Circle 6 groups the computations required to check initial and terminal conditions (e.g. contact established and destination reached, respectively, for RUB). Checks, however, are performed in circle 1. Circle 9 computes the rotation matrix and the Jacobian of the compliant frame with respect to the base frame (J(q)). Circle 8 executes the force-control algorithm to obtain u, yF and ~F. Info on current action_command is the command identifier (e.g. RUB). A decomposition of the functions of circle 8 is shown in Fig. 6, whose clear rationale is given by the force algorithm scheme of Fig. 4.
Thanks to this modular organization, changing (for instance) the force-control algorithm is reflected in changes to just one or few circle functions, i.e. a small part with sharp boundaries. The same is true for changing the force sensor, and so on. Impedance control and on-line path-planning algorithms, needed to implement actions like
INSERT and FOLLOW, may be implemented using the same circles, introducing the proper functions in each circle. The overall controller AA therefore has the same structure, even if the fimctions of each circle are actually the logic sum of the functions relevant to each action.
5. DESIGN OF THE IMPLEMENTATION
ARCHITECTURE
As the final part of the case study, this section shows how the fimctional architecture proposed above may be implemented by the extension of a commercial controller. The specific case of the Comau C3G 9000 controller is considered. The new controller is named here C4G. The hardware modifications will be considered first, and then a software architecture will be proposed. The illustrative example will then focus on the AA relevant to the RUB action. The proposed solution is, however, also suitable for implementing the other actions defined in Section 3, provided that the relevant additional software modules are developed.
In the choice of a solution, economic constraints are of major concern when mass-produced systems like industrial controllers are considered. Moreover, taking into account that applications requiring sensor-based control will likely remain in a minority for many years, a modular solution featuring exteroceptive sensor-based controls as an add-on item to the standard controller seems the most interesting, and perhaps the only viable, one.