IEEE JOURNAL OF OCEANIC ENGINEERING, VOL. 25, NO. 3, JULY 2000 399 Tracking Control for Underwater Vehicle-Manipulator Systems with Velocity Estimation Gianluca Antonelli, Fabrizio Caccavale, Stefano Chiaverini, and Luigi Villani Abstract—In this paper, the problem of tracking a desired motion trajectory for an underwater vehicle-manipulator system without using direct velocity feedback is addressed. For this purpose, an observer is adopted to provide estimation of the system’s velocity needed by a tracking control law. The com- bined controller-observer scheme is designed so as to achieve exponential convergence to zero of both motion tracking and estimation errors. In order to avoid representation singularities of the orientation, unit quaternions are used to express the vehicle attitude. Implementation issues are also considered and simplified control laws are suggested, aimed at suitably trading off tracking performance against reduced computational load. Simulation case studies are carried out to show the effectiveness of the proposed controller-observer algorithm. The obtained performance is com- pared to that achieved with a control scheme in which the velocity is reconstructed via numerical differentiation of position mea- surements. The results confirm that the chattering on the control commands is significantly reduced when the controller-observer strategy is adopted in lieu of raw numerical differentiation; this leads to lower energy consumption at the actuators and increases their lifetime. Index Terms—Manipulators, motion control, observers, track- ing, underwater vehicles. I. INTRODUCTION U NDERWATER tasks involving an autonomous vehicle equipped with a manipulator give rise to challenging control problems involving nonlinear, coupled, and high-di- mensional systems. As typical in robotics, the execution of such tasks can be formulated in terms of a control problem regarding the manipulator’s end-effector motion, for which several techniques have been proposed. In recent years, advanced control techniques have been developed for autonomous underwater vehicles (AUV’s) and remotely operated vehicles (ROV’s), aimed at improving the capability of tracking desired position and attitude trajecto- ries. Improvement of tracking performance typically requires control schemes based on the knowledge of the system’s dynamics (i.e., inverse dynamics control laws, feedforward compensation), e.g., as in [1], [2], or based on adaptive actions [3]. However, as is typical in feedback control systems, the achievable performance is highly dependent on the accuracy of sensor measurements. Manuscript received November 17, 1998; revised August 18, 1999. G. Antonelli and S. Chiaverini are with the Dipartimento di Automazione, Elettromagnetismo, Ingegneria dell’Informazione e Matematica Industriale, Università degli Studi di Cassino, 03043 Cassino, Italy. F. Caccavale and L. Villani are with the Dipartimento di Informatica e Sis- temistica, Università degli Studi di Napoli Federico II, 80125 Napoli, Italy. Publisher Item Identifier S 0364-9059(00)06688-7. Subsea vehicles are typically equipped with acoustic sensors or video systems for position measurements, while the vehicle attitude can be obtained from gyroscopic sensors and/or com- passes. Velocity measurements are usually obtained from sen- sors based on the Doppler effect. Consider the case of an underwater vehicle-manipulator system (UVMS) involved in a manipulation task to be accom- plished with high accuracy, i.e., the end-effector reference trajectories have to be tracked with small errors. Examples of such tasks are maintenance of off-shore structures and the recovery of materials on the sea bottom. Under these operating conditions, the vehicle’s velocity is usually lower than typical cruising velocities. In order to achieve good tracking during the fulfilment of the task, accurate position and orientation measurements must be available at a relatively high update rate: this can be achieved, e.g., if the system is equipped with a vision-based sensing device. Moreover, the availability of accurate and noise-free velocity measurements is crucial for achieving the desired performance. Unfortunately, this is not guaranteed by Doppler effect sensors and sonars, especially during slow maneuvers. As a matter of fact, numerical differ- entiation of noisy position/orientation measurements leads to chattering of the control inputs, which may become unaccept- able when quantization effects are present. Such phenomena lead to high-energy consumption at the actuators, and thus tend to reduce their lifetime and increase the failures rate. On the other hand, the use of low-pass filters on the numerically reconstructed velocities could deteriorate the overall tracking performance; in addition, since the system to be controlled is nonlinear, tuning of the filter parameters guaranteeing closed-loop stability is not straightforward and can be achieved only by a trial-and-error procedure. Hence, it is worth devising algorithms for position and atti- tude control which do not require direct velocity feedback. This can be achieved by adopting a velocity observer which performs a nonlinear filtering of the measures from the position/orienta- tion sensors and gives a noise-free estimation of the velocities. Clearly, the controller-observer structure must be designed so as to ensure stability of the resulting closed-loop system and ade- quate tracking performance. A nonlinear observer for vehicle velocity and acceleration has been proposed in [4], [5], although a combined controller-ob- server design procedure has not been developed. On the other hand, a passivity-based control law is proposed in [6], where the velocities are reconstructed via a lead filter; however, this control scheme achieves only regulation of position and orien- tation variables for an underwater vehicle-manipulator system. In this paper, the problem of output feedback tracking control of subsea vehicle-manipulator systems is addressed. The output 0364–9059/00$10.00 © 2000 IEEE