Human-machine Cooperative Manipulation with Vision-based Motion Constraints - Visual Servoing via Advanced Numerical Methods - page 64

(a) (b)

Fig. 4.3 Two applications of virtual fixture: (a) the InvenioSix high-precision robot in a con-
figuration to support fine lead soldering; and (b) the latest generation SHR in a configuration
designed for eye surgery

paradigm to guide dynamic motion on both admittance and impedance style robots
[27], and to better accommodate dynamics induced by the human user [25].

4.5 Conclusion

In this chapter, we have outlined a broad theory of compliant virtual fixtures,
and have applied that theory to the specific case of vision-guided assistance. Our
earlier work suggests that such virtual fixtures can be a useful aid to dexterous
manipulation.
In many ways, this chapter is intended to point toward interesting further direc-
tions to be explored. First and foremost, we have begun to develop a general means
for translating control algorithms into virtual fixtures. However, the treatment to
this point has not been sufficiently formal to determine when such a translation is
possible. In particular, we have not described how to exhibit a set of preferred direc-
tions that are consistent with a control input. Further, we have not offered a formal
definition of guidance with virtual fixtures that would permit a general theoretical
statement of an equivalence between active control and passive virtual fixturing.
These remain interesting open problems.
On the practical side all of our experiments with vision-guided virtual fixtures
have been within a very specific setup. Numerous issues must be solved before a
robust, general implementation of vision-guided virtual fixtures can be achieved.
For example, in our previous work, gain shaping was essential to maintain stabil-
ity. Similarly, there needs to be careful gain shaping to accommodate the differing
scales of forces and torques. More importantly, the ergonomics of this wider class
of guidance modes remains to be explored.
Finally, it is important to point out that most virtual fixtures apply in a very lim-
ited task context. Thus, it is important to consider how to combine guidance modes