Soft robots and smart movement

Links table

Abstract and 1 introduction

2 relevant business

2.1 Air operation

2.2 Aerobic weapons

2.3 Control of soft robots

3 Paul: Design and Manufacturing

3.1 Robot design

3.2 Choose materials

3.3 Manufacturing

3.4 operating bank

4 Gain data and control the open episode

4.1 Device Preparing

4.2 vision capture system

4.3 Data set generation: table -based models

4.4 Open ring control

5 results

5.1 final version of Paul

5.2 Analysis of the work area

5.3 Perform the models based on the table

5.4 Bending experiments

5.5 Weight experiences

6 conclusions

Finance information

A. Experiments and references

5.3 Perform the models based on the table

The size of the table was set to achieve the accepted kinetic modeling in an experimental manner, as no previous references were available and the previous works in the literature were very variable in terms of the number of data required. Moreover, the possibility of an error in the air system or the temporary store of the vision acquisition system collapses, along with the possibility of always leakage in the parts, making it recommended to take data in small sessions and then unite all they. Since the data collection process automatically, this was not a big problem.

Although the possibility that the surrounding temperature was a factor that affected the robot movements that were considered during the process of collecting the data set, it has recently been proven that small differences in the temperature did not affect the behavior.

Table 3 shows the data taken, the total time needed to get it and the average time for each point. It should be borne in mind that not all the situations that were taken in the end were used, because if the camera does not properly discover the positions of the three beacon areas, it cannot be calculated by the three -phase trend and thus restored the code of error. Of 1,200 samples collected, 5 % had to be disposed of, leaving 1146 usable in the end. The average time for each point was, given all the collected data sets, 6.76 seconds, with a standard deviation of 0.63 seconds. The low contrast is proven between the effectiveness of the designed automatic method.

Once all data collections are combined, direct and reverse motor models are validated here. Verification of the validity of the direct form of sending a robot mixture of inflation times and measuring the distance between the location that was reached, which was captured by the cameras, which the table predicted. Repeat the experiment for 40 points, the graph of the results presented in Figure 15. The average error is 4.27 mm, the medium error is 2.72 mm, and the standard deviation is 1.99 mm.

It seems that the high standard deviation and the shape of the graph, which is touched towards low values ​​and a very long tail, indicates the presence of points where the model offers noticeable failures with others with very good results. The future attention line can be the detailed analysis of the work space to determine where the areas of low accuracy of the model are trying to search for failures, and may lead to greater density of points in the data set.

In the same way, the reverse motor model was tested. To do this, Paul was given a reference site and guidance to achieve it, the necessary times were calculated, using the procedures referred to in equations (16), (17) the inflation was performed. After that, the situation that was seized with cameras was compared to the required mode.

As expected, the presence of repetition, where the values ​​of the extensive position are achieved with very different groups of inflation, provides great uncertainty in the model, which cannot capture triangles.

Specifically, the reverse kinetic model has an average 10.78 mm error, an average error of 9.22 mm and a standard deviation of 5.98 mm. Although these errors may seem high, they are compared in Table 4 with other open -loop control units presented in the literature. We can clearly see that they are in line with the results obtained and that they are better than those that the smaller robots got, where one expects, due to the smaller work area, a higher accuracy (at least in data -based models).

However, it is worth highlighting two experiences in which Paul performed a very satisfactory performance, because the field of operations is limited to an area where the repetition was not found. It is available in the appendix video a.

At the beginning, the robot was forced to reach a set of points on the horizontal plane, forcing the lower end of the last part to be parallel to

airplane. In each of them, errors were achieved less than 7 mm. Figure 16 shows the results of the aforementioned experience. In order to facilitate the understanding of the experiment, the lighthouse has been changed to the laser index that indicates the required points, which have been marked with goals of half a diameter of 5 mm, allowing the accuracy achieved.

In the second experiment, shown in Figure 17, the points on the bottom horizontal level were considered as a reference, but without the assumption that the lower face of the robot should remain parallel to it. In this case, 2 cm resolution was achieved.

Although the reverse kinetic model displays acceptable results, it must be commented that, in all these experiments, due to the engineering and robot materials, while Paul reaches the required mode, it tends to get. I made an attempt to reduce it, despite everything, it is a very fundamental phenomenon for the robot that is difficult to solve. The proposed future line, in this sense, is to try to organize the stiffness of the robot by inserting the negative pressures that generate the void.