Method and apparatus for hybrid resetting states proportional-integral-derivative and lag controllers

Abstract

A controller includes a feedforward module to produce a feedforward component azr, where a is a high frequency gain, a proportional module is to produce a proportional component Kpvz, where Kpv is a proportional gain, and an integral or lag module with state resetting to produce an integral or lag component. The feedforward component azr, the proportional component Kpvz, and the integral component Kivzc are combined as an input u for the system to be controlled.

Description

FIELD OF THE INVENTION

This invention relates generally controlling systems, and in particular to resetting states of proportional-integral-derivative (PID) and lag controllers.

BACKGROUND OF THE INVENTION

Traditional controllers were designed for systems that output either discrete or continuous signals. More recent hybrid controllers provide integrated control for discrete and continuous signals. Hybrid controlled systems can enable higher performance, versatility, and autonomy. A specific problem in hybrid controllers is resetting the integrator and first order reset elements (FORE).

The state of a linear controller is reset to zero, or same other known value, whenever some threshold is reached. The Clegg integrator and the FORE are now being revisited within the framework of hybrid controllers to overcome fundamental limitations associated with linear feedback, such as overshoot and settling time bounds.

The following U.S. patents describe dynamic state resetting. U.S. Pat. No. 5,929,700 describe a method for a nonlinear notch filter employing a type of filter state resetting with reduced phase lag compared to a linear notch, which is used as part of a controlled system to improve the bandwidth of systems with resonant modes. U.S. Pat. No. 5,694,077 describes a method for resetting different types of filters such as low pass, high pass, and bandpass filters to reduce their phase lag.

SUMMARY OF THE INVENTION

The embodiments of the invention provide a method and system for resetting a state of hybrid controller using generalized integral and first order resetting control techniques for use with a proportional-integral-derivative (PID) and lag controller for first and second order dominant controlled system. This reduces to a Clegg integrator for integral control of a first order controlled system. In addition, different modifications to alleviate design trade-offs are introduced.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a block diagram of a hybrid resetting controller according to an embodiment of the invention;

FIG. 2 is a block diagram of a lag hybrid resetting controller according to an embodiment of the invention;

FIG. 3 is a block diagram of a high order hybrid resetting controller according to an embodiment of the invention;

FIG. 4 is a block diagram of a resetting controller according to the embodiments of the invention; and

FIG. 5 is a graph comparing states of an integrator in a controller with and without resetting.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

A dynamic hybrid resetting (impulsive) system can be modeled as{dot over (x)}=ƒ_c(x) if (t,x)∉S_r x⁺=ƒ_d(x), if (t,x)∈S_r  (1)where x generally indicates states of the system and controller. The states are reset selectively as described below. That is, some but not necessarily all states are reset. The dots above the variables ({dot over ( )}) and ({umlaut over ( )}), here and below, are the conventional way for indicating an order of the derivative as known in the art, (+) indicates a next state, an input to the system is u, time is t, the Lipschitz continuous function ƒ_c describes the continuous-time dynamics of the controller, and the function ƒ_d is the resetting law with S_r being a resetting set that defines the resetting condition based on time and/or state.

It is assumed that the resetting times t_k are well-defined and distinct, i.e., and t_k−t_k-1≧∈ for ∈>0. Thus, the controller state is non-zero with well-defined solutions. This can be achieved by imposing a sampling on the resetting condition, as would be the case in a practical implementation, or by additional time-based condition.

We consider a class of systems s that includes a chain of integratorsay⁽ⁿ⁾=u,  (2)where yⁿ is the n^th derivative of a targeted output y, and n≦2 is a selected dominant order of the system. The constant parameter a is a high frequency gain. It is assumed that output signals y, y^(n-1) are available, which is typical for a PID controller. Furthermore, a reference trajectory r and its first n derivatives r⁽¹⁾, . . . , r⁽ⁿ⁾ are known, bounded and, piecewise continuous.

An error variable is constructed asz=−(d/dt+K_pp)^n-1e, where K_pp>0 is a selected scalar, and e=r−y is a tracking error for a desired reference trajectory r.

We consider the following control law (input) for the controlled systemu=K_pvz−K_ivz_c+aż_r,  (3)where statesż_c=z, if zz_c≧0z_c⁺0, if zz_c≦0, andwhere K_pv>0 is a proportional gain, and K_iv>0 is an integral gain.

In the controller, the “feedforward” signal isż_r=y⁽ⁿ⁾−ż, where y⁽ⁿ⁾ is the n^th derivative of the targeted output y.

For order n=1, we haveż_r={dot over (r)}, and for n=2, we haveż_r={umlaut over (r)}+K_ppė.

FIG. 1 shows the controller according to the above models. The reference signal is r 101, the input to the system 105 is u 102, and the output of the system is y 106.

A feedforward module 110 produces a feedforward signal ż_r 106 to which a gain a 111 is applied to produce a feedforward component for the input u 102. FIG. 1 shows the feedforward signal for order n=1 and n=2.

A proportional module 120 produces a proportional signal z 107 with gain K_pv 121 based on the error variable e 105 and the reference signal r, which forms the proportional component for the input u.

An integral or lag module 130 produces an integral or lag signal 108 with an integral gain K_iv 131 to produce an integral component for the input u.

The feedforward module, the proportional module and the integral or lag module form a controller for the system 105. At any one time, the controller has a state, and according to the embodiments of the invention, the integral or lag module is configured to resetting this state. That is, in contrast with the prior art, the module 130 is capable of resetting the controller as described in further detail below. In the embodiments described herein, the state is represented by signal z_c, or z_ci in some embodiments with multiple states. Generally, a reset occurs when z_c or z_ci=0.

The output y 106 of the system 105 is subtracted from the reference signal r to produce the tracking error e for the proportional module 120.

Substituting Equation (3) into Equation (2) yields the following hybrid resetting closed loop ({dot over (x)}_c1) controller and system

$\begin{array}{cc}{\stackrel{.}{x}}_{\mathrm{cl}}=A_c{x}_{\mathrm{cl}},\mathrm{if}{\mathrm{zz}}_c\ge 0x_{\mathrm{cl}}^{+}=A_d{x}_{\mathrm{cl}},\mathrm{if}{\mathrm{zz}}_c\le 0,\mathrm{where}{x}_{\mathrm{cl}}={\left[z,z_c\right]}^T,\mathrm{and}{A}_c=\left[\begin{array}{cc}-K_{\mathrm{pv}}/a& -K_{\mathrm{iv}}/a\\ 1& 0\end{array}\right]A_d=\left[\begin{array}{cc}1& 0\\ 0& 0\end{array}\right].& \left(4\right)\end{array}$

We consider the following Lyapunov functionV=az²+K_ivz_c².

Computing {dot over (V)} for the continuous-time part yields{dot over (V)}=−2K_pvz²≦0,whereas, for the resetting dynamics of Equation (4) can be expressed asΔV=V(x_c1⁺)−V(x_c1)=−K_ivz_c²≦0.

Using standard stability results for impulsive (resetting) hybrid controller, this proves the Lyapunov stability of the closed loop controller and system.

By hybrid extensions of the invariant set theorem, the state of the controller converges asymptotically to the fixed-point (z=0, z₁=0), and thus the tracking error e approached zero asymptotically.

Combining a Resetting and a Non-Resetting Integrators

We consider the following control lawu=−K_pvz−K_iv1z_c1−K_iv2z_c2+aż_r,  (4)where statesż_c1=z, if zz_c1≧0z_c1⁺=0, if zz_c1≦0ż_c2=z where K_iv1>0 is the resetting integrator gain corresponding to state z_l1, and K_iv2>0 is the non-resetting integral gain corresponding to state z_l2.

Thus, we obtain the following hybrid resetting controlleru=−K_pvz−K_ivz_c+aż_r ż_c=z, if zz_c≧0z_c⁺=b_d, if zz_c≦0.  (5)

The stability of this controller can be analyzed with the following Lyapunov functionV=az²+K_iv(z_c−d/K_iv)².

Computing {dot over (V)} for the continuous-time part yields{dot over (V)}=−2K_pvz²≦0.

Whereas, for the resetting dynamics we have:ΔV=V(x_c1⁺)−V(x_c1)=−K_ivz_c²≦0.

This proves Lyapunov stability of the closed loop controller of Equation (5) with the combined integrators method.

Furthermore, by hybrid extensions of the invariant set theorem, the controller converges asymptotically to the fixed-point

$\left(z,z_{,I1},z_{I2}\right)=\left(0,0,\frac{d}{K_{\mathrm{iv}2}}\right),$ and thus the tracking error approaches zero asymptotically. Therefore, zero steady-state rejection of constant disturbances is possible.

Hybrid Resetting Lag Compensators

Here, the resetting PI compensator, which is used for order n=1 and within a PID controller for n=2, is modified to yield resetting lag compensators. A resetting lag controller can be achieved by replacing the resetting integrator of the FORE, such thatu=−K_pvz−K_ivz_c+aż_r, and  (6)the states of the controller areż_c=z−a_cz_c, if (z−a_cz_c)z_c≧0z_c⁺=0, if (z−a_cz_c)z_c≦0and where the controller time lag or pole is a_c≦0.

This controller is shown in FIG. 2, where the integral module 230 determines the integral gain component. All other components are as for FIG. 1.

Stability of the controller can be shown using the following Lyapunov functionV=az²+K_iv1z_c²,

Computing {dot over (V)} for the continuous-time part yields:{dot over (V)}=−K_pvz²−K_iva_cz_c²<0

Whereas, for the resetting dynamicsΔV=V(x_c1⁺)−V(x_c1)=−K_ivz_c²≦0.

This proves exponential stability, and input-to-state stability, of the closed loop hybrid controller because {dot over (V)}<0.

Higher Order Resetting Controller

Consider the following control law

$\begin{array}{cc}u=-K_{\mathrm{pv}}z-\stackrel{N}{\sum _i}{K}_i{z}_{\mathrm{ci}}+a_{{\stackrel{.}{z}}_r}& \left(7\right)\end{array}$ where statesż_ci=z−a_ciz_ci, if (z−a_ciz_ci)zz_ci≧0, andz_ci⁺=a_diz_ci+b_di, if (z−a_cia_ci) zz_ci≦0,where the scalars a_ci≦0 and |a_di|≦1 for N controller states, and where a_ci=0 corresponds to an integrator, and a_ci>0 corresponds to the FORE if a_di=1, and a selected offset b_dj=0 for the j^th controller state. This way the controller can operate as a non-resetting integrator controller or a lag controller.

This controller is shown in FIG. 3, where the integral module 330 determines the integral gain component. All other components are as for FIG. 1.

Let b_dj=0, and consider the following Lyapunov function:

$V={\mathrm{az}}^2+\stackrel{N}{\sum _i}{k}_i{z}_{\mathrm{ci}}^2.$

Computing {dot over (V)} for the continuous-time part with the controller according to Equation (7) yields

$\stackrel{.}{V}=-K_{\mathrm{pv}}{z}^2-\stackrel{N}{\sum _i}{K}_i{a}_{\mathrm{ci}}{z}_{\mathrm{ci}}^2\le 0,$ whereas for the resetting dynamics

$\Delta V=V\left(x_{\mathrm{cl}}^{+}\right)-V\left(x_{\mathrm{cl}}\right)=\stackrel{N}{\sum _i}{K}_i\left(a_{\mathrm{di}}^2-1\right)z_{\mathrm{ci}}^2\le 0.$

This proves the Lyapunov stability of the system with the controller given by Equation (7). Whereas, when a_ci>0 for all i, then {dot over (V)}<0 and exponential stability of the hybrid controller is shown.

Analysis

FIG. 4 shows our controller in its simplest form. The controller includes a feedforward module 410 to produce a feedforward component z_r 411 of an input u 435 for a system 440 to be controlled. A proportional module is to produce a proportional component z of the input. An integral module is configured to produce an integral component z_c based on the proportional component. The components z_r, z, and z_c are combined 435 to produce the input u for the system 440.

The output y of the system is subtracted 405 from a reference signal r to produce an error signal e 407 for the feedforward module and the proportional module. The reference signal r is also provided to the feedforward module.

FIG. 5 compares the behavior of a PID controller with (501) and without (502) resetting. Note the resetting spikes in curve 501. In FIG. 5, the vertical access is integrator state, and the horizontal axis is time. If the integral module is continuously reset to zero, the controller does not achieve optimal steady-state tracking if there is a constant disturbance. Whereas, when the resetting is switched off after 0.1 seconds, then the integral module follows the value or response needed to overcome the disturbance yet retains the transient benefits of resetting.

Although the invention has been described with reference to certain preferred embodiments, it is to be understood that various other adaptations and modifications can be made within the spirit and scope of the invention. Therefore, it is the object of the append claims to cover all such variations and modifications as come within the true spirit and scope of the invention.

Claims

1. A controller for controlling a system, comprising: a feedforward module to produce a feedforward component azr, where a is a high frequency gain;a proportional module to produce a proportional component Kpvz, where Kpv is a proportional gain;an integral or lag module to produce respective integral or lag components, wherein the feedforward module and the proportional module and the integral or lag module form a controller having a state, and the integral or lag module is capable of resetting the state; andmeans for combining the feedforward component azr, the proportional component Kpvz, and the integral or lag components to produce an input u for the system to be controlled, wherein the feedforward module is provided with a reference signal r and an error signal e, and wherein where the error signal is a difference between the reference signal and an output y of the system, and the error signal is also provided to the proportional module.

2. The controller of claim 1, wherein the feedforward component is żr={dot over (r)},for order n=1, and żr={umlaut over (r)}+Kppė, for order n=2, and wherein Kpp is a gain, z=−(d/dt+Kpp)n-1e.

3. The controller of claim 1, wherein the input is u=−Kpvz−Kivzc+ażr,where a is a high frequency gain, Kiv is an integral gain, and the integral component is żc=z, if zzc≧0zc+=0, if zzc<0,such that the controller operates as a hybrid resetting controller.

4. The controller of claim 1, wherein the input is u=−Kpvz−Kivzc+αżr.

5. The controller of claim 1, wherein the input is

6. The controller of claim 1, wherein the input is u=−Kpvz−Kiv1zc1−Kiv2zc2+ażr,where a is a high frequency gain, and żc1=z, if zzc1≧0zc1+=0, if zzc1≦0żc2=z where Kiv1>0 is the integral gain corresponding to state zl1, and Kiv2>0 is a non-resetting integral gain corresponding to state zl2.

7. The controller of claim 1, wherein the input is u=−Kpvz−Kivzc+ażr,

8. The controller of claim 1, wherein the input is u=−Kpvz−Kivzc+ażr,where, and a is a high frequency gain, and Kiv is an integral gain.

9. The controller of claim 1, wherein the input is u=−Kpvz−Kivzc+ażr,where a is a high frequency gain, Kiv is an integral gain, and the integral component is żc=z, if zzc≧0zc+=bd, if zzc≦0,where bd is a selected offset, and controller operates as a hybrid resetting controller.

10. A controller for controlling a system, comprising: a feedforward module to produce a feedforward component azr, where a is a high frequency gain;a proportional module to produce a proportional component Kpvz, where Kpv is a proportional gain;an integral or lag module to produce respective integral or lag components, wherein the feedforward module and the proportional module and the integral or lag module form a controller having a state, and the integral or lag module is capable of resetting the state; andmeans for combining the feedforward component azr, the proportional component and the integral or lag components to produce an input u for the system to be controlled, wherein the input is u=−Kpvz−Kivzc+ażr,where a is a high frequency gain, Kiv is an integral gain, and the integral component is żc=z, if zzc≧0zc+=0, if zzc≦0,such that the controller operates as a hybrid resetting controller.

11. A controller for controlling a system, comprising: a feedforward module to produce a feedforward component azr, where a is a high frequency gain;a proportional module to produce a proportional component Kpvz, where Kpv is a proportional gain;an integral or lag module to produce respective integral or lag components, wherein the feedforward module and the proportional module and the integral or lag module form a controller having a state, and the integral or lag module is capable of resetting the state; andmeans for combining the feedforward component azr, the proportional component Kpvz, and the integral or lag components to produce an input u for the system to be controlled, wherein the input is u=−Kpvz−Kivzc+ażr,where a is a high frequency gain, Kiv is an integral gain, and wherein the lag component is żc=z−aczc, if (z−aczc)zc≧0zc+=0, if (z−aczc)zc≦0, andwhere ac is the lag pole, such that the controller operates as a resetting lag controller.

12. A controller for controlling a system, comprising: a feedforward module to produce a feedforward component azr, where a is a high frequency gain;a proportional module to produce a proportional component Kpvz, where Kpv is a proportional gain;an integral or lag module to produce respective integral or lag components, wherein the feedforward module and the proportional module and the integral or lag module form a controller having a state, and the integral or lag module is capable of resetting the state; andmeans for combining the feedforward component azr, the proportional component. Kpvz, and the integral or lag components to produce an input u for the system to be controlled, wherein the input is

13. A controller for controlling a system, comprising: a feedforward module to produce a feedforward component azr, where a is a high frequency gain;a proportional module to produce a proportional component Kpvz, where Kpv is a proportional gain;an integral or lag module to produce respective integral or lag components, wherein the feedforward module and the proportional module and the integral or lag module form a controller having a state, and the integral or lag module is capable of resetting the state; andmeans for combining the feedforward component azr, the proportional component Kpvz, and the integral or lag components to produce an input u for the system to be controlled, wherein the input is u=−Kpvz−Kivzc+ażr,where a is a high frequency gain, and żc1=z, if zzc1≧0zc1+=0, if zzc1≦0,żc2=z where Kiv1>0 is the integral gain corresponding to state zl1, and Kiv2>0 is a non-resetting integral gain corresponding to state zl2.

14. A controller for controlling a system, comprising: a feedforward module to produce a feedforward component azr, where a is a high frequency gain;a proportional module to produce a proportional component Kpvz, where Kpv is a proportional gain;an integral or lag module to produce respective integral or lag components, wherein the feedforward module and the proportional module and the integral or lag module form a controller having a state, and the integral or lag module is capable of resetting the state; andmeans for combining the feedforward component azr, the proportional component Kpvz, and the integral or lag components to produce an input u for the system to be controlled, wherein the input is u=−Kpvz−Kivzc+ażr,

15. A controller for controlling a system, comprising: a feedforward module to produce a feedforward component azr, where a is a high frequency gain;a proportional module to produce a proportional component Kpvz, where Kpv is a proportional gain;an integral or lag module to produce respective integral or lag components, wherein the feedforward module and the proportional module and the integral or lag module form a controller having a state, and the integral or lag module is capable of resetting the state; andmeans for combining the feedforward component azr, the proportional component Kpvz, and the integral or lag components to produce an input u for the system to be controlled, wherein the input is u=−Kpvz−Kivzc+ażr,where a is a high frequency gain, Kiv is an integral gain, and the integral component is żc=z, if zzc≧0zc+=bd, if zzc≦0,where bd is a selected offset, and controller operates as a hybrid resetting controller.