SYROTEK

On an e-Learning System for Mobile Robotics and Artiﬁcial Intelligence

Miroslav Kulich, Jan Faigl, Karel Ko

snar, Libor P

reu

cil

Department of Cybernetics, Faculty of Electrical Engineering, Czech Technical University in Prague, Czech Republic

Jan Chudoba

Center of Applied Cybernetics, Faculty of Electrical Engineering, Czech Technical University in Prague, Czech Republic

Keywords:

e-Learning, Mobile robotics, Collective robotics.

Abstract:

The paper deals with motivations and design leading to succeeding development of a system for remote learn-

ing of mobile robotics topics. Speciﬁcally, the designed SyRoTek system comprises a team of 12 tele-operated

mobile robots acting in 24/7 maintenance-free environment equipped with charging docks and reconﬁgurable

system of obstacles, all being observable and accessible via Internet. The SyRoTek system together with an

attached e-learning environment it is aimed to provide the features real data gathering and real robot motion

execution. The whole set-up is targeted on training purposes in basic and advanced courses in the ﬁeld of

Intelligent and Mobile Robotics and Collective Robotics as well as for test/veriﬁcation purposes in a research

domain.

1 INTRODUCTION

Robotics and autonomous systems have become an

inseparable part of up-to-date IT solutions. As be-

ing applied more and more frequently in these days,

need of training activities for these demanding tech-

nologies steadily raises. This situation forms condi-

tions and requirements how to approach and optimize

the training issue, respecting the nature of the robotics

systems in question.

Autonomous robotics systems typically require

applications of AI methods achieved through inde-

pendent processing of perceived information, un-

derstanding the environment, and followed by au-

tonomous reasoning and decision-making towards

a self-derived or given goal.

Autonomous (or intelligent) robotics in the ﬁrst

place aims to design and develop systems with capa-

bilities to operate the robot device under uncertainty -

to handle unexpected situation or developments in the

operating space of the system, changes of constrains

of a task to be fulﬁlled, etc.

Moreover, as for a next level autonomy may be

considered a system comprising multiple robot sys-

tems, where each of these is introducing its’ own

intelligence into a coexisting or even a cooperating

team. The later concept leads to collective robotics

tasks, whereas performance of such the system may

substantially be improved through higher ﬂexibility of

the robots team, via better resource planning and as-

signment, more reliable and optimized environment

sensing and other capabilities, which may be per-

formed faster, in more robust way, being best opti-

mized, and even sometimes done resolvable this way

at all.

As it can be seen, majority of the afore mentioned

tasks and their solutions are typically achieved on

conditions, that the robot provides capability of au-

tonomous processing of uncertain information, which

a real environment satisﬁes. The uncertainty itself

may be handled as a random process and there-

fore a simulation can be built up. Unfortunately,

none of feasible random processes allow us to cre-

ate an exhaustive simulator of real environments and

sensors. Although some widely used systems of-

fer a simulated environment and sensory simulations

(e.g. Player/Stage/Gazebo system (Gerkey et al.,

2003)), these do have a limited performance in certain

situations and remain efﬁcient only in early stages of

an intelligent robot design and development phases.

Multiple experiences have shown, that precise

simulation of sensor behaviour is an extremely hard

275

Kulich M., Faigl J., Košnar K., P

reu

cil L. and Chudoba J. (2009).

SYROTEK - On an e-Learning System for Mobile Robotics and Artiﬁcial Intelligence.

In Proceedings of the International Conference on Agents and Artiﬁcial Intelligence, pages 275-280

DOI: 10.5220/0001663102750280

 SciTePress

task, which remains unsolvable very often. There-

fore, existence of real senses, as sensory measure-

ments from real environments can not ever be fully

substituted by simulation.

The preceding ﬁnding leads to the idea, that re-

search/development as well as teaching activities in

the ﬁeld of intelligent robotics can be performed on

high quality level only and only if a real experimental

platform is made available. This comprises mainly

availability of real sensors being operated in real-

world environments and providing realistic data. To

move, orient and position the sensor in the environ-

ment a suitable carrier (mobile platform or actuator)

is needed. Integrating both the previous issues to-

gether and adding some control algorithms we ob-

tain a mobile robot, whereas the crucial control part

(the data processing algorithms, reasoning and plan-

ning) can be executed either on-board such platform

or off-board in an attached or remote computer. Hav-

ing multiple such sensor platforms with some neces-

sary control infrastructure (observation and caretak-

ing system) we achieve a multi-robot system, which

can be made ready either for local or remote con-

trol via Internet. As this setup is truly teleoperated,

it can effectively create an experimental part of an e-

learning system for the intelligent mobile robotics as

well as collective robotics domains.

In this paper we present the SyRoTek system -

a system for distance robotic learning which will al-

low its remote users to get acquainted with algorithms

from areas of modern mobile and collective robotics,

artiﬁcial intelligence, control, and many other related

domains. Advanced users will be able to develop own

algorithms and monitor behaviour of these algorithms

on-line during real experiments.

SyRoTek mobile robots move inside a restricted

area, which contains other elements like obstacles or

objects related to objectives of the actually solved

task. Moreover, several sensors (infrared, sonars,

cameras, etc.) are used to gather information about

the actual status of the play ﬁeld and particular objects

on it. Some sensors are placed on-board the robots,

while others are stand-alone getting global overview

of the play ﬁeld status. The user will be able not only

to observe gathered data using Internet interface, but

also control the robots in real-time. Unlike existing

e-learning robotic systems developed in the world in

which the user can only tele-operate robots, behaviour

of the robots in the SyRoTek system can be modiﬁed,

since the system allows to run own algorithms devel-

oped by the user.

The remainder of the paper is structured as fol-

lows. The next section gives a brief overview of exist-

ing robotic systems for e-learning. In section 3, main

ideas and architecture of the SyRoTek system are pre-

sented. Robots developed and play-ﬁeld are described

in section 4 and 5 respectively. Finally, typical assign-

ments solvable by the system are brieﬂy introduced in

section 6.

2 RELATED WORK

The SyRoTek project is focused on building a system

for distance learning. Many systems for remote robot

control as well as systems for e-learning were imple-

mented during last decades.

First robotic projects enabling their users to share

and control robots via Internet dealt with a single tele-

operated robots. Many of these robots were operating

many years so knowledge gathered during these years

allows creating more advanced e-learning solutions.

One of the ﬁrst robots controlled at distance and

available to public was Telegarden (Telegarden, 2008)

developed at University of Southern California. It has

been running since 1995 with 9000 users registered

to the system in ﬁrst month of operation. Telegarden

has a mechanism which informs its users about actual

state of the system, and planned drop-outs. More-

over, the users can interact with each other via fo-

rum. Number of contributions in the forum shows

that space for exchange experience among users is

an inseparable part of an arbitrary e-learning sys-

tem (see analysis in (Kahn et al., 2005)). Users cre-

ate their own community, manuals, documentation,

tips&tricks which play an important role for collec-

tive solving of problems.

Other system worth to mention is Bradford

Robotic Telescope operated at University of Bradford

(Telescope, 2008). The telescope is a part of an e-

learning course of which goal is to popularize astron-

omy. In addition to open up a unique equipment to

a broad public, the many research programs use tele-

scope for research of galaxies, supernovas, and black

holes. The system thus combines basic research with

education by sharing limited sources.

Project RHINO (Rhino, 2008) combines tele-

operation with visualization. Robot RHINO

(a robotic guide in a museum) is able to operate in

two diverse modes. In the ﬁrst one, the robot guides

visitors which can interact with the robot and inﬂu-

ence the tour. The second mode allows Internet users

to control the robot and to view the museum at dis-

tance. Although main research goal of the project is

to build an autonomous robots with cognitive func-

tions education aspect plays an important role in both

modes.

Robot Xavier (Simmons et al., 2000) developed at

ICAART 2009 - International Conference on Agents and Artificial Intelligence

276

Carnegie Mellon University is an autonomous robot

operating in indoor environments of university hall-

ways. Robot autonomy allows users to enter high-

level tasks (e.g. go to a speciﬁed position), which

are performed by the robot autonomously. After the

task is done, an e-mail with a photo of target place is

sent to the user. Concerning e-learning, an important

part of the project was web interface designed with

respect to deal with limited connection speed. The

authors also discuss aspects related to operation time

of robots. If the system should operate 24 hours/day

and 7 days/week then battery capacity, their charging

and other hardware and software services must be de-

signed with special attention.

Robotoy (Robotoy, 2008) - a robotic arm with

a gripper - developed at University of Wollongong al-

lows it users to control it via web interface. The user

can choose between two cameras from which it can

see robot’s working environment. Although the sys-

tem is relatively simple, it contains all basic compo-

nents of successful distant control. The robot is con-

trolled in command regime, i.e. the user enters a com-

mand which is immediately fulﬁlled.

A notable part of such system is a simulator which

introduces the robot to the user and allows the user

to test robot’s behaviour and its responses to user’s

commands off- line.

One of the most complex robotic e-learning lab-

oratories was developed at Swiss Federal Institute

of Technology in Lausanne (EPFL). The RobOn-

Web project (Siegwart and Sauc, 1999) is focused

on advanced robotic users. The authors deﬁne ﬁve

fundamental services of web interface: chat, video,

robot control, virtual robot representation, and log-

ging. Moreover, an user registration system is intro-

duced, which manages user’s access to robotic hard-

ware. Several conﬁgurations are parts of the project

varying mainly in used robot platform and sensors:

TeleRoboLab, AliceOnWeb, Koala on the Web, and

Pygmalion on the Web. TeleRoboLab is for example

an environment monitored with several cameras, in-

dependent global localization system, Koala robots,

and other controlled devices (movable doors, lights,

etc.). Play-ﬁeld in AliceOnWeb is realized as a small

city with houses, streets, crossroads, and squares.

Robots called Koala sized 22x21x20mm are localized

using a camera placed above the play ﬁeld.

3 SYSTEM DESIGN

The SyRoTek system comprises of ten main com-

ponents (both software and hardware) which are de-

picted on ﬁgure 1, and which will be described in the

following paragraphs.

Majority of e-learning systems is oriented on

a distant user. Objects realizing interface with the

user can be therefore considered as a core of such sys-

tems. Web interface which is user’s front-end gate to

the system is one of these objects. User’s computer

represents user’s work space. The user has access

from its computer to web interface, learning materi-

als, he is able to observe situation on the play ﬁeld,

send commands to control robots. The goal of Envi-

ronment visualization is to visualize actual play-ﬁled

situation to the user. It is realized by a set of video

streams which are produced by cameras monitoring

the play ﬁeld.

Learning materials as well as exercises for prac-

tical veriﬁcation of acquired knowledge are insepara-

ble parts of each modern e-learning systems. With

respect to the process how this material is created and

maintained and to the fact that learning material is in-

dependent from existence of its web presentation, Ex-

ercises and instructions is a standalone object.

environment − arena

system

robots

sensors

localisation database

of users

tasks

tutorials

control

computer

web

interface

visuali−

sation

user’s

computer

services

Figure 1: General concept and main SyRoTek modules.

It is expected that a number of SyRoTek users will

be large and thus a system for user administration.

The aim of Database of users object is to manage in-

formation about SyRoTek users (their login data, sta-

tus of all exercises, actually solved experiment, sen-

sory data and visual streams from experiments, etc.).

Moreover, the object handles a booking system for

user’s access to play ﬁeld, particular robots, sensors

and cameras.

The robots move in a deﬁned space (play ﬁeld)

called Arena. For successful pursuing of robotic

tasks, harmless robot control, navigation of a robot

to the docking station, obstacle avoidance, and for au-

tomatic evaluation of user’s solution of exercises it is

needed to determine positions of robots. Localization

is performed by the Localization object by processing

image from a camera located over the arena.

SYROTEK - On an e-Learning System for Mobile Robotics and Artificial Intelligence

277

Control computer provides user’s access to shared

SyRoTek hardware – it distributes sensor data to the

user and transmits control commands to the hardware.

It is not always possible or required to perform ex-

periments with real robots (during debugging, when

robots are already reserved by other users, etc.). Con-

trol computer therefore allows running the task in the

simulator. From user’s point of view, the simulator

works equally to the real system – the user can send

the same control commands and gets simulated sensor

data and video streams.

The SyRoTek system is designed to run in 24/7

regime. The aim of the Services object is to provide

functionality for distant administration of the system.

The object incorporates connections to other objects,

gathers information from them and creates log ﬁles

about all activities of currently connected user, and

statistics about system load. Moreover, the object is

responsible for backup of the system and its recovery

in case of failures.

4 ROBOTS

It is expected that no more than eight robots will op-

erate concurrently on the play ﬁeld at the same time.

Moreover several (4-8) robots can be prepared in the

docking station which limits robot size to 18cm.

The robots are designed taking modular principles

into account, which allows ﬂexible reconﬁguration of

robots and placement of various sensors on them. The

modularity is taken into consideration in hardware

frame construction, electronic modules, even the soft-

ware design, while it is most signiﬁcant is the sensor

replacement-ability. To typical sensors used belong:

incremental odometry, infrared distance sensors and

sonars. Extended conﬁguration of the robot include

Hokuyo laser range-ﬁnder, accelerometers, compass,

internal/external thermometers, and cameras (CMU-

cam, see (CMUcam, 2008)).

Computational performance is provided mainly

by two computers: on-board computer (Gumstix

Verdex with XScale PXA270 processor, running RT

Linux operating system) and single-chip micropro-

cessor based control computer (processor Hitachi

H8S/2639). On-board computer provides a commu-

nication with the user and other SyRoTek objects and

executes user applications and top-level service func-

tions. Control computer is responsible for controlling

robot drive units, gathering and distribution of chas-

sis sensor data, computing odometry and provide ba-

sic robot movement functions (e.g. velocity control

or trajectory following). The engine current senses

are analyzed by the control computer, providing in-

Figure 2: A robot without a cover and sensors.

formation about higher force against the movement

direction, indicating possible collision with an obsta-

cle. Moreover, the robots are equipped with other pro-

cessors responsible for controlling sensors and mon-

itoring robot status (batteries, temperature,charging,

etc.)

Robots have two wireless communication chan-

nels used for different purposes. Wireless network

card (WiFi) mounted on the on-board computer is

used mainly for the program upload and maintenance.

Because the latency of the WiFi may not be sufﬁcient

for the real-time control of the robot, radio channel

based on the Zigbee communication modules is used

for high-priority data whose latency is crucial, but

their volume is relatively low – control commands and

sensor data (except camera).

An uncovered robot without sensors is depicted on

ﬁgure 2.

5 ARENA

As mentioned above, Arena is the space, where the

robots perform their actions. There are antipodal de-

mands on arena size. The larger the play ﬁeld the

more complex tasks can be solved, robots can move

more freely, more users can solve their assignments,

etc. On the other hand, for localization system based

on camera, it is needed to overview the whole Arena

– the larger the area, the higher the camera should be

placed and the higher resolution it has to have. More-

over, Arena has to be situated in current classrooms,

which also determines its size. As a compromise,

the size was chosen to approx. 350x380cm including

ICAART 2009 - International Conference on Agents and Artificial Intelligence

278

docking station (see ﬁg. 3). It is expected, that obsta-

cles placed on Arena will be of two kinds. While ﬁxed

obstacles are designed to be stationary during longer

periods of time (weeks), movable obstacle can change

their position by moving up and down (the obstacle

can have different heights or can totally coincide with

plane of Arena).

Figure 3: Design of the Arena. Movable (pop-up) obstacles

are highlighted with a red color. Docking boxes are placed

on the top (hatched area).

Robots that currently perform no assignment are

situated in docking boxes. This is an area separated

with barriers from the place where the robots nor-

mally solve their tasks.

6 ASSIGNMENTS

The system allows to solve and test a broad spectrum

of tasks. The important part of the system is therefore

a collection of assignments that users (students) can

solve with it. The assignments are sectioned into sev-

eral categories (courses) according to their difﬁculty,

domain, and dependencies to other assignments.

Students should be introduced to the system ﬁrst,

so they go thru several simple demonstrations and ex-

ercises to become familiar with basic functionality

and features of the system. These exercises encom-

pass loading and running user’s code on the robot,

gathering sensor data, working with logging system,

etc.

Four introductory courses have been arranged in

order to afford students knowledge of fundamental al-

gorithms in key robotic domains. These are focused

on:

• Simple robot control (reactive, behavioural, tele-

operation, dead reckoning, etc.)

• ”classical” robot control (follow the carrot, pure

pursuit, vector pursuit, PID, motion models, etc. )

• Sensor processing and environment modelling

(continuous localization, Monte Carlo localiza-

tion, Kalman ﬁlters. occupancy grids, sensor

models, etc.)

• Path and motion planning (wall following, com-

binatorial planning, sampling-based planning, po-

tential ﬁelds, etc.)

• Communication, coordination, cooperation

Advanced courses are based on complex Top As-

signments (TA) that comprise from several fundamen-

tal problems mentioned above. These courses are

organized so that students learn all necessary funda-

mental algorithms (by solving corresponding assign-

ments) needed to solve TA of the course. TAs can be

divided into two groups: basic and advanced. Basic

Top Assignments are typical problems in robotics and

artiﬁcial intelligence, where solutions are well known

and described. To this class of problems belong for

example:

• Simultaneous Localization and Mapping

• Inspection, exploration, and coverage

• Pick & delivery

Each of these assignments can be solved with

a single robot or with a team of robots.

Advanced Top Assignments are problems whose

optimal or polynomial solution is not known. It is

therefore expected, that the student will either study

the literature to ﬁnd some approximate solution or

it will creatively develop its own approach. Typical

examples of Advanced TAs are e.g. the following

games:

• Treasure hunt

• Pursuit-evasion

• Capture the ﬂag

These problems are typically designed as multi-

robot, where cooperation and coordination of robots

plays an important role (although treasure hunt can

be performed by one robot).

Robot’s goal in Treasure Hunt is to locate a place,

where the treasure is and to navigate to this place. Ac-

cess to this place can be granted under speciﬁc con-

ditions that the robot must fulﬁll (e.g. ﬁnding a key,

opening door, etc.).

SYROTEK - On an e-Learning System for Mobile Robotics and Artificial Intelligence

279

Pursuit-evasion game is family of problems where

multiple robots (pursuers) collectively determine the

location of one or more other robots (evaders) and try

to catch them.

Capture the ﬂag generalizes the previous problem.

The aim of each of two groups playing the game is

to ﬁnd and capture the ﬂag defended by the opponent

group. In other words, each group pursuits and evades

simultaneously.

Each of the aforementioned assignments can be

considered in different levels of difﬁculty determined

by the environment (e.g. static x dynamic, orthog-

onal, grid-like or general), whether the environment

is known, and abilities and equipment of opponents.

These speciﬁcations lead in most cases to totally dif-

ferent solutions which increase variability of possible

assignments.

SyRoTek system is designated to work also in

open mode for trusted users. This mode grants full

access to all features of the system and it is intended

especially for researches and phd students for devel-

opment and veriﬁcations of their algorithms.

7 CONCLUSIONS

While this paper is written the research is still in

progress. Due to this fact, the paper presents only ﬁrst

results concerning system architecture, main ideas

and ﬁrst results.

An exhausting study of current state of the art in

robotic e-learning was originated during previous pe-

riod. The study deals with teleoperation, software

technologies and frameworks, Internet and web in-

terfaces, hardware components, sensors, and mobile

robots that can be potentially used in the project.

Based on the study, overall design of the system was

sketched together with main its components and their

functionality. Moreover, SyRoTek robots were de-

signed and their functional prototype was built.

Activities for the next time concern mainly to ﬁnal

design and production of 12 mobile robots, design and

implementation of fundamental software functional-

ity of robots an the control computer. Furthermore,

the assignments will be speciﬁed in more details, as

well as a concept of user’s access to the system will

be designed.

ACKNOWLEDGEMENTS

The presented work has been supported by the Min-

istry of Education of the Czech Republic under pro-

gram ”National research program II” by the project

2C06005. The support of the Ministry of Education

of the Czech Republic, under the Project No. 1M0567

to Jan Chudoba is also gratefully acknowledged.

REFERENCES

CMUcam (2008). http://www.cs.cmu.edu/∼cmucam.

Gerkey, B. P., Vaughan, R. T., and Howard, A. (2003).

The player/stage project: Tools for multi-robot and

distributed sensor systems. In In Proceedings of the

11th International Conference on Advanced Robotics,

pages 317–323.

Kahn, P. H., Friedman, B., Alexander, I. S., Freier, N. G.,

and Collett, S. L. (2005). The distant gardener: What

conversations in the telegarden reveal about human-

telerobotic interaction. In Proceedings of the 14th

International Workshop on Robot and Human Inter-

active Communication (RO-MAN ’05). Piscataway,

NJ,IEEE.

Rhino (2008). http://www.cs.uni-bonn.de/∼rhino/tour-

guide.

Robotoy (2008). http://robotoy.elec.uow.edu.au.

Saucy, P. and Mondada, F. (1998). Khepontheweb : One

year of access to a mobile robot on the internet. In

Proceedings of the 1998 IEEE/RSJ International Con-

ference on Intelligent Robots and Systems IROS 1998.

Piscataway, NJ, IEEE.

Siegwart, R. and Sauc, P. (May 1999). Interacting mo-

bile robots on the web. In Proceedings of the 1999

IEEE International Conference on Robotics and Au-

tomation.

Simmons, R., Fernandez, J. L., Goodwin, R., Koenig, S.,

and O’Sullivan, J. (2000). Lessons learned from

xavier. Robotics and Automation Magazine, pages 733

– 39.

Telegarden (2008). http://www.telegarden.org.

Telescope (2008). http://www.telescope.org.

ICAART 2009 - International Conference on Agents and Artificial Intelligence

280