04-02-2013, 10:45 AM
INTELLECTUAL CAMERA UNIT
[attachment=49482]
Abstract
Technological advances will allow the integration of smart devices which may
comprise sensor components, computational devices and a network interface. The built-in
computational component enables the implementation of a well-defined high level interface
that does not just provide raw transducer data, but pre-processed, application-related set of
process variables. The paper deals with two issues. Firstly, it describes an architecture which
encourages the design of multi-level control hierarchies exploiting the easy availability of
application related sensor information. This is based on encapsulated smart components, a
well defined communication interface and the easy access to this information by a variant of a
shared data space. Secondly, we describe the functions and the hardware of ICU which
constitutes the prototype of an optical sensor designed for vehicle guidance operating as a
smart component in such a system. The task of the sensor is to detect a guidance line and
directly produce the information needed by the steering system to control the vehicle.
Introduction
Technological advances will allow the
integration of smart devices which may
comprise sensor components computational
devices and a network interface. The builtin
computational component enables the
implementation of a well-defined high level
interface that does not just provide raw
transducer data, but a pre-processed,
application related set of process variables.
Consequently, the interfaces and the
functions of these smart components may
include functions related to overall control,
supervision, and maintenance issues. In
his excellent survey of vision sensors,
AlirezaMoini makes the point that “smart
sensors are information sensors, not
transducers and signal processing
elements”. Perhaps the most interesting and
challenging property of these intelligent
devices is their ability to spontaneously
interact with the overall system. This
enables a modular system architecture in
which smart autonomous components cooperate
to control physical processes
reactively without the need of a central coordination
facility.
Co-Operating Optical Sensors
There is a large variety of optical sensors
and vision sensors ranging from very simple
CCD and CMOS arrays to intelligent vision
chips and artificial retinas. They can
detect relatively simple things as
distributions of light intensities or detection of
marks or lines. More sophisticated sensors
enable motion detection and deriving the
speed of a device from optical flow analysis
directly embedded in the sensor array.
These devices are usually specialized to a
single function which they can perform faster,
better and with less energy as compared to
traditional vision systems which usually
convert a digital image in a an analogue
format coming from the first days of
television and then use a frame grabber
easily wasting 30 Mbyte of PCI bandwidth
again to create a digital image in the
processor memory.
Requirements For The System Control
Architecture
Traditional control systems usually center
around a single control unit (CU) which has a
sophisticated real world interface. The CU is
the last link in a chain of transducers,
converting a physical process variable, like a
temperature, a pressure or an optical signal
via a electrical signal to a digital
representation. Signal conditioning as
shaping analogue values, debouncing digital
inputs and improving an image received from
a camera all are performed by this CU. The
resulting digital information is a raw,
conditioned representation of a single
physical entity. Subsequent processing,
fusion of multiple sensors information, and
generation of control signals for the
actuators also is performed by the same
processor.
Functions Of The Icu
ICU was primarily designed for vehicle
guidance in an experimental traffic
scenario. In our first application ICU tracks a
guidance line (G-line) on the floor which may
include embedded colored marks. ICU
disseminates the position and orientation of
the G-line and the colour of the embedded
marks. The information delivered by ICU is
then directly used by the motor drive system
to reactively keep the vehicle to follow this
line, to slow down or to stop if a certain
colored mark is detected. Fig. 2 shows
the principles of detecting a G-line (the white
bar in the figure).
The Hardware Of Icu
ICU was intended as a low cost sensor
for simple imaging tasks which can directly
by used for control applications. ICU
should be adaptable to various simple
vision tasks. Therefore, we took a
camera/processor approach and put
emphasis on an easy and fairly universal
processor interface. Fig. 6 depicts the basic
components of ICU. We connect the
sensor to an asynchronous external port,
which is not the most efficient way but eases
the use of different micro-controllers. In fact,
we used the front end sensor and the
interface logic even with a simple microcontroller
as Motorola 68HC11 for
educational purposes. In the current version,
a 16-Bit controller C167 from Siemens [19]
(now provided by Infineon) is used featuring
a minimal instruction execution time of 100ns
and a peak rate of 10Mips.
Conlusion And Future Work
Design of control systems for complex
artifacts like autonomous vehicles, driver
assisting cars, airplanes and industrial plants
tend to become enormously complicated.
One step towards solving the problem is to
provide computational power at the
sensors and actuators making them smart
devices encapsulating all computations
needed to convert a physical process value
to an application related meaningful
information. The autonomy and
encapsulation properties encourage
hardware/software co-design on the
device level and component-based design
methodologies [20] for putting the building
blocks together.
[attachment=49482]
Abstract
Technological advances will allow the integration of smart devices which may
comprise sensor components, computational devices and a network interface. The built-in
computational component enables the implementation of a well-defined high level interface
that does not just provide raw transducer data, but pre-processed, application-related set of
process variables. The paper deals with two issues. Firstly, it describes an architecture which
encourages the design of multi-level control hierarchies exploiting the easy availability of
application related sensor information. This is based on encapsulated smart components, a
well defined communication interface and the easy access to this information by a variant of a
shared data space. Secondly, we describe the functions and the hardware of ICU which
constitutes the prototype of an optical sensor designed for vehicle guidance operating as a
smart component in such a system. The task of the sensor is to detect a guidance line and
directly produce the information needed by the steering system to control the vehicle.
Introduction
Technological advances will allow the
integration of smart devices which may
comprise sensor components computational
devices and a network interface. The builtin
computational component enables the
implementation of a well-defined high level
interface that does not just provide raw
transducer data, but a pre-processed,
application related set of process variables.
Consequently, the interfaces and the
functions of these smart components may
include functions related to overall control,
supervision, and maintenance issues. In
his excellent survey of vision sensors,
AlirezaMoini makes the point that “smart
sensors are information sensors, not
transducers and signal processing
elements”. Perhaps the most interesting and
challenging property of these intelligent
devices is their ability to spontaneously
interact with the overall system. This
enables a modular system architecture in
which smart autonomous components cooperate
to control physical processes
reactively without the need of a central coordination
facility.
Co-Operating Optical Sensors
There is a large variety of optical sensors
and vision sensors ranging from very simple
CCD and CMOS arrays to intelligent vision
chips and artificial retinas. They can
detect relatively simple things as
distributions of light intensities or detection of
marks or lines. More sophisticated sensors
enable motion detection and deriving the
speed of a device from optical flow analysis
directly embedded in the sensor array.
These devices are usually specialized to a
single function which they can perform faster,
better and with less energy as compared to
traditional vision systems which usually
convert a digital image in a an analogue
format coming from the first days of
television and then use a frame grabber
easily wasting 30 Mbyte of PCI bandwidth
again to create a digital image in the
processor memory.
Requirements For The System Control
Architecture
Traditional control systems usually center
around a single control unit (CU) which has a
sophisticated real world interface. The CU is
the last link in a chain of transducers,
converting a physical process variable, like a
temperature, a pressure or an optical signal
via a electrical signal to a digital
representation. Signal conditioning as
shaping analogue values, debouncing digital
inputs and improving an image received from
a camera all are performed by this CU. The
resulting digital information is a raw,
conditioned representation of a single
physical entity. Subsequent processing,
fusion of multiple sensors information, and
generation of control signals for the
actuators also is performed by the same
processor.
Functions Of The Icu
ICU was primarily designed for vehicle
guidance in an experimental traffic
scenario. In our first application ICU tracks a
guidance line (G-line) on the floor which may
include embedded colored marks. ICU
disseminates the position and orientation of
the G-line and the colour of the embedded
marks. The information delivered by ICU is
then directly used by the motor drive system
to reactively keep the vehicle to follow this
line, to slow down or to stop if a certain
colored mark is detected. Fig. 2 shows
the principles of detecting a G-line (the white
bar in the figure).
The Hardware Of Icu
ICU was intended as a low cost sensor
for simple imaging tasks which can directly
by used for control applications. ICU
should be adaptable to various simple
vision tasks. Therefore, we took a
camera/processor approach and put
emphasis on an easy and fairly universal
processor interface. Fig. 6 depicts the basic
components of ICU. We connect the
sensor to an asynchronous external port,
which is not the most efficient way but eases
the use of different micro-controllers. In fact,
we used the front end sensor and the
interface logic even with a simple microcontroller
as Motorola 68HC11 for
educational purposes. In the current version,
a 16-Bit controller C167 from Siemens [19]
(now provided by Infineon) is used featuring
a minimal instruction execution time of 100ns
and a peak rate of 10Mips.
Conlusion And Future Work
Design of control systems for complex
artifacts like autonomous vehicles, driver
assisting cars, airplanes and industrial plants
tend to become enormously complicated.
One step towards solving the problem is to
provide computational power at the
sensors and actuators making them smart
devices encapsulating all computations
needed to convert a physical process value
to an application related meaningful
information. The autonomy and
encapsulation properties encourage
hardware/software co-design on the
device level and component-based design
methodologies [20] for putting the building
blocks together.