Type-2 Fuzzy Control of an Automatic Guided Vehicle for Wall-Following

Fuzzy logic inference system (FIS) has been widely applied to the controller design for automatic guided vehicles (AGV) because FIS allows easier controller design under uncertainty and nonlinearity (Hwang et al., 2007; Godjevac & Steele, 1999; Baturone et al., 2008; Er &Deng, 2004; Ng & Trivedi, 1998). Wall following is a commonly adopted scheme for an AGV to navigate in the indoor or outdoor environments. Sonar system is usually the most popular hardware system installed on the AGV for wall following due to its costeffective functionality and computational efficiency. There has been some research applying FIS to the sonar-based wall following task (Tsui et al., 2008; Li et al., 2003; Juang & Hsu, 2009). The sonar constantly transmits ultra-sound signals during the wall following process. The ultra-sound signals cannot go through most of the objects, walls or structures in the environment, and thus are reflected back to the sonar. By calculating the difference between the time when the ultra-sound signals are transmitted and are received, the AGV is able to constantly detect the distance between the AGV and the object the ultra-sound signals are reflected from. For the wall following, the AGV is controlled to navigate along the wall while maintaining a fixed distance based on the received ultra-sound signals. If the surface or texture of the wall varies as AGV navigates in the environment, the ultra-sound signals might not be directly reflected back to the receiver or the intensity of received signals might not be constant all the time. The time difference of the transmitted and received ultra-sound signals is determined by calculating the time when the transmitted signal is above a threshold and the time when the received signal is above another threshold. The deflection of the ultra-sound signals due to the variation of object surface and the reduction of reflected signals due to the surface texture and material characteristic will cause the uncertainty of distance detection based on reflected ultra-sound signals. In other words, the calculated distance is corrupted by inevitable noise and disturbance contained in the received ultra-sound signals. Although fuzzy controllers are credited with a high degree of reliability for controlling such a complicated system as AGV, the type-1 fuzzy controller sometimes is not robust enough to cope with the uncertainty existed in the noise-corrupted sonar signals. In this paper, a type-2 fuzzy controller (Mendel, 2001; Mendel & John, 2002) is proposed to control both the left and right drive wheel of an nonholonomic AGV for the wall following. It will be shown in this paper that the proposed type-2 fuzzy controller resolves the inevitable noise problem due to its flexibility of processing controller’s input and output signals with uncertainty and