Further details on this ARM can be found in (Putz and Mau, 1992), where it has also been shown that the most prominent control concepts can be well represented as certain instantiations of this general architecture. For given user requirements (actions to be executed for a specific control application), the CDM recommends that the functional requirements should be derived according to the ARM framework. The resulting functional architecture (logical model) of the control system is called the application architecture (AA). An example is given in Section 4. In the following sections, the design of an industrial controller is developed following the CDM principles and guidelines. 3. ACTIVITY ANALYSIS FOR INDUSTRIAL ROBOTS Activity analysis performs a functional analysis of the application process(es) in order to define what has to be done by the robot (including the controller) under design. According to the CDM, the result of the activity analysis is the list of the activities (tasks, actions) the robot should be able to execute. The activity analysis for industrial robots has been based on available teclmical and scientific documentation, and especially on the expertise of engineers from COMAU Robotics, the major Italian robot manufacturer and FMS integrator, who were extensively interviewed. Potential new applications have been carefully investigated with respect to their technical feasibility and commercial relevance for the immediate future. It has been found that it would be commercially interesting to improve tile programming and control capabilities of the next generation of robots to allow easier and more effective application to tasks like polishing and deburring, arc-welding, peg-into-hole and conveyor- tracking. Such applications, in fact, require capabilities to generate a trajectory on-line or to control the interaction (force) with the environment that are not available in current industrial controllers. Sometimes they are tackled in industry using compliance devices such as remote center compliance (RCC), or very accurate (but expensive and inflexible) positioning mechanisms of the working parts, or special sensorized flange-mounted actuators that compensate for positioning errors (e.g., for arc welding) or for high contact forces (e.g., for deburring). Robot system activities relevant to such applications have been analysed and decomposed into the actions of the CDM catalog (Putz and Mau, 1992). Works by Weule and Timmermann (1990) and Mizugaki, et al. (1990) for robotized polishing, by Her and Kazerooni (1991) and Bone, et al. (1991) for deburring, and by Whitney (1982) have been taken as references. It has been found that, besides basic actions like MOVE and APPROACH/RETRACT (to control gross and proximity motion) and ATTACH/DETACH (to grip/release the tool), new actions should be implemented, involving exteroceptive sensor-control capabilities. Among the most representative are RUB, required for polishing and deburring, FOLLOW, for arc welding and tracking, and INSERT for part mating. Templates have been drawn up for each action and for each application, thus obtaining a complete description of a robot system's activities. As an example, the template for the RUB action for a polishing application is shown in Table 1. 4. ANALYSIS OF CONTROL FUNCTIONS FOR HYBRID POSITION / FORCE CONTROL The second step of the CDM consists in the specification of the controller's functionality in the form of an AA, to be obtained by summing up the functions required for each action listed in the activity script. For each action it is necessary to select a planning and a control algorithm, and then to split them according to the general ARM for robotics. As an illustrative example, this section shows the derivation of the AA for the RUB action. First, general controller requirements and constraints influencing the selection of planning and control algorithms are discussed, and then the selected algorithms are described. Finally, the structure of the AA is shown.
4.1 General requirements and constraints Industrial robot controllers are currently designed as positioning devices: independent positioning servomechanisms are implemented at each joint. This solution is well established, and is based on motion control technology and on the relevant commercial products and experience. For this reason it is assumed here that the independent joint control scheme will also be adopted in the next generation of controllers In applications like polishing and deburring the manipulator may have to work on very stiff materials. The force control algorithm should be robust with respect to high values of environment stiffness, and possibly should not require knowledge of the stiffness parameters. On the other hand, it should take account of the high torsional flexibility that commonly affects the joints of industrial manipulators. 4.2 ,Selection of control algorithms for the RUB action The execution of the RUB action involves the modelling of contact, the planning of trajectories and contact forces, and their actuation through a proper hybrid position/force controller. Contact modelling and action planning. It is assumed that the contact between the working tool z c contact area Z c X C X C Fig. 3 Contact model and the working surface extends to a given plane contact area, and a compliant frame is chosen with its origin in the center point or at a convenient point of the contact area, with the z c axis directed along the nornml to the area. This is sketched in Fig. 3. If the working surface and the tool are rigid, virtual displacements of the origin of the compliant frame along the normal z c, and virtual rotations of the compliant frame about axes x C and Yc are inhibited, while reaction force along z c and torques about x c and Yc can be generated. To ensure the proper contact between the working tool and the workpiece it may be necessary to control both the force and the torques. The user plans the action by defining the contact frame on the end effector, assigning the path of the origin O c on the working surface, the orientation of axis z c (parallel to the surface normal) and of axes x c and Yc (conveniently), and the set points of the contact force and torques. An off-line programming environment might help the introduction of such values in cases of working surfaces of complex shapes. Hybrid position[force control algorithm In the scientific literature, several force-control algorithms have been proposed that comply with the presence of position control loops (see for instance (De Schutter, et aL, 1988; Duelen, et al., 1992)). 工业机器人的结构设计英文文献和中文翻译(3):http://www.youerw.com/fanyi/lunwen_34681.html