curvasom, the MIDI can be mapped onto predefined movement sequences of virtual characters that are controlled parametrically. The composition engine will generalize the concept of Cuvasom into the domain of generic high level control of the multi-modal rendering of data, the control of ambient parameters, and control signals for the behavior of dynamic and static virtual objects including avatars and other aspects of the multi-modal interactive displays.

Tasks: Familiarization with the concepts of Curvasom and RoBoser. Implementation of a bridge between Curvasom and other kinds of effectors.

Type: Research

Knowledge: Basic knowledge about music composition, programming skills

Supervisor: Sylvain Le Groux, Ulysses Bernardet

Coordinator: Paul F.M.J. Verschure

Contact (email): ulysses.bernardet(at)upf.edu

 

Interactive visualization of connectome data in the eXperience Induction Machine

Description:

We are facing a data deluge in contemporary science where are ability to generate data outstrips that to analyze it. At SPECS we are addressing this issue in the CEEDS project wil the goal to help humans efficiently navigate and explore complex data spaces. The aim of this project is to develop and evaluate the interactive representation and exploration of brain connectivity (connectome) data (The human connectome: a structural description of the human brain. O Sporns, G Tononi, R Kotter. PLoS Comput Biol (2005) vol. 1 (4) pp. e42) in the eXperience Induction Machine in the context of the CEEDS project. This project will build up on a current project of visualization and navigation of iqr system files.

Tasks: Conceptualization of data understanding techniques, 3D visualization of data, navigation in XIM

Type: Research

Knowledge: Programming skills

Supervisor: Alberto Betella, Ulysses Bernardet

Coordinator: Paul F.M.J. Verschure

Contact (email): ulysses.bernardet(at)upf.edu

 

Self expression in virtual reality based on music and pictorial material

Description: The goal of this project is to develop self-expression tools for disabled people. In the context of BrainAble project (http://www.brainable.org) we want to provide VR users a number of non-verbal self-expression tools based on music and gestures (e.g. Mura et al.,2008) to complement conventional communication based on text. Audio-visual self-expression tools allow the user to communicate with family members, friends, therapists or other users within networked virtual community either online or offline. The focus lies on non-verbal self-expression tools such as automated music composition (RoBoser), selection of semantically and emotionally rich pictures or videos, or avatar behaviors (posture and gestures). Self-expression will be assisted using AI-based system which will help the user to create a personalized vocabulary of multi-modal expressions.

Type: Research

Supervisor: Sylvain Le Groux, Arnau Espinosa, Ulysses Bernardet

Coordinator: Paul F.M.J. Verschure

Contact (email): ulysses.bernardet(at)upf.edu

 

Mixed Reality 3D Tetris

Description: Lesions in the brain following a stroke or a traumatic brain injury often lead to motor and cognitive disorders. After a lesion in the brain, the recovery of function is possible by cortical plasticity mechanisms. The best way to drive this plasticity is still under discussion. However, several approaches seem to be effective such as intensive training, tasks directed towards specific deficits, online delivery of feedback, etc.. All these features can be put together by means of interactive scenarios that engage the subjects in specific tasks. We propose the creation of a scenario that recreates the well known tetris game, but in a rehabilitative approach. This scenario is to be developed using the Unity Gaming Engine

Type: Research

Tasks: Literature review on stroke, traumatic brain injury and recovery mechanisms. Analyze user response and impact. Develop Tetris with Torque.

Knowledge: Motivation to work in the rehabilitation field. Programming knowledge in C++. Ability to autonomously learn to use the Torque Engine.

Keywords: Motor and Cognitive Rehabilitation

Supervisor: Armin Duff

Coordinator: Paul F.M.J. Verschure

Contact (email): armin.duff(at)gmail.com

Brain-Computer Interfaces

Connecting your Brain: Playing Interactive Scenarios using BCI

Keywords: Cognitive Rehabilitation

Description: Lesions in the brain following a stroke or a traumatic brain injury often lead to disorders in memory and in the ability to plan and draw strategies. In a clinical context, engaging tasks that promote the recovery of these lost capabilities are still lacking. Some of these patients do not have any mobility on the affected motor areas, but some of their cortex is still functional. Therefore, we propose to use imagination as a method to train and activate these motor areas while a Brain Computer Interface reads out the correlated brain activity and the patient plays interactive scenarios directed towards these specific motor deficits. The tool to be used is a BCI interface (gMobilab) from Guger technologies.

Tasks: Propose and develop a BCI system to read intended movements and link it to the scenarios for the cognitive rehabilitation

Knowledge: Motivation to work in the rehabilitation field. Capability to autonomously learn to use and edit the NeuroVR or torque software. Some programming knowledge would be appreciated

Supervisor: Arnau Espinosa

Coordinator: Paul F.M.J. Verschure

Contact (email): dux.espinosa(at)gmail.com

Use brain computer interface (BCI) for non-verbal communication in a virtual environment

Keywords: Non-verbal communication, BCI, HCI

Description: Brain computer interfaces (BCI) provide whole new ways to interface humans to technology. This is especially the case when the technology is the multi-modal mixed reality space XIM. In this project a BCI system based on the electro-encephalogram, is used to control the behavior of an Avatar. This behavior might include facial expression and posture. Such an interface would be of immediate use to impaired patients, allowing them a rich, and active social participation in the virtual community developed within the BrainAble project (http://www.brainable.org). Key element of the project is collaboration with other partners of the Presenccia project, namely from the technical university of Graz, Austria

Type: Research

Knowledge: Cognitive science, programming

Supervisor: Arnau Espinosa, Ulysses Bernardet

Coordinator: Paul F.M.J. Verschure

Contact (email): ulysses.bernardet(at)upf.edu

Cerebellum

A digital neuromorphic implementation of cerebellar association learning

Keywords:  computational neuroscience, robotics, neural models, robotics, FPGA, neuromorphic engineering

Description: It is well known that the cerebellum is the part of the brain mainly responsible for the acquisition of a response in a classical conditioning experiment. Classical conditioning is the most elementary studied form of associative learning. In classical conditioning, the repeated presentation of a previously neutral stimulus (conditioned stimulus, e.g. a tone) jointly with an aversive one (unconditioned stimulus, e.g. an airpuff), leads an animal to associate the conditioned stimulus with a motor response (conditioned response) protecting him against the nocive stimulus. In our group we have been studying for nearly a decade the neural substrates of classical conditioning via neural network modelling and robotics studies (http://www.ini.ethz.ch/~connie/publications/CerebellumInAction.pdf). We have implemented various neural network models of the cerebellum able to reproduce the phenomenon of classical conditioning and able to control an epuck robot (http://www.e-puck.org/).The next step would be to implement the cerebellum in silicon, that is on a microchip. This would allow to implant the artificial cerebellum directly on the robot. We are currently collaborating with the ISS of Rome at the implementation of an aVLSI circuit reproducing the dynamics of the cerebellum. In parallel, we also aim at implementing the cerebellum on a more flexible digital platform, like for instance FPGAs. Our implementation would go along the lines of neuromorphic engineering philosophy: "Neuromorphic engineering is a new interdisciplinary discipline that takes inspiration from biology, physics, mathematics, computer science and engineering to design artificial neural systems, such as vision systems, head-eye systems, auditory processors, and autonomous robots, whose physical architecture and design principles are based on those of biological nervous systems" (http://en.wikipedia.org/wiki/Neuromorphic).

Tasks: 

  1. Implementation of a general purpose neural network simulator in HDL

  2. Testing of the performances of the simulator on mid-size neural network models

  3. if successful in the design, implementation of the cerebellum on silicon

  4. implement the control layer of a robot interfacing it with such a silicon cerebellum

Knowledge: Signal processing, computer science, electronics, physics, mathematics or biology and hardware description language (HDL). 

Supervisor: Ivan Herreros

Coordinator: Paul F.M.J. Verschure

Contact (email): ivanherreros(at)gmail.com

 

The computational neuroscience of learning: the Pons and the cerebellum

Keywords:  computational neuroscience, robotics, neural models, neuromorphic engineering

Description: Learning of discrete motor responses, such as the Pavlovian eye-lid response, has been extensively studied over many decades. The cerebellum as been identified as the putative site responsible for this type of associative learning. In our group we have been studying for nearly a decade the neural substrates of classical conditioning via neural network modeling and robotics studies (http://www.ini.ethz.ch/~connie/publications/CerebellumInAction.pdf). We have implemented various neural network models of the cerebellum able to reproduce the phenomenon of classical conditioning and able to control a real time robot (http://www.e-puck.org/). The Pontine Nuclei are supposed to be the site of convergence and integration of different somatosensory information (auditory, visual, tactile) from the cerebral cortex of one hemisphere to the contralateral cerebellar hemisphere. The objective of this project is to integrate our current model of the cerebellum with a biological plausible mechanism able to describe the role of the PN in the circuitry.

Tasks: 

  1. Review of the current literature on the Cerebellum focusing on

  2. Pontine Nuclei Implementation of a neural model of the PN

  3. Testing and evaluation of the model with simulations and on real time robot devices

Knowledge: Interest in: cognition, neuronal computation and biology, basic mathematical knowledge's and programming

Supervisor: Ivan Herreros

Coordinator: Paul F.M.J. Verschure

Contact (email): ivanherreros(at)gmail.com

 

Neural model of the cerebellar control of smooth eye pursuit

Keywords: computational neuroscience, robotics, neural models, robotics, interaction

Description: 
The Cerebellum is a sub-cortical brain structure necessary for smooth motor control. It is also well known that the cerebellum is responsible for the acquisition of a response in a classical conditioning experiment. Classical conditioning is the most elementary studied form of associative learning. In classical conditioning, the repeated presentation of a previously neutral stimulus (conditioned stimulus, e.g. a tone) jointly with an aversive one (unconditioned stimulus, e.g. an airpuff), leads an animal to associate the conditioned stimulus with a motor response (conditioned response) protecting him against the nocive stimulus. In our group  we have been studying for nearly a decade the neural substrates of classical conditioning via neural network modelling and robotics studies (http://www.ini.ethz.ch/~connie/publications/CerebellumInAction.pdf). We have implemented various neural network models of the cerebellum able to reproduce the phenomenon of classical conditioning and able to control an epuck robot (http://www.e-puck.org/).

The goal of the Master Student for this projects would be to adapt this model to a set up where both stimuli and responses care continuous. A typical case-study for such kind of tasks is the so-called eye-pursuit task in which a subject is trained to follow a moving target with its gaze. For more information http://www.sciencedaily.com/releases/2007/07/070702145312.htm)
 

Tasks: The work is divided in two stages, a first one more theoretical in which the model must be adapted to the new scenario, and a second one in which the performance should be demonstrated using real world platforms. Such platforms could be a pan-tilt camera, XIM gazers or a real-time robot. 

The work of the student will be closely supervised in the first more theoretical part of the project, where the alumn would have to design and implement in IQR (and possibly Matlab) the underlying neural network model. Afterwards, the student is expected to apply its own creativity for the generation of the real-world experiments.

Knowledge: Programming skills, interaction design.

Supervisor: Ivan Herreros

Coordinator: Paul F.M.J. Verschure

Contact (email): ivanherreros(at)gmail.com 

 

Kinect delay adaptation. Execution, learning and the role of the cerebellum.

Description: Despite the fantastic achievement that it represents, Kinect, the new Microsoft camera controller for Xbox, has a noticeable lag (namely, 160 ms). Inasmuch, the human player has to adapt to such a delay and time its actions ahead of its desired execution. This might seem a drawback for the recreational user but it represents a great opportunity for experimenters: Microsoft's Kinect allows the player to control an avatar using all his/her limbs and head; therefore, if we play a game such as beach volley, long jump or hurdling, we are adding a general delay to our entire motor system control.

We aim to investigate the extent and nature of an eventual adaptation occurring after prolonged exposure to Kinect. If such an adaptation occurs, we want to tell whether it affects the execution of learned movements, the acquisition of new stimulus-response associations, or whether it extends differently to conscious or automatized actions. By answering these questions we will also implicitly identify the brain structures responsible for our successful Kinect playing (ie: motor cortex or cerebellum). Any positive result will shed light on the versatility of our brain as motor controller, thus defining our hope of an easy transition to a fore-coming cyborg incarnation. For this research we will employ paradigms from experimental psychology, such as eye-blink conditioning, in addition to the commercial games provided with the Xbox Kinect.

Tasks: Review on motor adaptation literature. Besides carrying out the experiments, the student will be involved in some of the following tasks: experimental design(s), programming of the experiment protocol(s), data analysis.

Knowledge: Basic programming skills or the desire to acquire them.

Type: Research

Keywords: experimental psychology, neuroscience, cerebellum, kinect

Supervisor: Ivan Herreros

Coordinator: Paul F.M.J. Verschure

Contact (email): ivanherreros@gmail.com

Rehabilitation

Interactive Scenarios for the Rehabilitation of Memory and Planning Disorders

Keywords:

Description: Lesions in the brain following a stroke or a traumatic brain injury often lead to disorders in memory and in the ability to plan and draw strategies. In a clinical context, engaging tasks that promote the recovery of these lost capabilities are still lacking. Therefore, we propose the creation of versatile interactive scenarios directed towards these specific cognitive disorders. The tool to be used to develop the interactive scenarios is Unity. In this project basic scenarios will be developed and tested on different groups of users including neurological patients

Type: Research

Tasks: Literature review on memory and planning disorders following brain damage. Propose and develop the scenarios for the cognitive rehabilitation. Assess impact

Knowledge: Motivation to work in the rehabilitation field. Capability to autonomously learn to use and edit the NeuroVR or torque software. Ability do plan and design relevant scenarios

Keywords: Cognitive Rehabilitation

Supervisor: Belen Rubio/Armin Duff

Coordinator: Paul F.M.J. Verschure

Contact (email): armin.duff(at)gmail.com

 

Multi-modal cognitive neuro-rehabilitation systems

Description: Cerebral palsy patients can profit tremendously from intense training. The best way to drive this is still under discussion. However, several approaches seem to be effective such as intensive training, tasks directed towards specific deficits, online delivery of feedback, etc.. All these features can be put together by means of interactive scenarios that engage the subjects in specific tasks. We propose the creation of a scenario that integrates multi-model augmented feedback with particular focus on the rehabilitation of children with cerebral palsy. This scenario is to be developed using Unity

Type: Research

Tasks: Literature review on cerebral palsy, traumatic brain injury and recovery mechanisms. Design, implement and analyze rehab systems and user response

Knowledge: Motivation to work in the rehabilitation field. Programming knowledge in C++. Ability to autonomously learn to use the Torque Engine.

Keywords: Motor and Cognitive Rehabilitation

Supervisor: Cristina Campillo / Armin Duff

Coordinator: Paul F.M.J. Verschure

Contact (email): armin.duff(at)gmail.com

 

Psychophysics and stroke

Description: The brain of stroke patients undergoes massive reorganization. It is thus reasonable to suspect that this will include functional changes in modalities that are not directly affected by the stroke. In order to assess this hypothesis in more detail we want to perform psychophysical studies with stroke patients to assess how their ability to process the visual world is affected by their stroke.

Type: Research

Tasks: Literature review on stroke and visual psychophysics. Design, implement and analyze experiments

Knowledge: Motivation to work in the rehabilitation field. Programming knowledge in ePrime/Matlab/C++. Ability to autonomously learn to use these tools.

Keywords: Motor and Cognitive Rehabilitation

Supervisor: Cristina Campillo / Armin Duff

Coordinator: Paul F.M.J. Verschure

Contact (email): armin.duff(at)gmail.com

 

A new generation of music therapy systems

Description: Music can be a powerful stimulus to affect states of the brain. In this project we want to use existing interactive music composition tools developed at SPECS to investigate the impact of these tools on the states of the brain of coma patients. We will focus on issues of entrainment and expression.

Type: Research

Tasks: Literature review on coma, entrainment and music. Design, implement and analyze experiments.

Knowledge: Motivation to work in the rehabilitation field. Programming knowledge in ePrime/Matlab/C++. Ability to autonomously learn to use these tools.

Keywords: Motor and Cognitive Rehabilitation

Supervisor: Sylvain Legroux

Coordinator: Paul F.M.J. Verschure

Contact (email): sylvain.legroux(at)fulbrightmail.org

Exciting alternatives

Mapping human movement onto a humanoid robot

Description: A 3 dimensional tracking system (based on the XBOX kinect) will be implemented to localize in 3D information about  moving targets. The system will be used to track specific points of the body to reconstruct human body movements. These movements can be used to drive a humanoid robot (ROBONOVA-1) to reproduce the same movements

Type: Research

Tasks: The student will work on the reconstruction of 3D coordinates given the kinect. The software will track specific points of interest and compute its 3D coordinates. Additionally, the student will have to generate an inverse kinematic model of a humanoid in order to reconstruct tracked joint angles, that can eventually be used afterwards to map onto a small humanoid robot that has 16 degrees of freedom. In addition, we will use this system to track the flight performance of insects and robots. In addition, the project can be augmented with the Emotiv Brain-Computer interface system to control the robot.

Knowledge: Programming C++ and advanced mathematical background.

Keywords: tracking, data fusion, robotics

Supervisor: Armin Duff

Coordinator: Paul F.M.J. Verschure

Contact (email): armin.duff(at)gmail.com

 

Chaos based search and attention in a neuronal model of consciousness

Description: A while back the idea was proposed that chaos can be used as a method to search complex state spaces (Verschure, 1991, Chaos based Learning, Complex Systems). In this project we would like to generalize this idea to biologically constrained models of the cerebral cortex focusing on the idea of the Global Workspace (B Baars. 2005. Global workspace theory of consciousness: toward a cognitive neuroscience of human experience. Progress in Brain Research) and the Temporal Population Code (R Wyss, P Konig, Paul F M J Verschure.

2003. Invariant representations of visual patterns in a temporal population code. Proceedings of the National Academy of Sciences)

Type: Research

Tasks: The student will work on understanding the basic literature and on implementing a large scale simulation using the simulation environment IQR. In addition, the system will be analyzed using state of the art data visualization methods available in XIM.

Knowledge: IQR, Programming C++ and advanced mathematical background.

Keywords: Neuronal simulation, search, consciousness

Supervisor: Andre Luvizotto, Paul Verschure

Coordinator: Paul F.M.J. Verschure

Contact (email): luvizotto(at)gmail.com

 

A neuronal model of consciousness

Description: The notion of a Global Workspace (B Baars. 2005. Global workspace theory of consciousness: toward a cognitive neuroscience of human experience. Progress in Brain Research) proposes that consciousness emerges out of the critical interaction between a large number of neuronal systems or Global Workspace (GW). In this project we want to use components of existing models to create such a GW with a particular focus on the interaction between the thalamus and the cerebral cortex. The model will be applied to different kinds of robot tasks in the area of attentional processing and decision making.

Type: Research

Tasks: The student will work on understanding the basic literature and on implementing a large scale simulation using the simulation environment IQR. In addition, the system will be analyzed using state of the art data visualization methods available in XIM.

Knowledge: IQR, Programming C++ and advanced mathematical background.

Keywords: Neuronal simulation, search, consciousness

Supervisor: Andre Luvizotto, Paul Verschure

Coordinator: Paul F.M.J. Verschure

Contact (email): luvizotto(at)gmail.com

 

Recovering history at the Bergen Belsen memorial

Description: The Bergen Belsen concentration camp is one of the symbols of the barbarism of Nazism and the excesses of intolerance and ignorance. Shortly after its liberation the camp was burned down so that now, about 65 years later, no physical remains can be witnessed on its physical location. In order to allow current visitors to the memorial site to understand the layout of the camp and appreciate its historical significance we want to use modern interactive technologies to allow an augmented reality experience of the camp site using iPhones.

Type: Research

Tasks: The student will work on implementing the complete system and to perform field tests with it. This system will be based on a prototype already developed at SPECS.

Knowledge:  Programming, graphics and interaction design.

Keywords: History, interaction, exhibition, augmented reality

Supervisor: Melle hofman, Paul Verschure

Coordinator: Paul F.M.J. Verschure

Contact (email): paul.verschure(at)gmail.com

 

Develop models of immersion and presence using small hand-held devices augmenting real world environments.

Description: The increase usage of hand-held devices, such as iPhone and iPad, together with a constant availability of fast network connections raises the question how these new devices can be used in augmenting the experience of users in open air exhibitions or museums. This master project aims at developing prototypes using these kinds of hand held devices. The objective is to deliver made-to-measure content from large data-sets to tell stories about historical events on location. The interface should provide the content in such a way that the end user gets the feeling of being present in that place in time. Performance of the interface has to be measured and quantified.

Type: Research

Tasks

  • Develop Prototypes of Interfaces for Hand held devices integrating different types of content.

  • Direct a small test scenario.

  • Collect data to measure the feeling of presence and/or immersion. 

Knowledge

  • Programming Skills (preferably Objective-C)

  • Interaction Design

Supervisor: Melle hofman, Paul Verschure

Coordinator: Paul F.M.J. Verschure

Contact (email): paul.verschure@gmail.com

 

Model of artificial emotions implemented in the humanoid robot iCub

Keywords: Artificial emotions, non-verbal communication, humanoid robot, perception

Description: Non-verbal communication is a fundmental aspect of social interaction. The understanding of the psychological and behavioral mechanism are essential for the construction of a realistic model of artificial emotions. The project focus on the implementation of an artificial emotional model in the humanoid robot iCub. The architecture combines perception of human behavior, the processing of the stimulus stream that controls the elicitation of appropriate behavior. The performance of the model will be evaluated in an experiment where humans will free interact with the robot.

Knowledge: Basic programmin skills

Supervisor: Martin Inderbitzin

Coordinator: Paul F.M.J. Verschure

Contact: martin.inderbitzin(at)upf.edu