Experiments in control of flexible structures with noncolocated sensors and actuators

Abstract

Experimental apparatus has been developed for physically testing control systems for pointing flexible structures, such as limber spacecraft, in the event that control actuators cannot be colocated with the sensors. (An example is the Galileo spacecraft, whose television camera at one end of a flexible beam must be pointed by torquiiig at the other end of the beam). With colocation, good stable control is very easy to achieve. With noncolocation it is extremely difficult, particularly if structural damping is very low and spacecraft stiffness and inertia values are uncertain and changing, as they are typically. For the apparatus we have built, structural damping ratios are less than 0.003, each basic configuration of sensor/actuator noncolocation is available, and inertias can be halved or doubled abruptly during control maneuvers, thereby imposing in particular a sudden reversal in the plant's pole-zero sequence, a most difficult problem for the controller. First experimental results are presented, including stable control with both colocation and noncolocation. The inherent robustness of the former is clearly seen, as is the great difficulty of achieving robustness for the latter. (Schemes for doing so are now being explored, and future experiments will establish what the best achievable robust but nonadaptive control is, and will develop adaptive control.) What we hope to contribute here is a "red flag" warning about noncolocated control of flexible structures: there are configurations, indeed, simple ones, for which there may be no practical alternative to adaptive control.

Series elastic actuators

“…They found that the fundamental limit on the force control bandwidth is the link stiffness. Like Cannon [8] they found that non-colocated dynamics can lead to unstable systems. They also considered the effects of simple force control laws on a model of an actuator and the environment.…”

“…Cannon [8] was the first to investigate non-colocated sensors (where the sensor is not located at the actuator to be controlled). Cannon showed that colocation allows good stable control, while any significant dynamics between the sensor and the actuator can lead to instability.…”

An investigation of the active damping of suspension bridges

“…It turns out that cable-structure systems are much easier to control if collocated actuator-sensor pairs are used because this produces alternating poles and zeros in the open-loop transfer function of every channel of the control system [Cannon and Rosenthal 1984;Preumont 2011], drastically reducing the spillover and other problems associated with the high-frequency dynamics. This property was successfully exploited in several studies demonstrating the active damping of cablestayed bridges [Achkire and Preumont 1996;Bossens and Preumont 2001] and guyed space trusses Preumont and Achkire 1997;.…”

Untitled

“…All these studies were performed with noncollocated actuator-sensor configurations; this did not lead to any trouble in Yang's numerical study since perfect knowledge and a linear system were assumed, but Fujino's experimental results revealed that, even for relatively simple systems, instabilities tend to occur when the cable-structure interaction is large. It turns out that cable-structure systems are much easier to control if collocated actuator-sensor pairs are used because this produces alternating poles and zeros in the open-loop transfer function of every channel of the control system [Cannon and Rosenthal 1984;Preumont 2011], drastically reducing the spillover and other problems associated with the high-frequency dynamics. This property was successfully exploited in several studies demonstrating the active damping of cablestayed bridges [Achkire and Preumont 1996;Bossens and Preumont 2001] and guyed space trusses Preumont and Achkire 1997;.…”