You are on page 1of 19

Ocean Engineering 30 (2003) 21372155 www.elsevier.

com/locate/oceaneng

Design of a sliding mode fuzzy controller for the guidance and control of an autonomous underwater vehicle
J. Guo , F.-C. Chiu, C.-C. Huang
Department of Naval Architecture and Ocean Engineering, National Taiwan University, 73 Chou-Shan Road, Taipei, Taiwan, ROC Received 21 August 2002; accepted 27 November 2002

Abstract This work demonstrates the feasibility of applying a sliding mode fuzzy controller to motion control and line of sight guidance of an autonomous underwater vehicle. The design method of the sliding mode fuzzy controller offers a systematical means of constructing a set of shrinking-span and dilating-span membership functions for the controller. Stability and robustness of the control system are guaranteed by properly selecting the shrinking and dilating factors of the fuzzy membership functions. Control parameters selected for a testbed vehicle, AUVHM1, are evaluated through tank and eld experiments. Experimental results indicate the effectiveness of the proposed controller in dealing with model uncertainties, non-linearities of the vehicle dynamics, and environmental disturbances caused by ocean currents and waves. 2003 Elsevier Ltd. All rights reserved.
Keywords: AUV; Sliding mode control; Fuzzy control; Guidance

1. Introduction Control problems involving autonomous underwater vehicles (AUVs) present several difculties, owing to their non-linear dynamics, the presence of disturbance, and observation noises. Ocean exploration and the utilization of oceanic resources in shallow, conned water areas have received increasing interest in recent years. In

Corresponding author. Fax: +886-2-23929885. E-mail address: jguo@ccms.ntu.edu.tw (J. Guo).

0029-8018/$ - see front matter 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0029-8018(03)00048-9

2138

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

such regions, shallow water phenomena from the interaction among wave dynamics, tidal currents, coastal currents, and articial objects create a complex environment for operating unmanned underwater vehicles. Therefore, controlling AUVs to satisfactorily track trajectories in shallow waters remains a challenge. Several control strategies have been developed for controlling the motion of underwater vehicles, among them are, supervisory control (Yoerger et al., 1986), neural network control (Yuh, 1990), self-turning control (Goheen and Jefferys, 1990), LQG/LTR (Triantafyllou and Grosenbaugh, 1991), sliding mode control (Yoerger and Slotine, 1985; Fossen and Sagatun, 1991; da Cunha et al., 1995; Christi et al., 1990; Healey and Lienard, 1993; Lam and Ura, 1996; Lee et al., 1999), fuzzy logic control (Kato et al., 1993; Smith et al., 1994), and recently, the sliding mode fuzzy logic control (Song and Smith, 2000). More references could be found from the underwater robotics community, for example (Yuh, 1994; Yuh et al., 1996). This paper presents a design method based on sliding mode fuzzy logic control. As well recognized, fuzzy logic controllers are effective robust controllers for various applications. A merit of using fuzzy logic to design a controller is that the dynamics of the controlled system need not be fully known. However, the linguistic expression of the fuzzy controller makes it difcult to guarantee the stability and robustness of the control system. Sliding mode control can also be applied effectively in the presence of model uncertainties, parameter variations, and disturbances. A boundary layer is generally used to avoid chattering on the sliding surface (Slotine and Li, 1991). Designing a fuzzy logic controller based on the sliding mode theory assures performance and stability, while simultaneously reducing the number of fuzzy rules. Furthermore, fuzzy partition of the manipulated variables avoids the chattering problem of the sliding mode control method (Palm, 1994; Palm et al., 1996). The sliding mode fuzzy logic controller (SMFLC) is thus adopted herein as the basic controller structure.

2. Background In Yoerger and Slotine (1985), sliding mode control methodology is applied to control trajectories of remotely operated vehicles (ROVs). It is demonstrated that a sliding mode controller can be designed using a simple nonlinear model of the vehicle. The controller effectively deals with nonlinearities, and is robust even with imprecise models. Furthermore, the trade-off between performance and model uncertainty is predictable. In Fossen and Sagatun (1991), a hybrid controller combining an adaptive scheme and a sliding-mode term was designed to control the motion of an ROV. Vehicle dynamics, such as inertia, hydrodynamic forces, and restoring forces, were proven able to be estimated on-line, while the thruster nonlinearities and unknown thruster dynamics can be compensated by the switching controller. Meanwhile, in da Cunha et al. (1995), an adaptive control scheme for dynamic positioning of an ROV was developed based on a sliding mode control algorithm that only used position measurements. The applications of the sliding mode can be extended to the motion control of AUVs. Furthermore, the problem of controlling

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

2139

an AUV in the dive plane through the sliding mode based on observed states was addressed in Christi et al. (1990). Additionally, parameter variations of the motion equation resulting from speed changes are estimated on-line, while unmodeled dynamics are compensated by the sliding mode. In Healey and Lienard (1993), multivariable sliding mode design of an AUV autopilot system for the maneuvering of combined steering, diving, and speed was presented. Meanwhile, the simulation of a line-of-sight scheme for guiding the AUV between waypoints in ocean currents was also considered. Moreover, a switched control law that allowed a non-cruising type AUV to hover precisely was presented in Lam and Ura (1996). The controller uses a dead zone as well as a switching law to avoid the effects of measurement noises and the non-linearity of the thruster. In Lee et al. (1999), a discrete-time quasi-sliding mode controller is used for controlling the depth of an AUV, and the control algorithm is conrmed to be effective with model uncertainties and large sampling periods. In Kato et al. (1993), the fuzzy algorithm was applied to manage the guidance and control of a cable tracking AUV in both attitude control and cable tracking. The optimization of fuzzy rules based on the dynamic model of the AUV was also investigated. In Smith et al. (1994), fuzzy logic controllers were proposed for controlling and guiding a low-speed torpedo shaped vehicle. Heading, pitch, and depth are controlled simultaneously via three single-axis fuzzy logic controllers. A docking law based fuzzy logic was also assessed using a nonlinear simulation model. Recently, a time optimal design method base on SMFLC was presented in Song and Smith (2000). The shape of the AUV open loop step response was used as the switching surface for the SMFLC controller. It is argued that the design method provides time optimality and robustness, and it does not require system model of the AUV. This work presents an efcient controller for controlling the direction of a at, streamlined AUV, AUV-Hai-Min. The designated mission of the AUV-HM1 is area search and survey in shallow water. Typical search/survey scenarios covering large areas require route stability and good turning performance in the horizontal plane of motion. Directional control is thus fundamental to the system presented herein. The proposed design method requires designation of a shrinking or dilating factor for fuzzy input/output membership functions. We select shrinking and dilating factors based on specications on control precision and the robustness to external disturbance and system modeling error. This procedure is derived using the formulation of the sliding mode control described in Slotine and Li (1991). Compared to conventional fuzzy logic controllers and SMFLCs, this method has less number of control parameters, and is therefore much easier to implement. The paper rst describes the guidance law for the AUV-HM1. Then the design method based on fuzzy sliding mode is shown, and guidelines for selecting control parameters are illustrated. Guidance and control experiments in a water tank and in shallow sea were conducted to demonstrate the effectiveness of the method.

2140

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

3. Guidance law A one-degree-of-freedom vehicle model is used herein to describe the horizontal turning behavior of the AUV. The model includes drag, added mass, and thrust moment for yaw motion, Ir br r u d (1)

where I denotes the vehicles mass moment of inertia plus the added inertia of the body about the body-xed z-axis, r represents the body-xed rate for heading direction, b denotes the square-law damping coefcient, u is the moment generated by commanding differential thrust force on the left and right thrusters, and d represents the disturbance caused by ocean currents, modeling errors, and unmodeled dynamics. The line-of-sight guidance procedure is illustrated in Figs. 1 and 2. The guidance law comprises of the following components. 3.1. Inputs The heading error, ye(t), and rate of heading error, re(t), where ye(t) yi(t) yd(t) (2) and yi(t) is the vehicle heading, while yd(t) denotes the desired vehicle heading. yd(t) tan
1

yi(t) yd(t) ,0 yd(t) xi(t) xd(t)

2p

(3)

where (xi(t),yi(t)), (xd(t),yd(t)) represent the vehicles position and desired position, respectively.

Fig. 1.

Signal ow diagram of the guidance law.

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

2141

Fig. 2.

Denition of parameters in the guidance law.

3.2. Error and control signals The sliding error, SE, control command, u SE(t) e (1 l)y(t) ly(t),0 e l 1 (4) where the superscript represents the normalized variables. The normalization factors,Gy, and Guare used to map the signals on to a dened domain. SE SE ;y (t) Gy e ye(t) ;r (t) Gy e e y re(t) ;u (t) Gy u(t) Gu (5)

3.3. The fuzzy rule-base Rule i in the rule base is written as If SE Fsi then u Fu i, and msi(SE) mu i(u) (6) Where i = m,%, 1, 0, 1%,m, and Fsi and Fui are the ith linguistic members of the input and output variables, respectively. Here, we use 2m + 1 linguistic members to partition input and output variables. Both positive/negative sides have the same

2142

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

number of linguistic members. Triangular membership functions are used to represent linguistic members. The degree-of-membership function, mXi(X) is calculated by X Xi Xi Xi mXi(X)
1 1

for Xi

X Xi (7)

Xi+1 X for Xi X Xi+1 Xi+1 Xi 0 others

where X represents the sliding error, SE, or the control command, u. For example, if ve sections are used to partition the domain of input and output variables, the following rule-base can be obtained, If SE is PB, then u is NB. If SE is PM, then u is NM. If SE is ZO, then u is ZO. If SE is NM, then u is PM. If SE is NB, then u is PB. where PB represents positive big, PM denotes positive medium, ZO represents zero, NM is negative medium, and NB denotes negative big, respectively. 3.4. Arrangement of membership functions The core (or the highest) values Si and Ui for the ith membership functions are designed such that Xi i m s mf
|i|

(8)

% X 1 X0 X1 % Xms = 1, and where Xi = Si or Ui, 1 = X ms mXi(X) = 1. Here, sf (0,1] is the shrinking factor (Chen and Hsieh, 1996). This set up results in a series of shrinking-span membership functions. The same idea can be used to form a series of dilating-span membership functions by assigning core values as Xi i msgn(i) m df m
| i msgn(i) |

sgn(i)

(9)

where df ( 0,1] is termed the dilating factor. Fig. 3 illustrates examples of the shrinking and dilating-span membership functions. Note that when the shrinking factor sf = 1 , or the dilating factor df = 1, the membership functions are equal-span isosceles triangular membership functions.

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

2143

Fig. 3.

Examples of shriking-span and dilating-span membership functions.

This arrangement brings out the fact that there exist at most two rules being red at the same time. Furthermore, the two membership values have a sum of 1, mXi(X) mXi+1(X) 1 ,X [Xi,Xi+1] (10)

3.5. Defuzzication For a red rule, and the corresponding linguistic member Fui, dene a b c d h Ui
1

mui(u)(Ui Ui 1)

Ui

Ui+1 mui(u )(Ui+1 Ui) Ui+1 mui(u)

(11)

Since there are two linguistic members being red at the same time, Fig. 4 illus-

Fig. 4. Shaded regions represent areas covered by membership functions for defuzzication.

2144

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

trates the area obtained after the fuzzy inference. The commonly used center of area method (Palm et al., 1996) is employed herein to obtain the crisp value of the control signal u. Aui Cui c a 2 (b a)2 3 2(b a)(c b) c (c b)2 d a b (c a)(d c) 1 (d c)2 3 d a b h 2 (12) (13)

CuiAui A ui

(14)

where Aui, Cui are the centroid and area covered by the ith membership function of the output linguistic variable.

4. Controller design A sliding mode formulation is used to design controller parameters. The control law u(t) is designed so that the system trajectories are ultimately bounded under the region B = {SE, SE } ; 0. This goal can be achieved by satisfying the following sliding condition outside region B (Slotine and Li, 1991), SESE h(|SE| ); h 0 (15)

where the sliding error SE is dened in Eq. (5). The constant h is a design parameter that is inversely proportional to the time required for the SE to reach the boundary of region B from outside of B. The time derivative of the sliding error is determined from Eq. (4), in which the acceleration of yawing angle can be obtained from Eq. (1), SE l 1 l u I yd y I 1 l e br r d (16)

Dene r = yd SE

l y , Eq. (16) can be expressed as 1 l e d (I r br|r| )] 1 l [u I f(r,r, d)] (17)

1 l [u I

where f(r,r, d) = d (I r + br r ). If no sufcient model of r and d exists, the estimate of the following upper bound must be determined

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

2145

| f(r,r,d)|max

(18)

We re-express the sliding condition as SE or SE 1 l u I h( SE ) SE 1 l F I h 1 l F SE I h (19) The fuzzy rule base is designed so that the fuzzy controller has a similar form to the sliding mode controller with a boundary layer, u where sat( SE ) SE / if SE K(SE,l)sat( SE ) (20) 1 l [u I f(r,r,d)] SE 1 l u I |SE| 1 l F I h (|SE| )

sgn(SE / ) if SE

and K(SE,l) 0. When SE , the sliding condition holds, replacing u in Eq. (19) with the u in Eq. (20) obtains a stability and robustness bound for the magnitude of K(SE,l), K(SE,l) I / Gu 1 l h 1 l F I h / Gy SE (21)

where K(SE,l) = K(SE,l) / Gu. The method of determining the stability and robustness bound offers simplicity in selecting the input/output membership functions, and is similar to the technique presented in (Emami et al., 2000). When SE , the transfer function from SE to u can be tuned by designing the shape of the input and output membership functions. According to the design principle of sliding mode controllers, the balance condition (Slotine and Li, 1991) should be fullled inside the boundary layer to avoid the inuence of high-frequency unmodeled dynamics. Substituting Eq. (20) into Eq. (17) obtains the dynamics of SE inside the boundary layer ( 1 l K(SE,l) )( )SE I SE ( 1 l )f(r,r,d) I (22)

Letting the corner frequency of Eq. (22) be less than or equal to l / 1 l, the design criterion for the slope in the vicinity of the origin of the SE u phase space is obtained. Dene the K(0,l) to be the magnitude of K(SE,l) near the origin. K(0,l) shall satisfy the following inequality,

2146

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

K(0,l)

I 1 l 1

l l

Gy Gu

(23)

The slope of the transfer function at the origin can be adjusted by careful selection of shrinking and dilating factors. For example, Fig. 5 illustrates the transfer function of a normalized fuzzy logic controller with its inputoutput membership functions. Approximating the shape of the transfer function with piece-wise linear functions, discontinuous points of the piece-wise linear functions appear at core values of the input triangular membership functions (Palm, 1994). A reasonable selection of the parameter will be at the rst core values (i = 1 and 1, in Eq. (8) and Eq. (9)) of the input membership function. For example, if an input shrinking factor 0.5 is used, then = U1 = 0.25Gy, or = 0.25. Fig. 6 illustrates constant-slope contours at the origin of the SE u phase plane in terms of input/output shrinking and dilating factors. A plot such as Fig. 6 is useful for the selection of input/output shrinking and dilating factors. Five sections are used in Fig. 6 to partition the domain of input and output variables. From Eq. (21), the parameter Gu can be determined by letting SE , u 1, Gu I h 1 l 1 l F I (24)

To design the fuzzy controller, parameter values of h,Gu,Gy, and l are needed. Parameter h is inversely proportional to the time required to reach region B from an outside initial condition. Parameters Gu and Gy inuence the slope of the controllers transfer function, thus determining the behavior of the system around the origin. The value of l / 1 l species the rate of convergence on the sliding surface. In Slotine and Li (1991), a suggested choice for l / 1 l is one fth of the sampling frequency.

Fig. 5. The normalized input-output relationship of the fuzzy logic controller with its input/output membership functions.

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

2147

Fig. 6. Constant-slope contours at the origin of the SE u phase space in terms of input-output shrinking and dilating factors. The numerical numbers on the curves indicates the magnitude of the slope.

5. Experiments Tank tests and open sea trials were conducted to determine the effects of the design parameters. The testing tank was 120 84 m (L WD). Meanwhile, the testbed vehicle AUV-HM1 had dimensions of 2 10.6 m (L WD), as displayed in Fig. 7. The vehicle employed a 300 kHz Doppler velocity log and a ber-optics gyro as navigation sensors. The sea trial area had a water depth of 60 to 80 m. To test the effect of surface currents and waves on the effectiveness of the control algorithm, the vehicle was submerged at a depth of 2 m below the surface. To design the controller, the balance condition shall be satised at the origin of the SE u phase plane. One can select proper shrinking/dilating factors to fulll the balance condition. Gy is designed according to the tracking precision required. The design parameter h is determined to set the time required to reach the boundary

Fig. 7.

The testbed vehicle AUV-HM1.

2148

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

layer from an outside of the boundary layer initial condition. Gu sets the maximum output range of thrust forces. Gu is determined from Eq. (24), in which the uncertainty bound F represents the maximum uncertainty the system is able to tolerate. Finally, the stability and robustness property of the control system shall be checked using Eq. (21). 5.1. Effects of Gy, Gu, and l A directional command is given while the vehicle is moving with a forward velocity of 0.8 m/s in the water tank. Figs. 810 present the effects of Gy, Gu, l on transient responses and tracking precision. Gu is a scale factor to determine the magnitude of the thrust force. While advancing at higher speeds, the vehicle encounters larger disturbances and model uncertainties, higher Gu is needed to maintain satisfactory control performance. Higher Gu lead to faster transient responses and more precise tracking. Gy is proportional to the thickness of the boundary layer. Higher Gy leads to less tracking precision. Higher l lead to faster transient responses, and less steady state tracking errors. The parameter h is kept constant in these tests. Figs. 810 show different initial conditions. The time required to reach region B from an outside initial condition is inversely proportional to the value of h and the initial distance from region B. 5.2. The effect of xo During sea trials, the vehicle was ordered to follow a series of waypoint circles of origin [xkd,ykd], and radius xo. If the present vehicle position [xc,yc] satised the following condition

Fig. 8.

Transient responses of vehicle heading with different Gys.

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

2149

Fig. 9.

Transient responses of vehicle heading with different Gus.

Fig. 10.

Transient responses of vehicle heading with different ls.

[xc xkd]2

[yc ykd]2

x2 o

(25)

the guidance algorithm triggered the selection of the next waypoint. Fig. 11 presents the results of tracking experiments in ocean currents. Waypoints are arranged on a straight line. The controller is selected to be the same for all three experiments. We examine the tracking precision using three different waypoint circles radius xo under

2150

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

Fig. 11.

Waypoint tracking in ocean currents.

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

2151

the inuence of ocean currents and waves. Since precise tracking is difcult to achieve given model uncertainties and external disturbances, the radius of the waypoint circle must not be less than the achievable tracking precision. Achieving precise tracking obviously comes at the cost of poor transient performance. Consequently, a tradeoff between stability and tracking precision must be found. 5.3. The effect of the control gain Fig. 12 illustrates the state dependent control gains used in the sea trials. The upper plot of Fig. 12 represents a controller that satises the balance condition, while

Fig. 12. Controller gain of the sliding mode fuzzy controllers and the stability and robustness bounds for the sea experiments presented in Fig. 13.

2152

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

Fig. 13. Square path tracking in ocean currents.

the lower plot does not. Both controllers satisfy the stability and robustness condition. Fig. 13 presents waypoint-tracking results using the above controllers. Waypoints are arranged at the corner of a square route. The sampling frequency is 12.5 rad/s, and l is set at 0.7. The upper bound of the slope for the control gain at the origin is calculated using Eq. (23), yielding a value of 5.6. Two experimental results were chosen to demonstrate the effect of the control gain. The control gain at the origin corresponding to the upper plot of Fig. 12 is designed to be 5.5, while that corresponding to the lower plot of Fig. 12 is 1. Notably, while the AUV moved from one side of the square path to the other, the uncertainty bound F varied signicantly. At the moment of transition, the state and the sliding surface are separated by a signicant distance, and the unmodeled dynamics cannot change the sign of the control input. Consequently, we evaluated this bound only in the vicinity of the

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

2153

Fig. 14. The uncertainty f, and its bound F, calculated from the experimental data shown in Fig. 13.

origin of the sliding error. The magnitude of uncertainty bound F is inuenced by the waypoint selection, magnitude and direction of ocean currents, Vc, Dc, and the vehicles unmodeled dynamics. To check the stability and robustness of the control system, we need to measure F. The Fs corresponding to these control gains can be calculated using experimental data. The following parameter values are used: I = 24.13 Kgfmsec2, b = 32.50 Kgfmsec2, d max=1.02Kgfm. The moment of inertia, I, including the added moment of inertia, was measured using a Planar Motion Mechanism (Chiu et al., 1997), the damping coefcient b was derived from yaw step

2154

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

responses data, and the cross ow moment d at current velocity 0.5 m/s was measured by towing tank tests. Fig. 14 shows that F = 4.5Kgfm, which was calculated from all path segments, ignoring large variations in magnitude at transition periods. Based on Eq. (24), the value of Gu is set at 9.5 Kgfm, which, after a unit conversion from torque to command voltage, is 5 V.

6. Conclusions This study has demonstrated the feasibility of applying a sliding mode fuzzy controller to an AUV in shallow water in order to perform line-of-sight guidance in the horizontal plane. The state dependent control gain is specied by a set of shrinkingspan and dilating-span factors. Design parameters of this controller, such as the slope at the origin, and the stability and robustness bound, are specied using dened criteria. The effect of selecting different control parameters is evaluated through water tank and open sea experiments using an AUV testbed. The test results also conrm the effectiveness of the method.

Acknowledgements The authors would like to thank the National Science Council of ROC for nancially supporting this research under Contract No. NSC88-2611-E002-020.

References
Chen, C.-L., Hsieh, C.-T., 1996. User-friendly design method for fuzzy logic controller. IEE Proceedings Control Theory and Applications 143 (4), 358366. Chiu, F.C., Guo, J., Huang, C.C., Wang, J.P., 1997. On the linear hydrodynamic forces and the maneuverability of an unmanned untethered submersible with streamlined body. Journal of the Japanese Society of Naval Architecture 182, 151159 2nd report: lateral motion, in Japanese. Christi, R., Papoulias, F.A., Healey, A.J., 1990. Adaptive sliding mode control of autonomous underwater vehicles in the dive plane. IEEE Journal of Oceanic Engineering 15 (3), 152160. da Cunha, J.P.V.S., Costa, R.R., Hsu, L., 1995. Design of a high performance variable structure control of ROVs. IEEE Journal of Oceanic Engineering 20 (1), 4255. Emami, M.R., Goldenberg, A.A., Tueken, I.B., 2000. Systematic design and analysis of fuzzy-logic control and application to robotics, Part II. Control. Robotics and Autonomous Systems 33, 89108. Fossen, T.I., Sagatun, S., 1991. Adaptive control of nonlinear systems: A case study of underwater robotic systems. Journal of Robotic Systems 8, 393412. Goheen, K.R., Jefferys, E.R., 1990. Multivariable self-turning autopilots for autonomous underwater vehicles. IEEE Journal of Oceanic Engineering 15 (3), 144151. Healey, A.J., Lienard, D., 1993. Multivariable sliding mode control for autonomous diving and steering of unmanned underwater vehicles. IEEE Journal of Oceanic Engineering 18 (3), 327339. Kato, N., Ito, Y., Kojima, J., Asakawa, K., Shirasaki, Y., 1993. Guidance and control of autonomous underwater vehicle AQUA EXPLORER 1000 for inspection of underwater cables. In: 8th Intl Symp on Unmanned Untethered Submersible Technology, pp. 195211.

J. Guo et al. / Ocean Engineering 30 (2003) 21372155

2155

Lam, W.-C., Ura, T., 1996. Non-linear controller with switched control law for tracking control of noncruising AUV. In: IEEE AUV96, Monterey, California, USA, Lee, P.-M., Hong, S.-W., Lim, Y.-K., Lee, C.-M., Jeon, B.-H., Park, J.-W., 1999. Discrete-time quasisliding mode control of an autonomous underwater vehicle. IEEE Journal of Oceanic Engineering 24 (3), 388395. Palm, R., 1994. Robust control by fuzzy sliding mode. Automatica 30 (9), 14291437. Palm, R., Driankov, D., Hellendoorn, H., 1996. Model Based Fuzzy Control: Fuzzy Gain Schedulers and Sliding Mode Fuzzy Control. Springer-Verlag, Berlin. Slotine, J.-J.E., Li, W., 1991. Applied Nonlinear Control. Prentice-Hall. Smith, S.M., Rae, G.J.S., Anderson, D.T., Shein, A.M., 1994. Fuzzy logic control of an autonomous underwater vehicle. Control Engineering Practice 2 (2), 321331. Song, F., Smith, S.M., 2000. Design of sliding mode fuzzy controllers for an autonomous underwater vehicle without system model. MTS/IEEE Oceans 2, 835840. Triantafyllou, M.S., Grosenbaugh, M.A., 1991. Robust control for underwater vehicle systems with time delays. IEEE Journal of Oceanic Engineering 16 (1), 146151. Yoerger, D.R., Slotine, J.-J.E., 1985. Robust trajectory control of underwater vehicles. IEEE Journal of Oceanic Engineering 10 (4), 462470. Yoerger, D.R., Newman, J.B., Slotine, J.-J.E., 1986. Supervisory control system for the JASON ROV. IEEE Journal of Oceanic Engineering 11 (3), 392399. Yuh, J., 1990. A neural net controller for underwater robotic vehicles. IEEE Journal of Oceanic Engineering 15 (3), 161166. Yuh, J. (Ed.), 1994. Underwater Robotic Vehicles, Design and Control. TSI Press. Yuh, J., Ura, T., Bekey, G.A. (Eds.), 1996. Underwater Robots. Kluwer Academic Publishers.