Odsek za avtomatiko, biokibernetiko in robotiko





Seznam napovedanih predavanj (povzetki so pod seznamom):

 

Predavanja - sezona 2010/2011

------------------------------------------------------------------------------------------------------------------------------

30.06.2011, 13h: Lado lenart

SimMechanics ( v sklopu SIMULINK ) je orodje za racunanje direktne in inverzne dinamike odprtih in zaprtih kinematskih

verig z moznostjo linearizacije in vizualizacije. V seminarju se bomo dotaknili Hamiltonovega principa. Glavni del bo obsegal primerjanje klasicnih resitev s SimMechanics modeli. Posebej bodo obravnavani bloki v SimMechanics, ki omogocajo procesiranje omejitev v gibanju kinematske verige.

SimMechanics ( in SimuLink ) is a tool for forward and inverse dynamics calculation for open and closed kinematic chains with linearization and animation features. The Hamiltonion problem will be touched. The main part of the lecture is the comparison of classic solutions with SimMechanics models. The SimMechanics blocks follow, which enable the processing of constraints in kinematic chains.

------------------------------------------------------------------------------------------------------------------------------

09.06.2011, 12h: Joan Saez

Experiments in Cooperative Human Multi-Robot Navigation.

The problem considered is how a group of autonomous mobile robots and a human can move coordinately in a real-world implementation, i.e. in a large warehouse to assist fire-fighters in the event or danger of a fire. The key problem to be solved in the context of multiple robot is the inclusion of a human in a real multi-robot system and consequently the multiple robot motion coordination. A set of performance metrics is presented (system efficiency and percentage of time in formation) and a novel flexible formation definition whereby a formation control strategy both in simulation and in real-world experiments of a human multi-robot system is presented. The formation control proposed is stable and effective by means of its uniform dispersion, cohesion and flexibility.

-----------------------------------------------------------------------------------------------------------------------------

02.06.2011, 13h: Nejc Likar

Kuka Lightweight robot - From command to move

In the seminar, I will present the newest robot in the lab - Kuka lightweight arm. In the beginning I will briefly describe the robot and give some information of the arms. This will include also the introduction to Fast Research Interface(FRI) and KRL.

Later, I will present the real-time server which is keeping the FRI communication alive in one thread, and communicating with remote client in the other thread. 

On the remote side, the user is connected to the server and controlling the robot with Matlab / Simulink / C library.  At the end of the presentation we will have a discussion about the improvements of the system.

-----------------------------------------------------------------------------------------------------------------------------

19.05.2011, 13h: David Schiebener

Generation of view sequences for continuous localisation of multiple objects on an active humanoid robot head

-----------------------------------------------------------------------------------------------------------------------------

12.05.2011, 13h: Andrej Gams

Microsoft Kinect and IJS Real-time Kinect server

In the talk I will present the use of the Microsoft Kinect Sensor for controling robotic mechanisms. I will give some information on the device, drivers and OpenNI Library we use for tracking of people. I will also present the IJS real-time Kinect server, which sends the received data to client computers over the IDP protocol. I will also give a demo on the use for controlling the arms of a HOAP-3 humanoid robot, where I will mostly discuss on the transformation of the measured angles into the robot joint angles.

------------------------------------------------------------------------------------------------------------------------------

05.05.2011, 13h: Borut Lenart

Predstavitev CAD programske opreme Pro-engineer za modeliranje in preračunavanje strojnih elementov

------------------------------------------------------------------------------------------------------------------------------

31.03.2011, 13h: Goran Škorja

V predstavitvi bom okvirno opisal pristop za sintezo gibanja humanoidnega robota, ki bazira na osnovi človekovega vizualno motoričnega učenja. Ideja je, da bi se humanoidni robot uporabljal kot neko novo orodje, ki bi ga človek intuitivno uporabljal. Ko bi se človek naučil upravljati robota brez nekih težav (kot na primer voziti avtomobil) bi se lahko robotovo želeno obnašanje doseglo skozi prakso oz. izkušnje. Uspežno izvajanje naloge človeka preko humanoidnega robota pravzaprav pomeni, da so ustrezni ukazi robota izvedeni in odkriti s strani človeka. Da bi človek lahko ustrezno upravljal humanoidnega robota pa potrebuje nekakšno povratno informacijo o stanju robota. V ta namen bom prikazal poseben eksperimentalni pristop z uporabo pritiskovne plošče in Stewartove platforme.
Ob tej priliki bom na predstavitvi tudi prikazal uporabo Optotrak sistema ter pripadajočo programsko opremo za zajemanje gibanja v realnem času.

------------------------------------------------------------------------------------------------------------------------------

17.03.2011, 10h: Mojca Amon

Metabolic responses to environmental hypoxia

The observation that high altitude exposure may lead to considerable weight loss, has led to the suggestion that it might be beneficial to incorporate hypoxic training in weight management programs for obese individuals. Netzer et al. (2008) demonstrated that mild physical exercise in normobaric hypoxia (15% O2) for 8 wks caused a significantly greater weight loss in obese persons than exercise in a sham hypoxic environment. Recently, Lippl et al. (2010) demonstrated that hypobaric hypoxia seems to play a major role in weight loss after prolonged altitude exposure, although the physiological mechanisms remain unclear.

To our knowledge, no studies have been carried out to date, regarding the treatment of obesity under 10-days hypoxic normobaric people-friendly environment without exercise training. The mechanism of weight loss is not entirely due to an imbalance between energy intake and expenditure.

Tommorow on the Department meeting,  several factors that plays a role in weight regulation and have been subject of interest of our research group in »Planica« study, will be presented.

References:

  1. Netzer et al., 2008. Low intense physical exercise in normobaric hypoxia leads to more weight loss in obese people than low intense physical exercise in normobaric sham hypoxia. Sleep Breath. 12(2):129-134.
  2. Lippl et al., 2010. Hypobaric Hypoxia Causes Body Weight Reduction in Obese Subjects. Obesity.18(4):675-81.

 

------------------------------------------------------------------------------------------------------------------------------

03.03.2011, 13h: Leon Žlajpah

Introduction of Statflow for Robot Control

Stateflow extends MATLAB/Simulink with a design environment for developing state machines and flow charts. Stateflow provides the language elements required to describe complex logic in a natural, readable, and understandable form. It is tightly integrated with MATLAB and Simulink, providing an efficient environment for designing embedded systems that contain control, supervisory, and mode logic. In the lecture fa short introduction of Stateflow and its basic functionality will be presented. Then the application of Statefow in the robot control systems will be given. As an illustrative example the design of a control system for a robot (Mitsubishi PA10) with a vision system playing minigolf will be presented.

 

------------------------------------------------------------------------------------------------------------------------------

17.02.2011, 13h: Jan Babič

Sensorimotor Learning - Between Robots and Humans

In this talk, I will introduce a concept of obtaining complex robot motions by implementing the sensorimotor learning capabilities of humans. The idea is to consider the target robotic platform as a tool that can be controlled by a human. Provided with an intuitive interface between the human and the robot, the human learns to perform a given task using the robot. After the learning, the skilled control of the robot by the human provides data that are used for construction of an autonomous controller that controls the robot without any guidance of the human.

 

Besides, I will present some ideas on the neural correlates of human-in-the-loop robot control and show how the interfaces built for robot skill synthesis can also be used in the reverse direction for probing motor control mechanisms employed by the central nervous system.

 

------------------------------------------------------------------------------------------------------------------------------

10.02.2011, 13h: Miha Deniša

Učenje osnovnih gibov v robotiki

Na seminarju bom predstavil svoje diplomsko delo, ki obravnava nadzorovano in nenadzorovano učenje periodičnih in neperiodičnih osnovnih gibov iz domačega okolja.

Za diplomsko delo smo osnovne gibe zapisali kot dinamične gibe (DMP), saj je  tak zapis primeren za hrambo in omogoča enostavno modulacijo gibov pred ali med samim izvajanjem. Zaradi razmeroma majhnega števila parametrov, ki določajo obliko trajektorije, je zapis z dinamičnim generatorjem gibov priročen za nenadzorovano učenje.

Pri nadzorovanem učenju so bili demonstrirajoči gibi zajeti z vidom in nato, pretvorjeni v dinamični zapis, izvedeni v simulacijskem okolju. Naučeni periodični gib brisanja površine z gobo in neperiodični gib natakanja tekočine v kozarec smo izvedli tudi na realnem sistemu z robotsko roko Pa 10.

Za nenadzorovano učenje je bilo izbrano spodbujevalno učenje, bolj natančno gradientna metoda strategije. Gradient pa se je določal z metodo končnih razlik. V simulacijskem okolju so bili preizkušeni različni pristopi učenja in določanja nagrade. Sledila je izvedba učenja natakanja tekočine iz plastenke v kozarec na realnem sistemu.

Ključne besede: osnovni gib, nadzorovano učenje, nenadzorovano učenje, dinamični generator gibov, spodbujevalno učenje, gradientna metoda strategije, metoda končnih razlik.

 

------------------------------------------------------------------------------------------------------------------------------

03.02.2011, 13h: Igor Mekjavić

PHYSIOLOGICAL ASPECTS OF EXPOSURE TO HYPERBARIC ENVIRONMENTS: with particular reference to thermal balance
In contrast to the normobaric environment, the hyperbaric environment poses a major challenge to all physiological systems. Ambient pressure (PB) per se modifies the functioning of neural structures, a phenomenon termed the  ‘high pressure nervous syndrome’, whereas its rate of change (compression/decompression) may cause barotrauma and formation of  bubbles in tissues, the consequences of which can be negligible, or may lead to pain, paralysis and ultimately death. At elevated PB, the increased partial pressure of the constituent gases of air become narcotic (nitrogen, N2) and toxic (oxygen, O2; carbon dioxide, CO2), requiring substitution of the inert gas component N2, with less narcotic gases (e.g. helium, He), and precise control of the inspired partial pressures of O2 and CO2 (PO2, PCO2). The present review will provide a primer in diving physiology, but will focus specifically on the problem of thermal balance in hyperbaric environments. It will introduce a strategy by which the functioning of physiological control systems such as temperature regulation may be investigated.
    The principle aim of the thermoregulatory system is the maintenance of the temperature of the body’s internal environment. It regulates this modality (temperature) by initiating effector mechanisms for heat production and heat loss in response to temperature information provided by the modality specific (temperature) sensors. However, regulation of body temperature is dependent not only on thermal factors (modality specific factors), but also on factors unrelated to temperature (modality non-specific factors). Thus, to understand the challenges posed by the hyperbaric environment on physiological systems, it is necessary to understand how the numerous factors associated with the hyperbaric environment affect a particular physiological system, such as temperature regulation. Thus, within the framework of body temperature regulation, where temperature is the modality sensed and regulated, the factors associated with the hyperbaric environment, that can modify this regulation are PB, PN2, PHe, PO2, PCO2. These are then the environmental “modality non-specific factors”, also termed the environmental non-thermal factors in the context of the thermoregulatory system. In addition to the environmental modality non-specific factors, there are also physiological “modality non-specific factors”, which contribute to the regulation of body temperature. These include factors such as hydration status, plasma glucose, fatigue, ageing, etc. To investigate the manner in which an ambient (ie. hyperbaric) modifies the regulation of a particular modality of the body’s internal environment (ie. temperature), we must consider the contribution of the modality specific (ie. thermal) and modality non-specific (ie. nonthermal) environmental and physiological factors, as well the interaction of individual regulatory systems (ie. thermal and cardiovascular).
    The presentation will summarise the manner in which solutions to practical questions regarding the effect of the hyperbaric environment on physiological systems, has led to a better understanding of the mammalian temperature regulation. In particular, how this knowledge is being incorporated into neural models, which may eventually assist biometeorologists in better forecasting the influence of normobaric weather conditions on health, well-being, and in some cases survival.

------------------------------------------------------------------------------------------------------------------------------

20.01.2011, 13h: Tadej Petrič

Inteligentno vodenje robotov z uporabo nelinearnih oscilatorjev

Predavanje bo namenjeno predstavitvi moje teme doktorske disertacije z naslovom »Inteligentno vodenje robotov z uporabo nelinearnih oscilatorjev«. V uvodu bom najprej podal krajši pregled raziskav s področja vodenja ritmičnih nalog. Predstavil bom dosedanje delo in pridobljene rezultate ter nakazal smer nadaljnjih raziskav, kjer bom izpostavil pričakovane prispevke k znanosti.

 

------------------------------------------------------------------------------------------------------------------------------

06.01.2011, 13h: Denis Forte

Učenje robotov s posnemanjem

Na seminarju bom predstavil temo svoje doktorske disertacije »Učenje robotov s posnemanjem«. V uvodu bom najprej podal krajši pregled raziskav s področja strojnega učenja in metod za učenje robotov ter iz tega izpostavil ciljne izvirne prispevke k znanosti. Večji del predstavitve se bo nanasal na do sedaj opravljeno delo in dosežene rezultate ter podrobnejšo razlago problematike posameznih prispevkov.

Možni prispevki k znanosti:

1. Metoda posploševanja v realnem času.

2. Statistično izbiranje značilk, ki karakterizirajo nalogo (do sedaj so bile te značilke vključene ročno).

3. Povezava z reinforcement learning (ojačenim učenjem), parametrični zapis naloge.

 

------------------------------------------------------------------------------------------------------------------------------

23.12.2010, 13h: Tadej Debevec

Can subsequent hyperoxia and hypoxia augment erythropoietin synthesis?

Improved oxygen flux as a consequence of increases in Hematocrit and total Hemoglobin mass is among the main mechanism accountable for improvements in performance following different hypoxic training modalities. These hematological benefits are induced by augmented erythropoiesis, induced by stimulation of hematopoietic hormone erythropoietin (EPO). While renal tissue hypoxia is an established trigger of de novo EPO production, it has recently been suggested that EPO synthesis may be stimulated also by relative hypoxia, induced by an acute exposure to hyperoxia followed by a return to normoxia. This phenomenon has been termed the “normobaric oxygen paradox”. The purpose of the study was to evaluate whether such relative hypoxia, induced by acute and successive exposures to hyperoxia and hypoxia can stimulate EPO synthesis. The findings of this study do not support the theory of a “normobaric oxygen paradox”. Therefore, combining hypoxic and hyperoxic breathing does not seem to be a promising means for erythropoiesis augmentation.

 

------------------------------------------------------------------------------------------------------------------------------

02.12.2010, 13h: Mitja Babič

Vodenja elektroelastičnih polimernih aktuatorjev

V predavanju bom predstavil raziskave, ki sem jih opravljal za doktorat, slednje se nanasajo na dokaj neraziskano podrocje vodenja elektroelasticnih polimernih aktuatorjev. Predstavil bom problematko vodenja in elektronskih gonilnikov dielektricnih elastomerskih aktuatorjev, slednja zahteva razvoj elektronskih naprav visokih napetosti in srednjih tokov ter uporabo adaptivnih algoritmov vodenja. 

Prikazana bo nova in ucinkovitejsa strategijo vodenja, ki temelji na elektroniki gonilnika s topologijo zapornega pretvornika z galvansko locenim izhodom. Na koncu bom podal nekaj idejnih zasnov za nadaljni razvoj.

 

------------------------------------------------------------------------------------------------------------------------------

18.11.2010, 13h: Andrej Gams, Nejc Likar, Tadej Petrič

ROS: Robot Operating System

V predavanju bomo najprej na kratko predstavili kaj je ROS, kdo ga uporablja in razvija in za katere stvari je namenjen. Razložili bomo osnove delovanja in komunikacijo med senzorji in robotom s pomočjo ROS node-ov (Andrej in Nejc). Na koncu bomo demonstrirali tri stvari: premikanje virtualnega robota v Gazebo simulatorju, vključitev knjižnice openCV za razpoznavanje objektov ter vizualizacijo senzorskih informacij v okolju rviz.

Predavanje sloni na predavanjih in vajah s predstavitvene delavnice v Muenchnu (link, wiki) novembra 2010.

 

------------------------------------------------------------------------------------------------------------------------------

11.11.2010, 13h: Bojan Nemec

Introduction to reinforcement learning in robotics

Predavanje je namenjeno uvodnemu spoznavanju problematike vzpodbujevalnega ucenja s poudarkom na uporabi v robotiki. V predavanju se bomo seznanili z osnovnimi gradniki nenadzorovanega ucenja, predstavili bomo SARSA in Q algoritme ucenja ter pokazali, kako se algoritmi ucenja, ki so namenjeni obravnavi diskretnih dogodkov uporabijo tudi za obravnavo casovno zveznih sistemov.  Nadalje bomo pokazali se nekatere postopke ucenja, ki so namenjeni casovno zveznim sistemom. Podali bomo nekaj primerov uporabe predstavljenih algoritmov v robotiki.

 

------------------------------------------------------------------------------------------------------------------------------

04.11.2010, 13h: Miro Vrhovec

Tehnologija višinskih sob

V nekaterih prejšnjih odsečnih  seminarskih predstavitvah so kolegi predstavili svoje raziskave v hipoksičnem okolju. V svoji predstavitvi z naslovom Tehnologija višinskih sob bom predstavil opremo, s katero lahko v normalnem bivalnem prostoru simuliramo razlićne nadmorske višine s pomočjo spreminjanja parcialnega tlaka kisika. Prikazani bodo tudi sistemi, ki jih je Institut Jožef Stefan instaliral v Sloveniji  skupaj z dobaviteljem opreme Bcat iz Nizozemske.

Predstavitev bo potekala v slovenskem jeziku.

Devices for altitude simulation by means of changing Oxygen partial pressure  in normal living facilities

In some previous department presentations researches in hypoxic environment has been introduced. In presentation with title Devices for altitude simulation by means of changing Oxygen partial pressure  in normal living facilities will be presented. Hypoxic systems installed in Slovenija by Institute Jozef Stefan and Dutch company Bcat will be shortly introduced.

Presentation will be given in Slovene language.

 

------------------------------------------------------------------------------------------------------------------------------

21.10.2010, 13h: Igor Kovač 

Industrijska robotika, fleksibilna avtomatizacija in proizvodna merilna tehnika

V seminarju želim zelo na kratko predstaviti moje delo na področju industrijske robotike, fleksibilne avtomatizacije in merilne tehnike.

Pri tem bodo poudarki dani na robotsko podprtem merjenju (Robot Aided Measurement - RAM), robotsko podprti rekonfiguraciji (Robot Aided Reconfiguration - RAR), robotiki v gradbeništvu in robotiki v proizvodnih sistemih. Predstavljeni bodo posamezni znanstveni in aplikativni projekti z rezultati in podane potencialne možnosti za prihodnje delo in iskanje sinergij z ostalimi projekti na odseku.

 

------------------------------------------------------------------------------------------------------------------------------

14.10.2010, 13h: Aleš Ude

EU cognitive systems projects IntellAct and Xperience

IntellAct addresses the problem of understanding and exploiting the meaning (semantics) of manipulations in terms of objects, actions and their consequences for reproducing human actions with machines. It comprises three building blocks: (1) Learning: Abstract, semantic descriptions of manipulations are extracted from video sequences showing a human demonstrating the manipulations; (2) Monitoring: In the second step, observed manipulations are evaluated against the learned, semantic models; (3) Execution: Based on learned, semantic models, equivalent manipulations are executed by a robot. The analysis of low-level observation data for semantic content (Learning) and the synthesis of concrete behavior (Execution) constitute the major scientific challenge of IntellAct.

The goal of Xperience project is to replicate - in a technical system - the ability of humans to rapidly create new concepts and react to unanticipated situations using their experience. “Imagining” and “internal simulation”, hence generative mechanisms which rely on prior knowledge are employed to predict the immediate future and are key in increasing bandwidth and speed of cognitive development. Current artificial cognitive systems are limited in this respect as they do not yet make efficient use of such generative mechanisms for the extension of their cognitive properties. The Xperience project will address this problem by structural bootstrapping, an idea taken from child language acquisition research. Structural bootstrapping is a method of building generative models, leveraging existing experience to predict unexplored action effects and to focus the hypothesis space for learning novel concepts.

 

------------------------------------------------------------------------------------------------------------------------------

12.10.2010, 13h: Eva Stergaršek Kuzmič

Odkrivanje novih predmetov in grajenje njihovih modelov z aktivnim robotskim sistemom

Bistven element kognitivnih robotskih sistemov je zmožnost spoznavanja predmetov, s katerimi robot pride v stik prvič, in učenja predstavitev za takšne predmete, ki so uporabne za kasnejše razpoznavanje. Naša predpostavka je, da je odkrivanje novih predmetov izvedljivo na učinkovit način z aktivnim robotskim sistemom, ki poleg vizualnega sistema za zajemanje slik prizorov vsebuje elemente za manipulacijo delov prizora. Manipulacija služi za ločitev predmeta iz okolice in za akumulacijo objektnih značilnic, na osnovi teh pa je mogoče modeliranje predmeta in ugotavljanje njegove funkcionalnosti.
V tem delu podajamo postopek odkrivanja in segmentacije predmetov s pomočjo kombinirane rabe vizualnih in manipulativnih funkcij kognitivnega robotskega sistema. Postopek smo preverili z več eksperimenti z realnim robotskim sistemom.
V predlaganem pristopu je postopek spoznavanja predmeta sestavljen iz treh procesov. Prvi proces predstavlja oblikovanje hipotez o obstoju predmetov. Oblikovanje hipotez sestavlja več podprocesov: izločanje in gručenje točk vizualnih značilnic, določanje 3-D lokacij in iskanje ravnin.
Drugi proces definira optimalne manipulativne akcije za verifikacijo predhodno določenih hipotez o obstoju predmeta. Z več zaporednimi manipulativnimi akcijami je mogoče znanje o predmetih kopičiti, na podlagi teh pa določati modele predmeta. V okviru tega dela manipulativne akcije predstavlja potiskanje z robotsko roko.
Tretji proces hipoteze o obstoju predmeta dejansko verificira na podlagi novih informacij, pridobljenih z manipulacijo. Verifikacija temelji na ugotavljanja konsistentnosti detektiranih značilnic predmeta glede na gibanje togega telesa. V primeru, da je kateri izmed procesov neuspešen, lahko sistem oblikuje in ovrednoti dodatne hipoteze, zato predpostavka, da so procesi popolni, ni potrebna. Za zagotovitev robustnosti pristopa smo za več podprocesov uporabili verjetnostne metode, kot je na primer RANSAC (RANdom Sample Consensus).
Pridobljeni rezultati so pokazali, da je predlagan pristop uspešen pri odkrivanju in segmentaciji novih predmetov in da je znanje, ki ga pridobimo z več zaporednimi manipulativnimi akcijami, mogoče zbirati v popolnejše modele predmeta.


Ključne besede: detekcija predmetov, segmentacija predmetov, učenje z raziskovanjem, učenje konceptov, RANSAC.