The application of this Model Predictive Sliding Mode Control scheme (Fig.
9.1) to a teleoperation system is straightforward, if we modify the desired impedance in order to reflect the slave environment force
\(\varvec{F}_e\) on the master (
9.9), and ensure master reference tracking
\(\varvec{\tilde{x}} = \varvec{x}_s - k_p\varvec{x}_m^d\) for the slave (
9.10), where
\(k_f\) and
\(k_p\) are force and position scaling factors, and
\((\cdot )^d\) indicates a delayed quantity due to the communication delay between the two systems.
$$\begin{aligned} \varvec{M}_m\varvec{\ddot{x}}_m + \varvec{D}_m\varvec{\dot{x}}_m + \varvec{K}_m\varvec{x}_m = \varvec{F}_h - k_f\varvec{F}_e^d \end{aligned}$$
(9.9)
$$\begin{aligned} \varvec{M}_s\varvec{\ddot{\tilde{x}}} + \varvec{D}_s\varvec{\dot{\tilde{x}}} + \varvec{K}_s\varvec{\tilde{x}} = -\varvec{F}_e \end{aligned}$$
(9.10)
Tuning guidelines for the master and slave parameters can be obtained by analyzing the hybrid matrix
\(\varvec{H}\) describing the interconnected system, while guaranteeing teleoperation stability in presence of delays.
$$\begin{aligned} \begin{bmatrix} F_h \\ -\dot{x}_s \end{bmatrix} = \varvec{H} \begin{bmatrix} \dot{x}_m \\ F_e \end{bmatrix} = {\begin{bmatrix} \frac{M_ms^2 + D_ms + K_m}{s} &{} \frac{k_f}{(1+s\tau )^2}e^{-sd_2} \\ -k_pe^{-sd_1} &{} \frac{s}{M_ss^2 + D_s*s + K_s} \end{bmatrix}} \begin{bmatrix} \dot{x}_m \\ F_e \end{bmatrix} \end{aligned}$$
(9.11)
where
\(\tau \) is the force feedback filter constant, and
\(d_1\),
\(d_2\) the communication delays. From (
9.11), Llewellyn’s absolute stability condition requires that all impedance dynamic parameters are greater than zero, and that the following condition is satisfied, with
\(d_{rt}\) the round-trip communication delay.
$$\begin{aligned}&\Lambda (\omega ) = \frac{2D_mD_s\omega ^2}{D_s^2\omega ^2 + (M_s\omega ^2-K_s)^2} + \frac{k_pk_f}{1+\omega ^2\tau ^2}\cdot \\ \nonumber&\cdot \left( \frac{(1-\omega ^2\tau ^2)\cos (d_{rt}\omega )-2\omega \tau \sin (d_{rt}\omega )}{1+\omega ^2\tau ^2} -1\right) \ge 0, \quad \forall \omega \ge 0 \end{aligned}$$
(9.12)
Given the previous equation, we can show that the optimal tuning that maximizes teleoperation transparency is obtained by solving the optimization problem
$$\begin{aligned} \begin{aligned}&\max _{D_m,M_s,D_s,K_s} \quad \omega _0\\ s.t. \quad&0< D_m \le \overline{D}_m \\&0< \underline{M}_s \le M_s\\&0< D_s\\&0 < K_s \ll 2\sqrt{D_mD_s/(k_pk_f(2\tau +\bar{d}_{rt})^2)}\\&\tau = M_s\sqrt{k_pk_f/(D_mD_s)}\\&D_s = 2\sqrt{M_sK_s} \end{aligned} \end{aligned}$$
(9.13)
where
\(\omega _0\) is the largest zero crossing frequency of (
9.12) when the force feedback is unfiltered. This formulation and tuning of the teleoperation controller is able to provide better performance transparency without sacrificing stability, compared to modern time-domain passivity approaches [
16].