DUAL-BLADE WAFER HANDLING ROBOT 

Context in source publication

Context 1

... dual-blade wafer handling robot is an industrial robot designed for silicon wafer manufacturing. Unlike most industrial robots that have open kinematic chains, the dual- blade wafer handling robot has closed kinematic chains. The robot is not equipped with an actuator at every joint, which is quite different from other industrial robots. In the robotics literature, these robots are referred as "parallel robots" [ 1- 2] . Typically, parallel robots offer greater efficiency and faster acceleration at the end-effecter, and therefore are more suitable for fast assembly lines. However, because of the complexity of related kinematics and dynamics analyses, the derivation of dynamic equations of motion for parallel robots is more involved. Formulation of such equations that are suitable for controller design is still an open question. One popular method [3- 4] is to first separate the closed chain into several open chains, and then derive the equations of motion. This method will lead to a set of differential algebraic equations that are not easy to solve. Uicker [5] has tried to formulate the equations of motion in terms of all dependent generalized coordinates, and then eliminated all holonomic constrains to end up with differential equations with independent generalized coordinates. Getting explicit motion equations is generally impossible because of the complexity of parallel robot kinematics. It should be noted, however, that once explicit dynamic equations of motion can be obtained, it is possible to make the extension of existing robot control strategies to parallel robots [6]. In this paper, we exploit Lagrange's equation of motion to derive the dynamics model of a dual-blade wafer handling robot. An explicit form dynamic model is obtained. With the assumption of accurate lengths of links, we transform the dynamics model into a decoupled form to enable dynamic identification by least square regression [7]. Both the Lagrange's equations of motion and identification procedure are verified through numerical simulation. Friction is also molded to make the dynamic model more reliable. Identification experiment is performed on an actual robot prototype. The result has demonstrated the correctness of the dynamics model and the feasibility of the identification method. The dual-blade wafer handling robot adopts two frog-leg like five-bar mechanisms, as shown in Figure 1. During the silicon wafer manufacturing, the wafer is put on the forked end, which can be extended and rotated driven by joints 1 and 5 (two joints in the middle). The relationships between the dependent and independent generalized coordinates are called loop equations. If loop equations can be solved explicitly, a dynamic model can be formulated in an explicit form. For most parallel robots, loop equations can only be solved by iteration process. Fortunately, five-bar mechanism is one of the few existing parallel robots that have analytical form solutions for loop equations. The kinematics should be studied first before dynamics [8- 9]. The five-bar mechanism adopted in the dual-blade robot is slightly different from a common one [10]. It is driven by the two rotational joints 1 and 5. The motion of joints 3 and 4 are bounded to ensure the robot has desired motion, i.e., the symmetric motion with respect to the extensional axis (without this constraint, this mechanism will have one uncontrollable degree of freedom). This robot can only perform one rotational motion and one extension motion as shown in Figure 2. Analysis of kinematics would be easier if we consider the two type of motion separately. The extension motion is considered firstly, and then the rotational motion. The kinematics is as shown in Figure 2. 1 and 5 are chosen as independent generalized coordinates. From the simple geometrical relationship, we can get the solution of loop ...