Characteristics of the Configuration Space Control

This mode of control has several distinct advantages (see also Results, Section  5):

• From the operator standpoint, the task is simplified greatly: instead of dealing with a complex jointed kinematic structure, the operator has to solve a simple maze-searching problem with complete information, which humans are very good at.
• One explanation for the task simplification is that the responsibilities are divided in this mode - the operator can think of the motion planning only, while the computer takes on the problem of collision analysis.
• The arm's actual motion is quickly and easily calculable from user input, guaranteeing good real-time performance.
• Unlike in W-space, performance here does not seem to depend much on the obstacle layout. Indeed, this mode has consistently yielded near optimal performances by the human operator in a variety of settings. This is consistent with the fact that humans can easily ``see'' the path in a bird-eye view of a fairly complex maze, while they have difficulty visualizing a path in a simple scene with an arm manipulator (see Figure  3). The operator easily discards many ``dead-end'' directions in the maze representation, but find it difficult to identify them in Figure  3.
• The mode requires very little training, mostly to get used to the peculiarities of flat presentation of two-torus - e.g. to the fact that once the point reaches the top edge of the C-space square, it appears at the bottom edge. In fact, performance has been just as good for an inexperienced user as for an experienced one.
• Unlike the W-space control, the subject can often easily see if a solution (a path) exists. In fact, it is this kind of decision-making that the operator uses extensively along the way to discard potential dead-ends.

A few drawbacks deserve to be noted of this mode, although their impact is not nearly as great as those in W-space control:

• The fact of dealing with an abstract (C-) rather than physical (W-) space may make it difficult for the operator to address some global navigation tasks, such as choosing targets for the arm to reach. This problem is easily avoided if the corresponding W-space view is drawn in parallel with the C-space used by the operator (see Figures  3 and  5).
• While extremely helpful in 2D, the mode is not likely to easily generalize to more complex multi-link systems see discussion in Section  5.2).
• Computation of C-space is an expensive operation which must be performed to satisfy the complete information model (see Section  5.2 for details on the proposed uncertainty model).