Operation environment of mobile robots with supervision control.
Andreev, Victor Pavlovich ; Pryanichnikov, Valentin E.
Abstract: Based on analysis of data flows and functional structure
of information--measuring and control systems of mobile robots with
supervisory control, proposed a combined approach to constructing such
systems and multi-camera computer vision systems.
Key words: mobile robots, operation environment, computer vision
systems, supervisory control, network with mobile nodes
1. MOBILE ROBOT CONTROL SYSTEM
Let us consider data flows in an information-measuring system of
the mobile robot (MR) on the example of the operation environment shown
in Fig. 1. As any complex control system, a MR involves subsystems
estimating the state of the external environment and the state of the
control system as well as the formation of feedback and control signals.
The information-measuring system involves different sensors (data
sensing devices): odometers, ultrasonic sensors, TV cameras, thermal
imagers, special scanners (gamma locators), etc.
In systems with supervisory control, the data analysis and control
goal formation are conducted by human beings. Therefore, the information
about the environment should be delivered in a form that is customary
(i.e., as a halftone color image) and one should use computer vision
systems (CVSs) as the main sensor.
The signals arriving from different sensors are subjected to the
influence of various noises and distortions (fluctuation noises, optical
distortions, distortions caused by sensors, etc.). Thus, the system
should contain a signal preprocessing unit responsible for filtering
these noises and correcting the distortions. A distinguishing feature of
the MR is that the incoming information has a dynamical character;
consequently, the signal preprocessing unit must be real-time. Because
such a huge data flow cannot be handled by an ordinary computer, the
signal processing at this stage is conducted with the help of
special-purpose processors.
Among signal preprocessing algorithms, the most efficient are the
adaptive algorithms using the result of analyzed information about the
permanently changing external environment to adjust its parameters
(Andreev, 2010). The parameters of signal preprocessing can be affected
also by the goal of control. The instructions for changing the
preprocessing parameters arrive from an internal interface through an
instruction control unit, thus making an internal control loop.
The video signals from TV-cameras and the signals from other
sensors must be converted into a structured "data train" that
should be prepared for transmitting through communication channels
(communication environment). By a structured data train, we mean a
continuous data flow from all sensors of the MR and control instructions
that are ranged into a sequence in line with definite rules. The data
transmission through internal chains of a MR can be performed using
wired broadband digital channels of data transmission, when digital
images and other data refined from different kinds of noises go directly
into the multiport memory of the data-analysis unit. When a structured
data train is transmitted through radio channels, the remote interface
encounters a number of problems stemming from the communication
environment. When analog radio channels are used, a problem of bandpass
arises, which in this case should be enough for transmitting TV video
signals from a set of television cameras. In addition, analog radio
channels are known to have low fault-tolerance.
The transition to digital data transmission through radio channels
made it possible to solve theses problems. However, an absolutely stable
radio channel will never be organized; therefore, the internal interface
is assigned to treat a finite set of navigation problems in an
autonomous mode for the case when radio control is lost. Depending on
the complexity of given problems, the data analysis and goal generation
can be performed using both remote and internal interfaces.
The remote interface is responsible for converting the incoming
data into a standard form, analyzing these data, generating a goal of
control and corresponding instructions or a program of control, and
transmitting them into the MR for execution, thus forming an external
contour of control. According as the control type (remote or
supervisory), the incoming data are analyzed and the control goal is
generated by a human control or automation together with human. In both
cases, the incoming information should be reflected on control desk
monitors in a form that is most convenient for human perception. All
arriving video streams must be converted into images on the screens of
monitors. The latter must also involve images reflecting the indications
from other MR-based sensors. The goal of control is made by a human
control. Therefore, the control desk as a part of remote interface
should contain control units meeting ergonomic requirements: buttons,
tumbler switches, joysticks, etc. The instructions from control units
are transmitted to the MR. All functions mentioned above are implemented
by a control desk computer. The performance of this computer depends
primarily on the number of video streams that need to be decoded and
converted into a sequence of half-tone images generated with a standard
frame rate (25/30 frames per second). In the case of supervisory
control, this computer is assigned to perform data analysis and
generation of control goal within the framework of human instructions.
[FIGURE 1 OMITTED]
The internal interface is responsible for converting incoming data
into a standard form, analyzing data, generating a control goal (in the
cases of supervisory control and loss of radio communication), and
generating suitable instructions or program of control and their
transfer to the MR-control instruction generating unit for execution,
thus forming an internal contour of control. These functions require a
computer but with modest computing resources because this requires
considerable energy costs under limited energy possibilities of on-board
power supply of the MR.
The unit of generation of control instructions is responsible for
the transformation of decisions made from the operation of internal
interface into a system of instructions determined by specific features
of the unit of generation of control signals. This may include, for
example, a USB to RS-232 or RS-485 converter. The unit of generation of
control instructions is actually an interface transformer allowing one
to use microprocessor units of different manufacturers.
The unit of generation of control signals is a transformer of
control instructions into electrical signals that are sent to executive
mechanisms through respective amplifiers. An example of the system of
generators of control instructions and signals is a PTZ camera
controller equipped with a rotary tilting mechanism and an optical
scaling system.
2. FUNCTIONAL STRUCTURE OF MR
An analysis of data flows in the MR makes it possible to specify
the required functional structure of its equipment:
--a set of sensors with a corresponding system of signal generation
(TV camera, thermal imager, gamma-radiation scanner, odometers,
ultrasonic sensor, etc.);
--a system of data collection and filtering of signals from
different sensors for the correction of distortions caused both by the
fact that the gain-transfer characteristics of signal sensors are
nonideal and by signal-generation;
--a system of radio channel for communication between the on- board
unit and control desk;
--a repeater designed generally for acquiring a stable radio
channel at large distances;
--a controller desk with a radio channel generation unit, a system
for converting video signals into half-tone images (including a video
data mapping control system), and a system of control bodies (for
example, joysticks);
--a system of self-contained power supply for electronic units and
drivers of executive mechanisms;
--for supervisory control, a computer is installed onboard
MR;--remote units used for making the operation of the MR control
easier. These units contain controllable TV-cameras placed immediately
close to the place of MR operation and allowing to observe it as an
onlooker.
3. SPECIFIC FEATURES OF CVS FOR MR
Let us consider some key aspects of CVS construction for. Here, we
take into account that the completeness and reliability of data mapping
are necessary for minimizing the errors arising in the generation of
control goals.
The completeness of data is ensured by the possibility of a full
coverage of the locale, and the reliability of data depends on both the
technical characteristics of sensors and techniques of data retrieval
and representation. The technical characteristics of sensors influence
on data completeness as well. It's very important for
self-organizing systems (Katalinic, et al., 2002).
The full coverage of the locale on the MR is provided by the system
of optical radiation sensors: TV-camera with an optical system of
omnidirectional scanning; TV-camera with a rotary tilting mechanism;
several TV- cameras installed onboard the MR; TV on remote units.
For MR with supervisory control, it will be most appropriate to
combine the variant with a controlled multicamera CVS using TV-cameras
with a rotary firing mechanism and the possibility of optical scaling
(PTZ). The remote units (RUs) must be self-contained and consequently
have a suitable power-supply system and a unit of radio channel
generation for communicating with the controller desk. These RUs
("satellites") can be placed on a simplified moving chassis
(on the robot-observer).
The use of multi-camera CVSs brings up the question of multi-stream
video transfer through a radio channel. One of the traditional answers
is to use a broadband analog radio channel. The use of IP-cameras with
hardware compression of images makes it possible to solve the problem of
multi-stream video more cheaply and reliably. In this case,
analog-digital conversion occurs making it possible to use the most
advanced Ethernet-based methods. Then, the problem is reduced to the
creation of a local computing network with mobile nodes represented by
TV-cameras installed on onboard units and remote units as well as, on
the one hand, an onboard computer and, on the other hand, a computer on
the controller desk (Andreev et al., 2010). This CVS has the following
properties. 1. Multi-stream video: video signals can be transferred
simultaneously from several TV-cameras without reducing the quality of
images, which is achieved by using highly efficient compression
algorithms for the sequence of frames. 2. Uniting CVS elements through
digital radio channels: the Wi-Fi or Wi-Max standards are used. 3.
Scalability: the possibility of easy upgrading of any MR resources
(IP-remote units). 4. High- quality images: the possibility of using
high-resolution TV- cameras with good color rendering. 5. Digital
systems of image processing: a special digital processors and versatile
computers. 6. Enhanced noise protection: digital channels of data
transmission prevent video signals from distortion. 7. Functionality:
digital channels make it possible to transfer not only video but also
audio signals and control signals. 8. Distributivity: computing
resources can be distributed between MR units. 9. Multiuser mode: the
possibility retrieval of data on several controller desks (including
those connected to the system via the Internet). 10. Control via the
Internet: the possibility of analyzing the operation of MR and
controlling over its executive mechanisms at any distances.
4. CONCLUSION
Within the frameworks of the concept and under contracts signed
with different agencies and universities, our laboratory has developed
AMUR-series mobile robots and CVSs installed on heavy-duty mobile robots
(Brokk-110D, Brokk-330). These systems proved their efficiency,
including for conditions of operating in emergency situations MR
(Andreev et al., 2009).
5. REFERENCES
Andreev V., Kirsanov K., Pronkin V., Pryanichnikov V., Prysev E.
(2009). Computer vision for mobile robot on the base of Wi-Fi. //
Intelligent and Adaptive Robots: Radiotechnika, ISSN 2070-0814,No 1,2,
pp.49-63, Moscow
Andreev V., Pryanichnikov V., Prysev E. (2010). Multi-access
control of distributed mobile robotic systems on the base of networking
technologies. Annals of DAAAM for 2010&Proceedings of the 21st
International DAAAM Symposium, ISSN 1726-9679, pp. 15-16, Zadar, Croatia
Andreev V. (2010). Video Signal Correction for Scanning Photocell Array in the IR-Computer Vision Systems. Annals of DAAAM for
2010&Proceedings of the 21st International DAAAM Symposium, ISSN
1726-9679, pp. 17-18, Zadar, Croatia
Katalinic, B., Kurdic, V. (2002) Concept of Design and Scheduling
of Self-Organizing Complex Flexible Assembly System, Proceedings of the
4th International Workshop on Emergent Synthesis--IWES 02, ISBN 961-4238-49-3, pp. 89-96, March 12-13, 2002, Kobe, Japan
