Set up of an Automated Multi-Colour System for Interior Wall Painting

Abstract

It is shared that construction projects are getting bigger and more complex, hence also the productivity of the construction industry must be improved, while preserving its labour from hazardous job sites. Such requirements can be accomplished by the adoption of robotized products, which, however, need to be quickly developed and marketed. In this paper, first the issue of a new miniature laboratory for developing lightweight and well-coordinated robotized systems is pursued, then a novel robot device for high quality multi-colour interior wall painting carried by a robot arm is developed and successfully tested. Thanks to the new 1:6 scaled down laboratory and its six degree of freedom robot arm on an hexapod for horizontal moves, we tested the opportunity to introduce also in the building sector miniature robots that can change the ergonomics standardly adopted by construction workers. It is analyzed how and why switching from full size to miniature robots is convenient in construction. In addition, a new system adding further features to robotized painting has been conceived. Our new multi-colour spraying end-tool was developed and fixed on the robot arm, in order to be able to reproduce coloured artworks. Finally, a methodology to reproduce colours from digital format of artworks is presented, showing how accurate and efficient is this new robotized spraying device.

Keywords

miniature painting robot scaled down laboratory multi-colour spraying end tool

1. Introduction and problem statement

The main advantages that can justify the adoption of robots in the construction industries are generally grouped as follows (Kahane, B. and Rosenfeld, Y., 2004):

–
improving productivity and save human labour;
–
allowing higher quality standards for buildings.

When construction workers and robots are properly integrated in building tasks, the whole construction process can be better managed and savings in human labour and timing are obtained as a consequence. In addition, it would offer the opportunity to reduce or eliminate human exposure to difficult and hazardous environments, which would solve most of the problems connected with safety when many activities occur at the same time. Finally, automated machines are able to produce higher quality products, as they would promote the passage from craft production to industrial production, with consequent higher accuracy and lower levels of defects.

As listed in the next section 2, there are a lot of contributions pertaining to the improvement of efficiency and safety in construction, which are brought about by automation, provided it will be able to produce highly autonomous robots for building sites. The main requirements presently pursued are: developing dedicated robots to execute the same activities that workers are usually in charge of; equipping robots with the sensors and control systems required to make them autonomously move throughout every job site.

However we think that the other objective, namely that of quality improvement, should be reconsidered to be not only connected with better accuracy in production, but also strongly linked to the possibility to adopt new technologies and ergonomics for the execution of building tasks. As a consequence, when we think about building automation we should take the opportunity to reconsider new procedures and new tools to accomplish every task.

Recent research in the building sector has been mainly devoted to automating interior finishing tasks. Among them we have chosen to start working on interior wall painting. Thus far, some very interesting robots for interior wall painting have been manufactured, like in (Kahane, B. and Rosenfeld, Y., 2004). They have been conceived to be able to make wall spraying in such a way that reproduces what performed by human workers. Big efforts have been made to improve their efficiency and the robustness of their control systems, too. We think that such very valuable research in construction can be further improved by exploiting the possibility to use automatic systems to re-think new procedures for executing building tasks. In the case of wall painting, we are studying the feasibility of a new approach for interior wall painting, which gives users the possibility to paint with a multicolour device, that would be able to reproduce artworks similarly to what is done by printing machines: this is also intended as an example of how known industrial technologies can be adjusted for convenient applications in construction. Moreover, architects will be free to design new environments, even because this system can be used as well for standard spraying on interior walls. In this paper we have tested one four-color spraying system.

Another big advance that can be brought about by automation in the building sector, is the one of rethinking new ergonomics. Robots are able to work at different scales than human workers: (Bach, Fr. W., Rachkov, M., Seevers, J. and Hahn, M., 1995) maintain that there are some tasks, including painting of big construction areas, which are difficult and expensive to perform using standard ergonomics. They suggest the use of climbing robot, equipped with many sensors to coordinate one another.

However there are a lot of other situations where human sized ergonomics is again the best approach: hence we have set up two parallel laboratories, one for testing at full size and another one which is thought to test miniature robotic systems, as the two solutions will complement each other.

In this paper all the experiments have been carried out at the smaller scale, because it is considered to be the most critical. Our successful experimental campaign shows how convenient the use of such new technologies can be: this way it is possible to widen the set of performances allowed for the painting tasks, without renouncing to the efficiency and the other advantages traditionally obtained.

This paper is organized as follows: section 2 summarizes the state of the art in this field of research; section 3 outlines the research that we have already performed, whose tests are shown and commented in section 4. Conslusions and future research are summed up in section 5. All the tests take advantage of the extensions listed in section 7 and references in section 8.
2. State of the art

Interesting examples of automated construction processes are the CIC (Yasuyoshi M. and Roozbeh K., 1993) and the SMART systems (Yamazaki Y. and Maeda J., 1998). CIC is the acronym for Computer Integrated Construction: it requires the use of a mobile contruction system moving throughout the building site. Even if it requires the adoption of ready made products and it would be difficultly applied for different construction technologies, however it laid the foundation of automation in the building sector with wide use of robots and control systems. Another similar example is represented by the S.M.A.R.T. system, developed by Shimizu Corporation, which encompasses also the design and management phases in the automated construction process.

The very interesting contribution given by the aforementioned (Kahane, B. and Rosenfeld, Y., 2004) and (Warszawsky, A. and Rosenfeld, Y. 1994) are more specific for the execution of building tasks. They are developing the multi-purpose TAMIR robot, which has been used to perform a number of experiments on wall painting using a spray gun with its on/off control nozzle operated by connections of the robot's controller. Through full scale experiments it was shown that, according to the break-even point technique, robots are always more profitable than human work when highly autonomous machines are adopted. This holds because one operator may supervise also two different robots; in this way it is convenient also for cheap labor markets. This is due to the fact that human workers are released to the simultaneous performance of complementary subtasks, which might improve the profitability of the whole system. TAMIR was tested also in the execution of other tasks like plastering, as performed in (Forsberg, J., Aarenstrup, R., Wernersson, A., 1997), about which there is however less information available. The main purpose of these experiments was connected with the necessity to develop highly autonomous machines for building sites: even if that problem is not completely solved, however first good results have been obtained, which perorate the feasibility of this technique. All these experiments follow the common approach of testing human ergonomics prototypes.

At the same time, another school of thought is working on miniature robots, which can be used for a number of purposes. Some of them are climbng robots for inspection and diagnostic (Bach, Fr. W., Rachkov, M., Seevers, J. and Hahn, M., 1995); climbing robot for curtain wall cleaning (Zhang, H., Zhang, J., Zong, G. 2006), for miniature navigation (Yilma, B., Seif, M. A. 1999). These robots have been developed because they are more lightweight and move at high speed, thus allowing operations that would be very cumbersome, and sometimes impossibe, in case of use of heavy and slow automatic machines. Among the wide variety of applications where they can be used, there is also the possibility to adopt them in construction: they could be used to access high rise building walls or dangerous environments and perform many operations, including wall painting. They can constitute also one valid alternative to the painting robots thus far presented, because their lightness enables them to quickly move throughout every job site, despite the presence of obstacles or other operators.

Other applications of miniature robots in different fields of research strengthen the validity of our proposal and the interest in this approach. In (Stoeter, S. A., 2005) the development of a control system for miniature robots is investigated. The author observs that untethered miniature robots with physical dimensions around or lower than 10 cm do not permit powerful onboard computers because of their size and powerful constraints. This limit has reduced in the past their functionality to that of remote controlled robots. In this contribution control is performed by a larger companion robot, that closes the loop using a video-based approach system. Such a framework was experimentally validated by performing the task of staircase climbing, which resulted to be feasible. Another robotic miniature system has been proposed for reconnaissance and surveillance by (Hougn, D.F. et al., 2000). It has been conceived to be able to carry out a wide range of missions, thanks to its multiple mobile modes and sensing capabilites. Data are then transmitted to a “ranger” robot or to a human leader. Both the “ranger” and miniature robots are able to accomplish together very complex operations.

The other aspect that will be studied in the next paragraphs is concerrned with multi colour painting. This concept is not new in the field of manufacturing, like the ones offered for purchasing in many websites, e.g. (Goldspray, S.L., 2001). But this concept is original for construction. Manufacturing industry presents some examples of industrial plants for color mixing, that are aimed at reproducing every type of hue, saturation and brightness, starting from some primary colors. However those plants are very heavy and dangerous, as they store big amount of paints and work in airless mode, hence at high pressure. Moreover they are simply connected to a spraying gun, with no control system. The step forward that we would like to pursue by integrating a new light multicolor device within robots for contruction is given by the possibility to reproduce the functioning of printing machines for interior walls of buildings. It can be accomplished in case the multicolor device is opportunely coupled with robot arm's movements, as it will be widely explained in the next section 3.

3. Research course

3.1. The new experimental approach

Our approach is focusing on two main concepts: allowing new ergonomics to better perform construction works and exploiting all the features uniquely provided by automation to produce high quality works, otherwise not feasible by human workers.

The development of new ergonomics is necessary just to improve the feasibility of automation, which is otherwise limited when it tries to reproduce human ergonomics. In order to better express our concern, we could think about what happened with painting: some prototypes have been manufactured reproducing the same working procedures presently adopted in building sites: they require the same tools and size of working area. At this stage the researchers have obtained very good results: they have set very good automatic machines that can work as humans do, hence with the same limits. To cite some of them we can notice that robots are generally quite heavy and cannot access every place in building sites; they would need strong and long braces to hold robots when painting external walls of high rise buildings. That means that human size robots must be combined with other nimbler machines in order to provide a complete service for building painting. When both technologies are available, robots will be able to completely replace human work.

We have also one example of this kind, which is fully described in (Zhang, H., Zhang, J., Zong, G., 2006) and summed up in Fig. 1: even if wall cleaning has historically been performed by men working on suspended platforms like in Fig. 1-a, the coming of new technologies helped inventors think about the new and more efficient technology depicted in Fig. 1-b. The second case is more feasible and efficient because it does not ask for the use of strong and heavy supporting devices to lift operators up to very high levels; hence equipment costs are lowered and a smaller area of work is required. In addition, no worker is required to access high level and dangerous platforms like in the first case, eliminating risks of injuries. The same solution would not have been pursued in case heavy cleaning robot had been developed for such a purpose.

Fig. 1.

Human ergonomics (a) and automated ergonomics (b) adopted for glazed envelope cleaning of high rise buildings.

The same concept holds for painting robots moving through building sites or close to external walls: heavy machines can reproduce human ergonomics and work properly just in case they have the required room for their movements and stay on strong floors. In case accessing some rooms to be painted is very difficult or dangerous due to the conditions pertaining the specific building site, then heavy machines would be not suitable and the use of different products would be required. Hence our research is aimed at developing two kinds of automatic technologies for painting. The first is relative to a reduced ergonomics for painting, where some small and lightweight robots are coordinated one another. The second is based on a human-like ergonomics, reproducing the same movements and adopting the same tools as it is done by human workers. In the first case a multitask control system will be developed, which is thought to be able to manage the movements of every single robot and to coordinate their relative positions: it will require also the use of many sensors, some of them aimed at recording the extension of painted surface (Fig. 2). This second part of the control system is not necessary in case of robots working according to human ergonomics.

Fig. 2.

Multitask control for miniature robots.

The two solutions will be complementing each other, meaning that they will be opportunely selected according to the features of the specific environment that will be painted. Considering the necessity of performing experimental campaigns, two robots have been set up inside as many laboratories depicted in Fig. 3. The first one is widely described in the next subsection 3.2, and it has been used to develop the scaled down painting end tool for the miniature approach. The other one, shown in the second part of Fig. 3, was not tested again, as it will be employed in the next step of our experiments, being more urgent to study the feasibility of a miniature painting end-tool at this first stage. Once that painting in small and full scale is shown to be possible, then further experiments concerning the two integrated technologies will be made. Moreover we will address the problem of high quality multicolor painting: this way we exploit all the capabilities that are put at disposal by the adoption of automated technologies. Multicolor means not only that architects can choose which color to attribute to walls, but that they could decide to reproduce drawings, whose resolution is determined in this paper: anyway, the smaller the end-tool the better the resolution.

Fig. 3.

Robots for scaled down (a) and full scale (b) experiments.

Hence that approach is preferred at the miniature scale, because resolution is higher than at full scale, as it behaves in inverse proportion to the size of sprayed paint flux.

3.2. Conceiving the miniature laboratory

The main component of the 1:6 scaled-down laboratory is its miniature robot, which is depicted in Fig. 4. It is a 6 degree of freedom robot controlled by a 32 channel servo controller connected to the serial port of a PC, having a nominal reach of 0.4 m and a payload of 40 N. The robot is made up of a supporting rotating base fixed on a two degree of freedom hexapod for horizontal moves (not exploited for the experiments proposed in this contribution), an arm with three rotary motions whose end is equipped with a rotary wrist and a gripping hand (Fig. 4-a). The whole robot owns 6 servos (one for each degree of freedom excluding the hexapod). For this testing the gripping hand was substituted with the aerograph, that utilized the servo formerly used for gripping to rotate horizontally. Every servo is controlled by a small preassembled servo controller with some interesting features. It has high resolution (1uS) for accurate positioning and extremely smooth moves. The range is 0.75uS to 2.25uS for an angular range of about 170°. The motion control can be immediate response, speed controlled, timed motion, or a combination. A unique “Group Move” allows any combination of servos to begin and end motion at the same time, even if the servos have to move different distances. Every servo's position or movement can be queried to provide feedback to the host computer. Thanks to this arrangement it was possible to reproduce the movements that are used to paint plane walls with aerographs like in Figg. 4-b and 4-c. The robot is powered by one current and voltage regulator (range from 0 to 12 V and from 0 to 6 A).

Fig. 4.

Scaled down robot (a) and trajectories used for wall painting (b and c).

The laboratory is surrounded by small plaster made surfaces to reproduce building walls mounted on three sides.

Finally, one compressor was used to send high pressure air flows towards the aerograph installed on the end tool of the robot and to the color mixing device described in the next paragraph.

3.3. Mixing device of the miniture laboratory

Every elementary painting equipment is made up of at least the following components: one spraying gun, a tank containing paint and a device to lead paint out of the tank and send it to the spray gun. The first step towards the development of the mixing device needs the choice of the best painting technology. There are three main approaches for painting: Airless; Aerography; Airmix.

The airless painting system works at very high pressure: paint is sent to the spray gun at about 1.2·10⁶ Pa (to be created by a pump), whose precise value depends on its physical and state characteristics (density, viscosity, temperature, surface tension, solvent volatility, speed). In this case the jet stream shape is uniquely determined by the nozzle slot, which is generally chosen as vertically oriented. Instead the flow rate of paint is determined by the nozzle extension, which is generally limited from 3.1·10⁻⁸ to 1.7·10⁻⁶ m² (Perrin, H. and Zanassi, R., 1982). Paint flows through the nozzle at an average speed of about 140 m/s and it is held at about 0.25 m from the wall. Instead aerography produces a jet where air is mixed with paint, which is in turn sucked by air into the gun. It is different from the airless type, that directly atomizes paint through high pressures; in this case paint is sucked by high pressure air (about 4·10⁴ Pa) flowing through the spray gun towards the nozzle. Sometimes the gun is fed by gravity or by low pressure, in order to help paint to pass through the gun, but the principle does not change, because in this second case high pressure air is always in charge of paint atomization. This technique is older than the previous one but is not suitable for quick drying paints and produces bigger drops and thicker films, hence lower quality products. In addition, its application is not simple because it generates a paint cloud around the nozzle and it could be dangerous for workers. The airmix system tradeoffs between the two former possibilities, and it is very useful for high precision works.

It is practiced through a nozzle that owns three holes: one in the centre working in airless mode; other two holes on the sides blowing high pressure air to mold the whole paint jet into the desired shape. Paint is blown between 3 and 6·10⁵ Pa, and air is tuned according to the wished final result, whose pressure's magnitude is lower than for aerography (generally between 1 and 3·10⁴ Pa). The success of this new system is however correlated to another requirement: it would revolutionize the concept of building painting just in case it will be able to draw high resolution images. In such a case the robot should be able to change the color of the sprayed painting while moving its arm, which means that it would reproduce also complex textures. In this case the system would overcome other similar applications, known just in the industrial field: they basically use very heavy paint distributors, which would be difficult to move. Moreover, they mix colors inside the distributors, which is far from the nozzle, requiring long time delays to switch from one color to another. Fig. 5 shows the concept on which it is based, where we can notice that it is too heavy to be moved throughout building sites and so large that a long delay between the time when it starts to change the color mix (through a change in pressures) and the time it comes out into a different color sprayed on the wall is required.

Fig. 5.

Example of mixing device for industrial applications and two details of the main figure, included in box a and box b.

For that reason our system is equipped with a high precision aeroghraph, that was adapted from a normal aerograph in such a way that it can be used for airmix spraying.

Fig. 6 shows the functioning of the whole system: the four primary colors have been put inside high pressure tanks to be sent to the mixing board (Fig. 7), which turns up or down its valves to proportion the colors' relative amounts, previous to their blending into the mixing device close to the gun (Extension [1]).

Fig. 6.

Sketch of the final multicolor mixing device.

Fig. 7.

The mixing board.

All the system has been designed to work with four primary colors: cyan, magenta, yellow and black. Before showing the whole experimental set up it is necessary to understand how the mixing board in Fig. 7 works. The four color tanks push paint towards the mixing board at high pressure of 1.5·10⁵ Pa, which is equipped with four ball valves. Every valve is connected to a servo (deflecting torque equals to 0.132 kg·m) which turns up or down the amount of paints flowing through it, establishing the percentage of primary colors to be mixed in the following device and hence deciding the hue of the final color. Every servo is driven by the same controller as for the robot, such that color proportions are coordinated with the robot's movements.

From the mixing board the primary colors move towards the mixing device in Fig. 8-a, which is a modelled aluminum parallelepiped containing one transversal positioned steel spiral, helping the colors to be mixed into the final one by changing the flow movement into a turbulent one. This device is very close to the spraying gun (Fig. 8-b), that is the place where paint is mixed with air flowing at 3.5·10⁵ Pa as soon as it goes out of the gun. To be noticed that the mixing device is positioned at a very short distance from the gun, equals to 0.08 m, in order to lower its inertia of color changing. Hence paint is partially atomized in the gun by the high pressure set in the nozzle, and that process is completed at the time when it is sprayed, that's to say just out of the gun before hitting the wall. In Fig. 9 it is possible to appreciate the whole experimental equipment, as installed in the scaled-down laboratory. High air pressure (6·10⁴ Pa) is sent by the air compressor towards the two pressure regulators: the first is used to regulate air pressure at 2.8·10⁴ Pa before it is directed towards the nozzle; the second limits air pressure at 1.5·10⁴ Pa before it enters the two storage thanks. This way paint comes out from the tanks at a final pressure of 0.8·10⁴ Pa, that was considered to be good for this kind of paint and gun.

Fig. 8.

The mixing device (a) and its relative position to the spraying gun with the nozzle for airmix painting (b).

Fig. 9.

The whole multicolor miniature painting system.

No pressure drops have been recorder along the other path followed by air. All the system is equipped with three safety valves. Then paint is sent to the mixing board, where the four servos drive their valves to the required opening, according to the inputs from the control system. The four paint colors flow towards the mixing device, that increases turbulent actions in order to spray a homogeneous painting, that is athomized at the gun-nozzle level, fixed on the robot arm, which is in turn powered by the voltage regulator. The control system gives the warranty that the robot's movements are coordinated with the valves' openings to reproduce the required color textures.

4. The experimental campaign

4.1. Set up of the spraying parameters

Even if the system has been designed for four colors, however all the experiments have been performed with three colors: cyan, magenta and yellow. We know that in this case a narrower range of colors can be reproduced, but it is enough for our experimental purposes.

Preliminary experiments to set the opportune spraying parameters have been performed as detailed in (Naticchia B., Giretti A. and Carbonari A., 2006). There are two different sets of paramters: those ones which are fixed by standards and by the technology adopted, and those ones depending on the expected quality for the final product. The first group embraces type of paint (in this case waterbase paint), type of application (in this case airmix), spraying pressures (in this case 2.8·10⁴ Pa for air and 0.8·10⁴ Pa for paint), end-tool trajectories (please refer to Fig. 4), distance from the wall (in this case 0.08 m). Instead variable parameters are given by the distance step between every line of the trajectories and the couple speed-paint flow, which determines the paint thickness on the wall. A detailed study to set the most opportune values for these variables has been carried out in the way already detailed in the aforementioned paper. The quality of the final work has been evaluted using two indicators:

–
saturation: indicates color intensity and purity;
–
thickness uniformity: evaluates thickness variation on the wall surface, that must be as constant as possible to avoid that visual perception changes with respect to the direction from which the surface is looked at.
–
The first was measured using an image processing software, restituting how far from the totally 100% saturated condition is a surface after its coating. The second is important to avoid too big variations of paint thickness on the wall, that could cause a particular reflection of light, that would be perceived by human eyes as a vertical or horizontal stripped surface. The procedure for parameters optimiziation asks for the collection of one statistical sample (formed by at least 10 measures) whose characteristics vary according to the parameter to be set. In the previous experiments speed of arm motion was left to vary between 0.036 and 0.11 m/s. For our experiments we chose to work at an average speed of 0.073 m/s, as it trades off between a low speed strategy (that reduces the size of the pixel) and the necessity to move fast in order to prevent the formation of droplets (in case too thick paint is sprayed on walls). At that speed it was found out that the paint flow giving the best results is equal to 3.4·10⁻⁷ m³/s, as it sprinkles on the wall a paint layer thinner than 0.044·10⁻³ kg/m², which has been shown to be the maximum allowable limit in (Naticchia B., Giretti A. and Carbonari A., 2006). The distance step between every line of the trajectories was set equals to 0.045 m, as suggested by the results published in the same paper. Once that these parameters have been set, other experiments to verify the accuracy and to table the features of the multicolor system have been perfomed, as expalined in the next paragraph.

4.2 Tests on the four colour spraying device

Given the preliminary experiments of the previous section, the multicolor system has been characterized and tested, in order to infer its features. Tests have been performed regarding:

–
analysis of precision;
–
level of resolution;
–
accuracy in color reproduction.

The analysis of precision has the main aim of establishing the accuracy of the system when it is asked to paint a given shaded rectangular portion of wall. First, the behavior of the ball valve has been analyzed: its characteristic curve describing the paint flow rate (Q) with respect to its rotation around its axis (measured in percentage) from the closed position has been drawn.

The valve has been turned on through steps of 15 degrees, and the corresponding flow rate has been measured as pictured in Fig. 10. The relation coming out from Fig. 10 is of basic importance: thanks to that, the controller will be able to compute the paint flow sprayed by the gun, starting from the opening of each valve of the mixing board (which is equipped with four identical ball valves). Working the relation in the opposite way, it is possible to compute the percentage opening required for each ball valve if a pre-established paint flow is required. This information has been used to compute the relative opening of each valve when a known color mixing is required. Once we know the value of the total flow (Q = 0.34·10⁻³ 1/s), it is necessary to work out single values for each primary color. The following relation must hold, meaning that the sum of all the paint flows cannot exceed the maximum:
$1 = \frac{Q_{C}}{Q} + \frac{Q_{M}}{Q} + \frac{Q_{Y}}{Q}$
(1)

Fig. 10.
Characteristic curve of the ball valves.

Where Q is the total paint flow rate, while Q_C, Q_M and Q_Y are the flow rates assigned to each primary color (cyan, magenta and yellow). From the theory of colors explained in the next paragraph it is possible to compute every relative flow rate and then the relation in Fig. 10 can be used to compute the required opening for each valve. At this stage it is possible to perform analysis of precision. It has been made by generating a sample of size 9, formed by nine shaded portions of walls of the kind in Fig. 11, changing from magenta to yellow and other 9 shaded from cyan to magenta. Through a software for imaging processing, the saturation of each sample on a 14 equally spaced column grid has been measured. In Fig. 11 we can appreciate that on each point of the grid, standard deviations measured for magenta on the selected points are low, and on the average equal to 8%. Even better results have been obtained for the other two primary colors, where cyan deviation from the mean are on the average equal to 5.5%. Table 1 sums up the first set of those results.

Fig. 11.
One of the samples for precision evaluation.

Table 1.
Precision evaluation at each point of the grid superimposed to sprayed samples like in Fig. 11.

Mean 86 84 83 79 71 68 63 58 54 47 41 36 35

s.d 2.5 3 2 2.9 4.4 3.3 4.4 4.1 6 7.5 5.5 5.1 7.3

% s.d. 2.9 3.5 2.4 3.7 6.1 4.9 7 7 11 16 13 14 21

The time delay has been measured by filming the movements of the servos on the mixing board and the colors sprayed on the walls: the time difference between color switching of the ball valve and the time when it is observed on the wall is the unknown time delay.

One of these tests is shown in Fig. 12. It has been found out that the switching time is equal to 3 s, corresponding to one pixel (0.13×0.20) m wide, which is the finest resolution feasible with this system.

Fig. 12.
System resolution.
4.3 Procedure for colour reproduction

Mean	86	84	83	79	71	68	63	58	54	47	41	36	35
s.d	2.5	3	2	2.9	4.4	3.3	4.4	4.1	6	7.5	5.5	5.1	7.3
% s.d.	2.9	3.5	2.4	3.7	6.1	4.9	7	7	11	16	13	14	21

One of the most important requirements for such a system is that it can accurately reproduce the colors of every pixel of drawings. For that purpose a general procedure for color reproduction has been set, that is based on the theory of colors in (Perrin, H. and Zanassi, R., 1982).

As we are dealing with a system working like printing machines, every color can be thought as composition of the three primary ones: cyan, magenta and yellow (generally also black is included to obtain a wider range in the visible spectrum, however in this experiment it was neglected). We will explain the general procedure while applying it to one real case. First of all one color belonging to the visible spectrum has been generated with the help of a software for image processing (Fig. 13-a), having the following normalized coordinates (C,M,Y) = (0.43, 0.06, 0.51). Every normalized coordinate is computed from the original CIE (Commission Internationale de l'Eclairage) coordinates (c,m,y) by defining for cyan:

C = \frac{c}{c + m + y}

(2)

Fig. 13.

Original color (a) and its reproduction (b).

and the same holds for magenta and yellow. Then it must be considered that the primary colors used by our system are not the same primary colors considered by the image processing software. Hence three wall samples have been painted using these primary colors and obtaining the following normalized coordinates:

\begin{aligned} 1^{st} primary color (C^{*}) = (0.67, 0.2, 0.13); \\ 2^{nd} primary color (M^{*}) = (0.12, 0.80, 0.08); \\ 3^{rd} primary color (Y^{*}) = (0.05, 0.00, 0.95) . \end{aligned}

(3)

The relative amount of the system's primary colors (C*,M*,Y*) to be used for the reproduction of the original color (C, M, Y) have been computed by solving the following system:

{\begin{cases} C^{*} + α_{c} M^{*} + β_{c} Y^{*} = C \\ α_{M} C^{*} + M^{*} + β_{M} Y^{*} = M \\ α_{Y} C^{*} + β_{Y} M^{*} + Y^{*} = Y \end{cases}

(4)

Where the coefficients α and β allow for the non purity of our primary colors. For instance, if we know that the system must spray a certain amount of cyan (C), we know that it is present not only in its cyan (C*) but has also other contributions in its magenta and yellow (α_c=12% and β _c =5%), as suggested by eq. (3), from which all the other coefficients have been determined. The unknown variables have thus been determined to be (C*, M*, Y*) = (0.476, 0.025, 0.54). Then applying eq. (1) in the following form:

1 = C^{*} + M^{*} + Y^{*} = \frac{Q_{C}}{Q} + \frac{Q_{M}}{Q} + \frac{Q_{Y}}{Q}

(5)

and by the knowledge of the optimum flow rate Q, it was possible to compute the single flow rate values Q_C, Q_M and Q_Y; from these values and from the diagram in Fig. 10, valves' openings were determined and the final color was sprayed on a plaster wall sample (Fig. 13-b), as shown in Extension [2] and in Fig. 14.

Fig. 14.

Four clockwise directed snapshots of the miniature robot's spray painting on a plaster wall.

Observing the two figures it is possible to appreciate that they look very similar. But this check has been performed also using a scanner acquisition system: it gave back the normalized values (0.53, 0.58, 0.31) for Fig. 13-b, which have been corrected by the knowledge of the shift introduced by that scanner (previously determined scanning pure primary colors from (Perrin, H. and Zanassi, R., 1982)), that is (+0.20, +0.06, −0.26). Hence the corrected value of the color sprayed by our system was (0.33, 0.1, 0.57), whose difference with the original one is (–0.1,+0.04, +0.06), that is the approximation introduced by the whole mechanical system.

4.4 Results

The first positive result pertains the ability of the small and lightweight robot to carry its end-tool, even when it sprays at quite high pressures, causing a strong recoil. Furthermore, analysis of precision gave very good results, as every shift is never higher than 10%. Also the finest pixel, (0.13×0.20) m wide, would allow users to perform quite detailed drawings, when painting very big walls. As far as reproduction of colors, we found that the shift introduced by the system is never higher than 10% (just in the case of cyan). Moreover we can notice that a very good coverage of the plaster wall has been obtained, even if just one layer of paint has been laid. We know that better results can be obtained by applying two layers of coating on walls, that will be experimented in later steps.

5. Conclusion and Future Research

Thanks to the mixing equipment developed throughout the research step described in this paper, we have shown that automated painting can be not only aimed at improving productivity, but also quality. In addition, two kinds of complementing ergonomics can be thought, one at reduced scale and the other one at human scale. Our future research will be aimed at discerning how it is possible to perform the most efficient choice among them, according to the working place and to the technology chosen. Moreover, it will be necessary to develop two technologies with very similar performances, in order to provide a whole system with interchangeable tools (reduced and human scale).

The small scale technology has been shown to have high precision, quite high resolution and very good accuracy in color reproduction. We know that even better results could be obtained by developing new painting products, more suitable for the narrow nozzles used for this small applications, by improving the precision of the mechanical system, of the nozzle and better choice of the primary colors.

When moving to full scale experiments, it wll be necessary to reproduce the same precision level, which should be obtained by providing a very accurate control system, similar to the one developed for small scale experiments: adoption of one mixing board equipped with four valves, each of them connected to a servo which turns up or down the amount of paints flowing through it and establishing the percentage of primary colors to be mixed for the final color. A robot arm with high precision is required, too. Accuracy in color reproduction should be faced in a way similar to what was done in this contribution: an automated system to translate the normalized coordinates of the colors to be reproduced, into the movement speed of the robot end tool and valve openings of the mixing board. Much more work will be probably necessary to obtain high resolution. Because of the size of full scale robots, probably also the resolution of the human scale robot will be lower. Of course we will be not able to obtain pixels as small as (0.13×0.20) m, however we think to be able to exploit the features of the airmix system to mold its shape as needed: air pressure and air flow passing through the two holes on the sides of the airmix nozzle can be tuned according to shape and dimension wished for the final paint flow. This will determine the threshold under which the adoption of the small scale system will be preferred to the human scale approach. Another peculiarity of the small scale system is of course the ability to access some difficult places of buildings under construction, where human size robots could not be allowed.

Footnotes

6. Acknowledgments

We have been indebted to prof. Mario De Grassi, full professor at the Polytechnic University of Marche – DACS Department of the Faculty of Engineering, for his helpful support and suggestions. We would like to thank the editor of this special issue for his kind invitation to publish our paper. Also the comments of the anonomous reviewers revealed to be critical for the improvement of the clarity and quality of the contents of this paper.

Index to Multimedia Extension

The multimedia extensions to this article can be found by following the hyperlinks from: www.ars-journal.com [1]

mixing board.wmp: video showing how the mixing board works.

[2]

system validation.wmp : video showing the robot at work when spraying the final wall portion used for validation of the procedure for color reproduction.

References

Bach

Fr. W.

Rachkov

Seevers

and Hahn

, (1995). High tractive power wall climbing robot, Automation in Construction, Vol. 4, No. 3, October 1995, pp. 213–224, Print ISSN: 0926-5805.

Forsberg

Aarenstrup

Wernersson

, (1997). A construction Robot for Autonomous Plastering of Walls and Ceilings, Proceedings of the 14th International Symposium on Automation and Robotics in Construction (ISARC), June 8–11 1997, Pittsburgh, PA, USA.

Goldspray

S.L.

, Manufacturing Company in Torrente Valencia - Spain, available on line at: www.goldspray.com/InglesWEB/Index2.htm, 2001.

Hougn

D.F.

Benjaafar

Bonney

J. C.

Budenske

J. R.

Dvorak

Gini

French

Krantz

D. G.

P. Y.

Malver

Nelson

Papanikolopoulos

Rybski

P. E.

Stoeter

S. A.

Voyles

Yesin

K. B.

, (2000). A miniature robotic system for reconnaissance and surveillance, Proceedings of the 2000 IEEE International Conference on Robotics and Automation, pp. 501–507, April 2000, San Francisco, CA.

Kahane

and Rosenfeld

, (2004). Balancing human and robot integration in building tasks, Computer Aided Civil and Infrastructure Engineering, Vol. 19, No. 6, November 2004, pp. 393–410, Print ISSN: 1093–9687, on-line ISSN: 1467-8667.

Naticchia

Giretti

Carbonari

, (2006). Set up of a robotized system for interior wall painting, Proceedings of the 23rd International Symposium on Automation and Robotics in Construction (ISARC), October 3–5, 2006, Tokyo, Japan.

Perrin

Zanassi

, (1982) “Verniciatura a.spruzzo”, vol. 5(2), La Rivista del Colore, Milano.

Stoeter

S. A.

, (2005). “Autonomous Stair-Climbing With Miniature Jumping Robots”, IEEE Transactions on Systems, Men and Cybernetics-Part B : Cybernetics,Vol. 35, no. 2, April 2005, pp. 313–325, ISSN: 1083-4419.

Warszawsky

and Rosenfeld

, (1994). Robot for interior finishing works in buildings : feasibility analysis, ASCE Journal of Construction Engineering and Management, Vol. 120, No. 1, March/April 1994, pp. 132–151, Print ISSN: 0733-9364.

10.

Yamazaki

Maeda

, (1998). The SMART system : an integrated application of automation and information tehcnology in production process, Journal of Computers in industry, Vol 35, No. 1, February 1998, pp. 87–99.

11.

Yasuyoshi

and Roozbeh

, (1993). Experiencing computer integrated construction, International Journal of Construction Engineering and Management, Vol. 119, No. 2, pp. 307–322, May/June 1993.

12.

Yilma

Seif

M. A.

, (1999). Behavior-based artificial intelligence in miniature mobile robot, Mechatronics, Vol. 9, No. 2, March 1999, pp. 185–206, ISSN : 0957-4158.

13.

Zhang

Zong

, (2006). Effective pneumatic scheme and control strategy of a climbing robot for class wall cleaning on high rise buildings, International Journal of Advanced Robotic Systems, Vol. 3, No. 2, June 2006, pp. 183–190, ISSN : 1729-8806.