IEEE Robotics & Automation Magazine - June 2013 - 53

sensor technology. Indeed, several kinds of kinematic transducers are included in the design, the utility of which will be
discussed in the sequel. Moreover, we gave up any aesthetic
concern, at least, until now. The size of the robot was chosen to
conduct experiments in a limited space with, say, ten units,
but, at the same time, sufficiently large to require no complex
(and costly) miniaturization and to allow few limited
expansions.
Related Works
A wide range of mobile robots are available on the market
both for educational and research purposes. In the field of
educational robotics, small platforms are sold by several companies. Although these units are characterized as inexpensive,
they provide a wide set of tools that allow the students to get
experience with hardware. The Scribbler from Parallax is one
example: the dimension of the robot is 19 # 16 cm, and it
mounts dc motors that are actuated by a microcontroller. It
furnishes a cheap solution (below US$200), providing a minimal set of transducers that can be easily accessed through a
serial interface. Interestingly, the provided microcontroller is
capable to run concurrent programs. Another example is
given by the Pololu 3pi: this platform is provided with a
microcontroller that steers the platform up to 0.9 m/s. It also
provides a few digital and analog lines to exploit external
expansions. The LEGO NXT 2.0 is a more powerful robot: it
provides an ARM7 processor that is supported by a second
microcontroller. It can be interfaced with several transducers,
in particular, the sonar one is able to get distances up to
2.3 m. Another peculiarity is the possibility to control this
platform by Labview and MATLAB. At the same time, other
companies strictly related to universities introduced some
platforms. The Costbot, developed at University of Southern
California, Berkeley, is a miniplatform built on the sensor
network mote MICA from Crossbow [8]. This platform is
13 # 6.5 cm, and it basically provides motion capabilities to
the sensor network node.
The e-puck, developed at Ecole Polytechnique Federale
De Lausanne (EPFL), is controlled by a digital signal processor (DSP) that allows for direct control of more sophisticated
sensors [9]. It also provides expansions that extract spatial
information related to other agents. This robot is particularly
suitable for educational purposes: through DSP capabilities,
the students can also get experience with sophisticated realtime signal processing.
Although in all the previous examples the platforms
show limited sensorial and computational capabilities to
gather information about the surrounding environment,
other robots exploit more sophisticated and powerful
architectures and are therefore typically used to conduct
research experiments.
The Khepera III from K-Team is a popular tool in the
research community. This platform is small and 12 cm in
diameter but provides a decent computational power. The
processor runs at 400 MHz and is equipped with a Linux
operating system (OS). This robot can be equipped with

several expansions such as laser scanners or cameras. The
traction system is controlled by a DSP that provides an interface to the actuators and basic sensors [infrared (IR), light,
and sonar devices].
The Garcia robot from Acroname, similar to Khepera III,
is a small robot that can be equipped with laser scanners or
other sensors with a maximum payload of 2 kg. It also provides an arm that can be used to mount devices on top of it.
Another popular platform is the iRobot-Create from iRobot.
Although this robot only provides a traction system and some
proximity transducers, its reduced cost (under US$200)
makes this platform a good base for more performing robots.
In fact, it can carry a payload of about 2 kg, and, consequently,
it can be equipped even with a small laptop. Upon this principle, recently, the Ros Turtle-Bot from Willow-Garage and
Billibot (www.billibot.com) have been built. The former is
constituted by an iRobot Create equipped with a netbook and
Kinect sensor. The latter is provided with the Kinect as well,
but it has a more powerful processing unit and accommodates a small gripper. Both of them mount the robot OS
(ROS) and can be run directly onboard the robot.
All previous platforms show several compromises between
performances and costs. By size and computational resources,
SAETTA can be compared with Khepera III: they basically
have the same amount of computational power and comparable dimensions. The peculiarity of SAETTA is related to the
ability to mount different low-level sensors whose management is directly integrated in the platform without the need of
installing additional boards. At the same time, this robot cannot be compared with the higher-performing Turtle-Bot and
Billibot. These platforms are equipped with a more advanced
processor and greater power resource, which allow for exploiting more performing devices such as the Kinect sensor.
Currently in the prototyping phase, the components of
SAETTA cost about US$600; particular care has been focused
on the mechanics and low-level electronics; in fact, only 30% of
the costs is related to the high-level tier. SAETTA is going to be
produced by Spin Italia Srl Company (http://spinitalia.com/).

Figure 1. SAETTA robots: a part of the 12 units developed at DIA
Robotics Lab. (Photo courtesy of DIA, Rome.)

june 2013

IEEE ROBOTICS & AUTOMATION MAGAZINE

http://www.billibot.com http://www.spinitalia.com/

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - June 2013