EXPERIMENTAL SETUP OF FUZZY CONTROLLER!!!
The experiment continued until the boat entered the berthing stand-by zone, at which point the autopilot system switched automatically from the approach mode to the berthing control mode. (Note that in this study, the berthing capabilities of the autopilot were not evaluated, and thus the experiments terminated as soon as the system switched into the berthing control mode.)Figure 15(a) shows the trajectory of the FRP boat relative to the leading line during the automatically-controlled approaching maneuver when the membership functions defined by Figs. 8-10 are used. The results clearly show that the boat converges toward the leading line as it approaches the berthing stand-by zone. Therefore, the efficacy of the fuzzy-based autopilot system in accomplishing the approaching maneuver is confirmed. Figure 15(b) illustrates the variations in the input and output variables of the fuzzy controller during the approaching maneuver. In angle, Î¸2, maintains a relatively low value throughout.This result is the consequence of the small separation distance between the front and rear targets on the berthing wall (i.e. xd = 2.7 m). The profiles level-off after approximately t = 120 seconds since at this point, the boat entered the berthing stand-by zone, and thus the experimental measurements were discontinued. Figure 15(c) illustrates the time-varying distance of the ship to the berthing wall as estimated by the image-processing scheme (solid line) and measured by the DGPS system (dotted line), respectively. Although the estimated results are in broad qualitative agreement with the exact results, it is clear that quite significant errors exist. The time-varying discrepancy between the two sets of results is illustrated graphically in Fig. 15(d). From inspection, the error in the estimated distance value is found to vary in the range of approximately 10~20%. In addition, it is observed that the magnitude of the error tends to increase as the boat approaches the berthing wall. This result is to be expected since the rear target elevation angle Î¸y increases significantly as the ship nears the quayside, and thus the value of the ship-to-berth separation distance computed using (9) becomes increasingly sensitive to errors in its estimated value. The image-processing scheme is designed such that in the event that the targets disappear from the CCD screen (as the result of course changes, for example), the previous estimate of the shipto- berth distance is retained as the current estimated value. Thus, as shown in Fig. 15(c), the time-varying estimated distance profile has a pronounced stair-like characteristic. Since the FRP boat used in the current experiments is small (i.e. 4 m in length) and therefore highly responsive, large rudder commands result in an abrupt change of course and cause the CCD camera to lose track of the leading marks.
Thus, unless suitable precautions are taken, the imageprocessing scheme can not compute feasible values of the input variables Î¸1 and Î¸2.