State of the Art: Difference between revisions
No edit summary |
No edit summary |
||
(4 intermediate revisions by the same user not shown) | |||
Line 1: | Line 1: | ||
'''Window Cleaning Robots''' | '''Window Cleaning Robots'''<br/><br/> | ||
As mentioned in the introduction, window cleaning robots are currently on the market in the form | As mentioned in the introduction, window cleaning robots are currently on the market in the form | ||
of domestic and professional robots. The most obvious difference between the two types of robots | of domestic and professional robots. The most obvious difference between the two types of robots | ||
Line 54: | Line 54: | ||
in the upcoming chapters. Because the focus is put on the improvement of the motion algorithms, | in the upcoming chapters. Because the focus is put on the improvement of the motion algorithms, | ||
the current technological state of the abilities on the terrain of motion planning are analyzed in | the current technological state of the abilities on the terrain of motion planning are analyzed in | ||
the next section<br/><br/> | the next section.<br/><br/> | ||
'''Motion Planning Algorithms | '''Motion Planning Algorithms''' | ||
<br/><br/> | <br/><br/> | ||
The literature study shows that there is already a wide variety of motion planning approaches | |||
possible, which differ between the types of window cleaning robots. As mentioned before, these | |||
approaches are fairly simple and inefficient. The most common approaches to the movement of | |||
the robots will be discussed in this section.<br/> | |||
For professionally used cleaners the movements are mostly straightforward. They move in parallel | |||
lanes over the building’s façade from the top to the bottom of the building, as seen with the | |||
SkyScraper-I model.<br/><br/> | |||
More advanced methods are also developed in the form of multi-robot systems or sensory adjusted | |||
movement. These multi-robot systems can be divided into so-called ’parent-robots’ and | |||
’child-robots’. One parent-robot can transport n-number of child-robots to the designated site. | |||
Each robot can move only vertically or horizontally. The task of cleaning is performed by the | |||
child-robots that have modular facade maintenance tools. A variation of this also exist with a | |||
different system, which consist of horizontal and vertical moving robots. The horizontal robot carries | |||
out window cleaning work while moving along a horizontal (transom) rail. The vertical robot | |||
transports the horizontal robot to another level along a vertical rail. After cleaning a floor, the | |||
horizontal robot docks into the vertical robot where it is securely held in place with rail breaks. | |||
Then the vertical robot is transported up or down by a wire winch, making it able for the horizontal | |||
robot to clean another floor.<br/> | |||
For the sensory robots, a number of built-in guide-type robots must move along predefined rails. | |||
They can measure the contamination level of the building’s surface using detection sensors and | |||
can adjust their movement speeds accordingly. Thus, moving slower at places with high contamination | |||
level creating more focused cleaning of the façade.<br/><br/> | |||
The domestically used robots move in simple ways as well. The very basic domestic robots | |||
move in arbitrary directions until faced with an obstacle. When this happens the robot turns and | |||
faces an new direction to continue cleaning. This arbitrary moving is obviously not efficient that’s | |||
why the most common window cleaners do have some form of logical motion planning.<br/><br/> | |||
Most cleaners start at the top of the window, then proceed to move down the windows surface | |||
in horizontal lanes. Cleaning with water causes water drippage down the window due to gravity. | |||
Starting at the top of the window makes sure the robots are also adjusting for this the dripping of | |||
water down the window by cleaning their own spillage.<br/><br/> | |||
More advanced robots try to estimate the dimensions of the window by making ’smart movements’, | |||
which are then followed up by repeated movements to clean the complete windows surface. | |||
These smart movements are based on different methods. For example, these can be done | |||
using vertical positions estimation units, which are similar to the system used in old computer | |||
mouse devices. Where the robot uses a ball to track its movement relative to its starting point and | |||
this way knows the dimensions and its position on the windows surface.<br/> | |||
On top of most systems, a control loop is used to keep track of displacements. This method prevents | |||
cleaning areas twice or missing places, leaving areas uncleaned which is undesired.<br/><br/> | |||
It is also possible for these robots to have remote control features. This means the user can | |||
decide over the motion of the robot according to what’s desired. If the user feels the window | |||
cleaner missed a spot or should re-clean a certain area, the user can remotely steer the robot to | |||
the corresponding place on the window.<br/><br/> | |||
The reason this standard movement in horizontal lanes is flawed is a consequence of the robot | |||
only knowing its movement, and not having any knowledge about the state of the windows surface. | |||
This makes it so the robot does not know if the window is actually clean, after completing | |||
its cleaning pathway. As mentioned before the user can often manually correct mistakes or insufficiently | |||
washed areas, but it is expected by the users that such a window clean robot can | |||
fully clean a window autonomously. This means the robot needs a way to get information about | |||
the current state of the windows surface and, on top of that, a corresponding course of action to | |||
correctly adjust its cleaning according to this knowledge of the windows dirtiness. |
Latest revision as of 16:12, 2 April 2018
Window Cleaning Robots
As mentioned in the introduction, window cleaning robots are currently on the market in the form
of domestic and professional robots. The most obvious difference between the two types of robots
is the size. Window cleaning robots which are built for domestic use are much smaller than professionally
used skyscraper robots. WINDORO is the best-known example of a domestic
window cleaning robot, with dimensions of 200 x 200 x 50 mm. With these dimensions it is small
enough to fit on almost every window in an ordinary house. An example of a professional robot is
IPC Eagle’s HighRise505 which is 2908 x 930 x 2115 mm. A consequence of the size difference
can also be seen in the performances of both types. The HighRise is able to clean surface
area at a higher rate than the WINDORO. The main difference in behavior of domestic and professional
robots is that domestic robots are only supposed to clean one window at a time and
require to be manually transferred to a different window while most industrial robots are capable
of cleaning an entire facade without human interaction.
There are, however, also guiding vehicles which are specially designed to use domestic window
cleaning robots like the WINDORO. These tethered guiding robots, known as TGV’s,
help to reduce the risk of unwanted events from happening and are specifically used for high
rise buildings. The TGV’s help guide the window cleaning robots from window to window and
determine the orientation of the robot itself. These tethered guiding systems are often sought for
when the window cleaning robot has to move across window surfaces. Another example of a tethered
robot is the SkyScraper-I. This robot tackles, unlike the TGV, both issues of movement
across windows and cleaning the windows itself. For the first issue, reels are installed on the top
of a building to which tethers are attached, which are then in turn attached to the SkyScraper-I.
This way the robot can access every window on a specific side of the building. For the cleaning
of the windows, the SkyScaper-I uses long vertical oriented rods along with a squeegee can move
up and down. At each end of these rods, a rotating arm is attached with a roller. These arms
press the rollers to the window frame without making contact with the window. The squeegees
move up and down to clean the window.
Getting back to domestic window cleaning robots, the WINDORO the best known domestic window
cleaning robot, as mentioned before. The WINDORO adheres to the glass by means of two
magnet units on each side on the glass. The inner unit is responsible for navigation and works via
two silicon driving wheels which have a high coefficient of friction. The outer unit is responsible
for cleaning and works via four motorized rotating disks with a pad for cleaning. The outer unit
also consists of a water tank filled with liquid detergent. A nozzle sprays the volume of this tank
on the glass surface by using a small water pump.
The WINDORO also has a successor, the Smart WINDORO. This improved version has
renewed magnetic adhesion, magnetic force control, vertical position control and corner cleaning
mechanisms. The renewed magnetic adhesion reduces the energy needed in order to stick to the
window. The magnetic force control helps with windows of different thickness. Furthermore, the
vertical position control helps reducing time and energy by reducing double cleaned surfaces and
the new corner cleaning mechanism, which has rectangular disk instead of a circular disk as seen
in the old version, cleans window corners more properly.
There are more differences between window cleaning robots in general that can be considered,
the water supply can be implemented through a water tank or hose, the movement is possible through
tracks, external ropes or rails, energy through a battery or socket cable, cleaning through high-pressure water beams, brushing or movement of a wipe. Some robot are equipped with sensors
to measure dirt levels on the surfaces those robots pass and other sensors to better understand
the environment the robot is working in.
These are some of the possible characteristics of current window cleaning robots regarding mechanical
features. Besides all these differences in geometry, the robots also differ in software.
The way the robot moves over the window is called the motion planning of the robot. Motion
planning is dependent on the mechanical possibilities of the robot in its environment, as well as
the complexity of the algorithms it uses.
For this project, the focus is on domestic window cleaning robots, more specifically the motion
planning of these domestic robots. The set specification regarding the hardware will be explained
in the upcoming chapters. Because the focus is put on the improvement of the motion algorithms,
the current technological state of the abilities on the terrain of motion planning are analyzed in
the next section.
Motion Planning Algorithms
The literature study shows that there is already a wide variety of motion planning approaches
possible, which differ between the types of window cleaning robots. As mentioned before, these
approaches are fairly simple and inefficient. The most common approaches to the movement of
the robots will be discussed in this section.
For professionally used cleaners the movements are mostly straightforward. They move in parallel
lanes over the building’s façade from the top to the bottom of the building, as seen with the
SkyScraper-I model.
More advanced methods are also developed in the form of multi-robot systems or sensory adjusted
movement. These multi-robot systems can be divided into so-called ’parent-robots’ and
’child-robots’. One parent-robot can transport n-number of child-robots to the designated site.
Each robot can move only vertically or horizontally. The task of cleaning is performed by the
child-robots that have modular facade maintenance tools. A variation of this also exist with a
different system, which consist of horizontal and vertical moving robots. The horizontal robot carries
out window cleaning work while moving along a horizontal (transom) rail. The vertical robot
transports the horizontal robot to another level along a vertical rail. After cleaning a floor, the
horizontal robot docks into the vertical robot where it is securely held in place with rail breaks.
Then the vertical robot is transported up or down by a wire winch, making it able for the horizontal
robot to clean another floor.
For the sensory robots, a number of built-in guide-type robots must move along predefined rails.
They can measure the contamination level of the building’s surface using detection sensors and
can adjust their movement speeds accordingly. Thus, moving slower at places with high contamination
level creating more focused cleaning of the façade.
The domestically used robots move in simple ways as well. The very basic domestic robots
move in arbitrary directions until faced with an obstacle. When this happens the robot turns and
faces an new direction to continue cleaning. This arbitrary moving is obviously not efficient that’s
why the most common window cleaners do have some form of logical motion planning.
Most cleaners start at the top of the window, then proceed to move down the windows surface
in horizontal lanes. Cleaning with water causes water drippage down the window due to gravity.
Starting at the top of the window makes sure the robots are also adjusting for this the dripping of
water down the window by cleaning their own spillage.
More advanced robots try to estimate the dimensions of the window by making ’smart movements’,
which are then followed up by repeated movements to clean the complete windows surface.
These smart movements are based on different methods. For example, these can be done
using vertical positions estimation units, which are similar to the system used in old computer
mouse devices. Where the robot uses a ball to track its movement relative to its starting point and
this way knows the dimensions and its position on the windows surface.
On top of most systems, a control loop is used to keep track of displacements. This method prevents
cleaning areas twice or missing places, leaving areas uncleaned which is undesired.
It is also possible for these robots to have remote control features. This means the user can
decide over the motion of the robot according to what’s desired. If the user feels the window
cleaner missed a spot or should re-clean a certain area, the user can remotely steer the robot to
the corresponding place on the window.
The reason this standard movement in horizontal lanes is flawed is a consequence of the robot
only knowing its movement, and not having any knowledge about the state of the windows surface.
This makes it so the robot does not know if the window is actually clean, after completing
its cleaning pathway. As mentioned before the user can often manually correct mistakes or insufficiently
washed areas, but it is expected by the users that such a window clean robot can
fully clean a window autonomously. This means the robot needs a way to get information about
the current state of the windows surface and, on top of that, a corresponding course of action to
correctly adjust its cleaning according to this knowledge of the windows dirtiness.