Keywords: Control , Control architectures (thorough investigation: performance, influence parameters), control example

Abstract (copied from article):

Forcefeedback control is investigated for improving the quality of the haptic feedback in virtual reality applications. Advanced control design can increase the transparency of the haptic device at the haptic interface thereby increasing the realism of the simulation. Force feedback also enables the implementation of admittance control approachesheretofore consideredthedomainoflargeroboticplatforms.Thequality
ofthehapticinterfaceis''rstquanti''ed,andthestandard open-loopimpedanceapproachtohapticcontrolisreviewed.
Dual-inputcontrolschemeswithsensoryforcefeedbackarethen introduced,andtheresultingqualityofthehapticinterfaceis
derived.Afour-barlinkageexampleisusedtoillustratetheimprovementin''delityrealizedwiththevariousapproaches.The
tradeoffsencounteredinmovingtoforcefeedbackcontrollers for haptic applications are also discussed.

1. Introduction
The quality of a force-position interface with a virtual environment can be measured in (A) impedance accurancy (how close does the sensed impedance match the virtual impedance) and (B) resolution/fidelity (level of impedance discrimination that can be detected at the haptic device). For surgical simulation B is the most important.
Dynamics of the device is one of the barriers to achieve high impedance accurancy and precision, (natural frequencies distract the user from the virtual environment).l
Further reductions in the haptic device dynamics will likely come from active control (eg model feedforward or force feedback from a force-torque sensor mounted at the haptic interface)

2. Previous work
It has long been recognized that the dynamics of the haptic device interfere with the realism of haptic feedback, and the drive to achieve low dynamics or irtransparencyl has always been regarded as a desirable feature in hand controller design
Force-torque sensors have rarely been implemented in haptic devices due to the added mass and cost penalties. However, the advantages of incorporating them have been known from the start.

3. Haptic control design
Most of the simulators constructed so far concentrate on modeling the environment and use relatively simple control laws for the haptic feedback.
Discription PHANToM: The PHANToM, is a three degree of freedom device consisting of a four-bar parallel linkage attached to a vertical rotational axis. The parallel linkage allows the three motors to be mounted at the base giving an apparent mass at the tool tip of about 100 grams. Cable drives are used for torque amplication to reduce the effect of friction usually found in geared mechanisms. A continuous force of 1.5 N can be generated at the tool tip and peak forces of 10 N can be applied.

Haptic engineers typically employ three criteria when designing haptic mechanisms [Massie and Salisbury, 1994, ID051]:
A. free space must feel free: natural dynamics of device should not distract user from the scene (apparent mass and friction should be low. Phantom: <0.2N (**AFR: is the friction not smaller than 0.02N??**))
B. solid virtual objects must feel stiff: usually this is taken to be a minimum of 20N/cm
C. virtual constraints must not be easily saturated: for fingertip contact, a force of 10 N is rarely exceeded. However, for a grasp situation, the force can be much higher.

Two classes of control schemes:
A Impedance control: controlled system detects movement commanded by operator and control force applied to haptic device
B. Admittance control: controlled system detects force commanded by the operator and control velocity/displacement of the haptic device
NB. Sometimes, force is used as an additional input to the impedance controller or displacement is used as an additional input to the admittance controller.

Controllers demonstrated:
3.1 Open-loop impedance control (fig 6)
Open-loop impedance control with model feedforward (extra system model in control scheme should cancel out dynamic effects of tht haptic device)
3.2 Impedance controller with force feedback (extra input: force sensor. Now it's possible to add a control-loop that actually follos the desired force)
3.3 Admittance contrl (not pure admittance control but rather admittance control with position feedback): measure force, calculate desired tool position in outer loop, use inner-loop servo-controller to track haptic position with the desired position. The outer loop with difficult/slow reverse kinematic calculations can be run at a lower rate than the inner loop.

4. Impedance
Equation 23 shows a measure for the relative impedance error.

5. Example
Example: four-bar haptic device: modelling
5.3 Impedance analysis: nice controller evaluation. Read for yourself!!!

6.Result and discussion.
Read for yourself.
Note: the article mentiosn the effect of noice from the force-sensor. This is also described in article ID011!!

7.Conclusion
It cannot be concluded from this analysis that one control approach is clearly superior. The impedance controller with force feedback is promising if the stability problems associated with explicit force feedback can be overcome. Admittance control may offer more immunity to this problem because of its filtering effect on the force signal. The disturbance rejection properties of the dual-input admittance controller are attractive, but the need for high joint PD gains contraindicates its use at low values of joint inertia and sample rate. The effect of target impedance is
clearly divisive as the admittance controller is superior at low target impedances and the impedance controller is superior at high target impedances. The choice may end up being determined by the environment being simulated as well as the characteristics of the haptic device.