APPLICATION DESIGN ISSUES - Hard Real-Time Computing Systems: Predictable Scheduling Algorithms and Applications - page 412

OBJECT TACTILE EXPLORATION
When working in unknown environments, object exploration and recognition are es-
sential capabilities for carrying out autonomous operations. If vision does not provide
enough information or cannot be used because of insufficient light conditions, tactile
and force sensors can be effectively employed to extract local geometric features, such
as shape, contour, holes, edges, or protruding regions, from the explored objects.
Like the other tasks described above, tactile exploration requires the robot to conform
to a given geometry. More explicitly, the robot should be compliant in the direction
normal to the object surface, so that unexpected variations in the contour do not pro-
duce large changes in the force that the robot applies against the object. In the direc-
tions parallel to the surface, however, the robot needs to maintain a desired trajectory
and should therefore be position-controlled.
Strict time constraints for this task are necessary to guarantee robot stability during
exploration. For example, periods of servo loops can be derived as a function of
the robot speed, maximum applied forces, and rigidity coefficients, as shown in the
example described in Section 11.2.2. Other issues involved in robot tactile exploration
are discussed by Dario and Buttazzo [DB87] and by Bajcsy [Baj88].
CATCHING MOVING OBJECTS
Catching a moving object with one hand is one of the most difficult tasks for humans,
as well as for robot systems. In order to perform this task, several capabilities are re-
quired, such as smart sensing, visual tracking, motion prediction, trajectory planning,
and fine sensory-motor coordination. If the moving target is an intelligent being, like
a fast insect or a little mouse, the problem becomes more difficult to solve, since the
prey may unexpectedly modify its trajectory, velocity, and acceleration. In this situa-
tion, sensing, planning, and control must be performed in real time - that is, while the
target is moving - so that the trajectory of the arm can be modified in time to catch the
prey.
Strict time constraints for the tasks described above derive from the maximum velocity
and acceleration assumed for the moving object. An implementation of this task,
using a six degrees of freedom robot manipulator and a vision system, is described by
Buttazzo, Allotta, and Fanizza [BAF94].