Adaptive methodologies and robotics

In this paper the two following meanings of "Adaptive Control" are precised: the automatic modification of the transfert function of actuators in terms of the moving robot; the automatic correction of the trajectory by a changing robot environment. The main results of the conventional control technics (neglecting the coupling terms between joints and using conventional PID controlers) as well as the results of the dynamic control (supposing an a priori knowledge of the inertial properties of the manipulated object) are then described. The limits of these two approaches are given. The aim of the adaptive control consists in high-level performances linked with large variations of the process dynamic characteristics adaptation ability without any previous knowledge of the dynamic behavior. The three main approaches are emphasized : 1) the Model Reference Adaptive Control (MRAC) compares the value of the performance index with the measured value; 2) the self-Tuning Adaptive Control (STAC) modifies the control law parameters after an on-line identification of the controlled process; 3) the Identification Observation Adaptive Control (IOAC): in this approach the adaptive mechanism entailing modifications to the parameters of the adjustable system is also composed of an identificator working out the Reference Model (that is the observer by itself whose parameters are calculated at the end of the identification stage). The State of the Art is conveyed in the application to the manipulator robot control of the three previous approaches. These technics are developed either in the generalized or operational space. The main results which have been obtained from the algorithms developed in the 1), 2), 3) approaches described again. The current studies in the 3) approach are indicated. The main control block diagrams worked out by the research workers are summarized in a table under the form of a decisional tree, At last the industrial future of the Adaptive Control in Robotics - s discussed from the kinematics point of view for the articulated mechanical system and as far as the complexity of the calculations and performances are concerned for intelligent robots.