different options in the novel dual surface type-2 fuzzy logic controller.
7.6
Introduction to Dual Surface Type-2 Controller
In this set of experiments three variations of the dual surface type-2 twin surface controller were investigated and compared with each other. The three controllers were the threshold dual surface, the minimum dual surface and the maximum dual surface type-2 controllers. The purpose of the experiment was to discover if there were any differences between the controllers when they were used to control a micro robot.
7.6.1
Method for The Threshold Dual Surface Controller
The setup used in the dual surface type-2 controller investigation was the same as that used for the comparison of the three controllers. Prior to starting the experimental runs, the common robot control image was download to all the robots using the PonyProg download utility. Again four robots were used in the experiment. A threshold configuration file was created with the parameters set as follows, .pF=2 .pZ=2. This configured the robots to use the threshold option. The threshold parameter was set to seventy (.rT=70) in the configuration file. The configuration file was downloaded to three of the robots using the Microsoft Hyper terminal communications package. An average configuration file was created with the parameters set as follows, .pF=1 .pZ=2. This file was downloaded to the fourth robot. Two tests of the threshold dual surface controller were ran. The first test experiment was run without ramping being applied by setting the parameter .rA=0. So the robot controllers had to handle a straight step change from a standing start. Three robots were used to try and counter the reliability problems that had previously been observed. The aim was to get at least ten valid runs across the three robots for every test. By setting the fourth robot as an average dual surface type-2 controller it provided a control and it was possible to directly compare the threshold results with the results for a baseline robot as soon as the run was completed. The second test experiment was to run with ramping being applied by setting .rA=1. This was done directly through the Microsoft Hyper terminal for each of the four robots.
Having set up the robots the Robot Soccer Engine was used to control the runs. Again five sets of test runs were made one for each speed demand of 10, 20, 40, 80 and 120. If any obvious failures occurred, such as collisions, then the run was repeated. There was no supervisory commands made during the runs other than the start and stop commands.
7.6.2
Results for the Threshold Dual Surface Controller
The same criteria that was used in the three controller comparison experiments was used to determine if a run was valid. Also the Pythagorean adjustment was made to the results if the robot had initially slewed and then travelled on a straight trajectory. For the tests of the threshold dual surface controller it was found that the controller could not control a step speed demand of 120 on any of the threshold robots. Many extra runs were made when a step was applied, with robots colliding with each other and hitting the sides before the 127 cm minimum distance was achieved. When a ramp was applied for the speed demand of 10, the controller could not move the micro robots for the minimum distance of 127 cm without the micro robots stalling at some point in the journey making the speed calculations inaccurate. The results obtained from the test runs are given in Tables E.12 and E.13. In the threshold dual surface controller test number 102 for speed demand 20, the average dual surface control robot four failed to record any valid runs. So the results of the three controller comparisons from Table E.10, for the average dual surface type-2 controller, test 3, were used as the control.
The mean and standard deviations for the dual surface controller with a threshold value of 70 across the three robots running without a ramp are given in Table E.14.
The corresponding results for the average type-2 controller are given in Table E.15. The results show that the average type-2 controller produces a higher speed for the speed demand in three of the four speed tests. This result was rather disappointing given that in simulation the dual surface type-2 controller outperformed the average type-2 con- troller. Examining the results show that micro robot number three was consistently under performing when compared with the average speed of the other three robots. Studying the position data of all the robots, it was observed that there were times when the reported position had not altered. These mini stalls were seen more frequently in robot three com- pared to the others and would explain why a lower speed was obtained. This difference
7.6. Introduction to Dual Surface Type-2 Controller 175 had not been observed during the baseline comparisons. However the robot had experi- enced problems due to loose wheels and that they might have been overtightened when attached to prevent them from falling off. During the tests it was not noticed that robot three was usually behind the others, as would be expected from the results. For the thresh- old test with no ramp applied, the global means are lower except for the speed demand 20 runs in test number 102. The global means are all within one standard deviation of the mean of the average dual surface control robot four. Removing the poor performing robot three from the comparison does not change the number of global means that are lower, but moves all the means close to the mean of the control. For the ramp test the global means are lower except for the speed demand 40 runs in test number 111. Again removing robot three from the comparison does not alter the result. With the results so close this allows the claim that the threshold dual surface controller is a safe and creditable alternative to the classic type-2 fuzzy logic controller when used to control micro robots.
7.6.3
Method for the Minimum and Maximum Dual Surface Con-
trollers
Having completed the threshold tests the robot’s configuration was changed to operate as a minimum dual surface controller by setting .pF=3, using the Microsoft Hyper termi- nal communications package. Only the ramp option was used in these tests with speed demands of 20, 40, 80 and 120. This was due to the problems experienced during the threshold tests, with the response to the step change. When completed the robot’s con- figuration was altered to operate as a maximum dual surface controller by setting .pF=4, again using the Microsoft Hyper terminal communications package. The results of the two tests are given in Tables E.16 and E.17. In the minimum dual surface controller test for speed demand 20, the average dual surface control robot four failed to record any valid runs. So the results from the equivalent maximum test for robot four were used.
7.6.4
Results for the Minimum and Maximum Dual Surface Con-
trollers
The results show that there is a distinct difference between the means of the minimum and maximum surfaces, however they are not significantly different when a t-test was applied. The global means of the minimum dual surface controllers are all less than the global means of the maximum dual surface controllers. Comparing the global means of the minimum dual surface against the mean of the average dual surface control robot shows that the minimum global means are less than the means of the average dual surface control. This is also true for the three individual robots except for the speed demand of 80
runs for robot one. In this case the difference is 1.93cmsec−1 with a standard difference
of 1.844 for robot one and 2.76 for the control. The global means of the maximum dual surface controllers are not greater than the mean of the control. The differences are all within one standard deviation. If the poor performing robot three is removed from the comparison, then individually the means of the maximum dual surface controllers do exceed the control mean, except for the speed demand of 20 runs for robot two.
7.7
Comparison of Three Controllers
In this experiment three controllers were compared with each other. They were a PID con- troller, a Type-1 controller and the dual surface average Type-2 controller. The purpose of the experiment was to discover if there were any differences between the controllers when they were used to control a micro robot.
7.8
Method
The experiments were carried out on the robot soccer pitch. Three test runs were per- formed in the experiment. Each test run had four robots running the strategy. The first test run consisted of two robots running the PID controller and two robots running the Type-1 controller. The second test run consisted of two robots running the dual surface average Type-2 controller and the other two robots the Type-1 controller. In the third test run two robots ran the PID controller and the other two robots the dual surface average
7.9. Results 177