Two-stage adaptive robot position/force control using fuzzy reasoning and neural networks

Abstract – Many studies have been performed on the position/ force control of robot manipulators. Since the desired position and force required to realize certain tasks are usually designated in the operational space, the controller should adapt itself to an environment and generate the control force vector in the operational space. On the other hand, the friction of each joint of a robot manipulator is a serious problem since it impedes control accuracy. Therefore, the friction should be effectively compensated for in order to realize precise control of robot manipulators. Recently, soft computing techniques (fuzzy reasoning, neural networks and genetic algorithms) have been playing an important role in the control of robots. Applying the fuzzy-neuro approach (a combination of fuzzy reasoning and neural networks), learning/ adaptation ability and human knowledge can be incorporated into a robot controller. In this paper, we propose a two-stage adaptive robot manipulator position/ force control method in which the uncertain/ unknown dynamic of the environment is compensated for in the task space and the joint friction is effectively compensated for in the joint space using soft computing techniques. The effectiveness of the proposed control method was evaluated by experiments.

Keywords: robot manipulator; position / force control; soft computing; friction compensation.

INTRODUCTION

In order to perform sophisticated tasks with robot manipulators, the force as well as the position has to be precisely controlled to interact with the environment. Since the desired force and position required to perform certain tasks are usually designated in the task space, the control force vector should be given to the end-effector in the task space. A hybrid position / force control method is one of the most effective and fundamental control methods to realize accurate position / force control of robot manipulators in the task space. In the case of when the dynamic properties of the environment are unknown or uncertain, the robot controller has to adapt itself to the unknown dynamics of the environment since it affects the dynamics of the whole system. In order to cope with this problem, fuzzy reasoning and neural networks have been applied for position / force control to realize the effective adaptive controller. Applying fuzzy reasoning and neural networks, learning / adaptation ability and human knowledge can be incorporated into a controller. It is known that adaptive neuro control and adaptive fuzzy-neuro control are especially effective for position / force control with uncertain / unknown parameters of the robot manipulator and the environment.

If the friction of each joint of a manipulator is not sufficiently compensated for at the beginning, the adaptation time of the controller becomes longer. Therefore, accurate control cannot be expected until the controller adapts itself to the uncertainties. Most friction compensation techniques are model-based feed-forward compensation methods. It is difficult, however, to produce a precise friction model because of the complexity of the static and dynamic characteristics of friction such as the Stribeck effect, the Dahl effect, stick-slip motion, etc. In particular, in the case of the control of multi-link robot manipulators, the effect of friction is more complicated since the configuration change of a robot manipulator results in the change of the joint friction due to the change of the moment of the gravity force acting on each joint. Furthermore, the amount of external force acting on the end-effector affects the amount of joint friction. Consequently, precise friction models are expected to accurately compensate for joint friction.

In this paper, an effective position / force control method, in which on-line joint friction compensation is performed in the joint space and adaptation to uncertain / unknown dynamics of the robot manipulator and the environment is simultaneously performed in the task space using fuzzy reasoning and neural networks, is proposed based on the resolved acceleration control method. In this method, the position and force of the end-effector is controlled in the task space by the hybrid position / force controller and the desired joint torque is realized with adaptive friction models assuming that most of the joint torque error is caused by joint friction. Applying the two-stage approach of joint level adaptation and the task space level adaptation, the desired position and force of the end-effector is realized precisely and efficiently, even if parameters of the robot manipulator and the environment are uncertain or unknown. In this study, two kinds of adaptive friction models (fuzzy friction model and neural network friction, model) are introduced and evaluated.