Precision Pointing of Segmented Reflectors Using Decentralized Control

Building a larger-aperture telescope by fabricating the optical components in sections, rather than by attempting to polish the continuous surface from a single piece of glass, is an especially attractive concept. The problems of packaging a large optical system in its launch vehicle are more tractable if the system can be launched in a folded configuration and then deployed into orbit.

Although multiple mirror designs have many advantages, a number of major difficulties are associated with the technique. Specifically, the ability to provide phasing of the separate beams (keeping the optical paths from the multiple reflectors constantly to the fraction of the wavelength of light) is an especially difficult and as-yet-unsolved problem. Because multiple-mirror telescopes are independent devices, combining their images at one focal plane is also a difficult task. Not only do the images have to be precisely "stacked" on one another, but the individual focal planes must be coplanar as well. This problem requires special consideration in the optical design so that the individual focal planes can be properly aligned.

The performance of any optical system is a direct function of the accurate positioning of its components. A monolithic reflector depends on the mathematical properties of its material to provide the dimensional stability required for good optical performance. A reflector built from segments relies on its support structure for stiffness and rigidity and on an active control system for precision alignment of the optical surface. Therefore, an active control system is required to maintain the alignment of the segmented reflectors. This control system will also be necessary to achieve high-precision figure control and maintenance of the reflector surface within 1 micrometer of the calibrated parabolic reference figure in a dynamic disturbance environment. Several technical challenges in the areas of sensing, actuation, modeling, and controls arise in the design and implementation of such a control system.

Specifically, control of the optics associated with the NGST mission requires development of technologies for precision pointing, vibration attenuation, fault identification, controller reconfiguration, adaptive/robust control, decentralized control, neuro-fuzzy control, and system identification. For the past seven years, Dr. Boussalis and the SPACE team of Co-PIs, as well as undergraduate and graduate students, have worked on the design and fabrication of the SPACE testbed and the development of advanced decentralized control algorithms. To enable reflector shape control, the team has developed and experimentally validated technologies for fault identification, adaptive/neural control, and system identification on the state-of-the-art testbed. The results obtained will become the foundation of the proposed research focused on precision pointing.

Figure below shows the major features of the SPACE testbed including the primary and secondary mirrors, the supporting structure, and the isolation platform. The testbed is designed to emulate a Cassegrain telescope of 2.4-meter focal length with performance comparable to an actual space-borne system. The system's top-level requirements figure maintenance of the primary mirror to within 1 micron RMS distortion with respect to a nominal shape of the primary mirror, pointing accuracy of 2 arc seconds, a high level of disturbance rejection (100:1), and attenuation of vibration due to gravity, thermal and seismic effects, and control structure interaction. The primary mirror is composed of a ring of six actively controlled hexagonal panels arranged around a central panel. The central panel is fixed and serves as a point of reference to the moving panels. The testbed is a control-oriented experimental system and due to the difficulty and added expense of actual optical quality segments made from glass, the panels are made of aluminum honeycomb plates. The required paraboloid surface is thus maintained by positioning the flat panels as tangents to the surface.

The primary mirror is supported by a specially designed lightweight truss structure whose structural dynamic characteristics are representative of a large, flexible space-borne system. The entire testbed is supported on a triangular isolation platform made of aluminum honeycomb core with stainless steel top and bottom skin.

The SPACE testbed active secondary mirror is attached to the primary by a tripod. This critical feature has been added to the testbed to make its performance and functionality more comparable to an actual space-borne telescope. It has been designed to provide three-axis active control, with control system hardware that consists of a number of reluctance actuators and position sensors that move and control the secondary mirror.

The SPACE testbed is fitted with an optical scoring system consisting of the secondary mirror, a laser source placed in the center of the primary, and an array of optical sensors that will detect any deviation from a reference position. The testbed's control system is composed of a data acquisition system, computers, and actuators. The control system consists of an ensemble of 24 edge sensors mounted on the peripheries of the six actively controlled segments; 18 high-bandwidth, high-force-output, voice-coil linear actuators mounted on the truss, three per each moving panel affecting three degree of freedom motion. These actuators are fitted with collocated position sensors. The sensor signal is processed through a pre-amplifier for signal conditioning and then to an A/D converter. A digital signal processor, a Pentek 4285 with four Texas Instruments TMS320C40 processors, is used to process the data, implement control algorithms, and send commands to a D/A converter and a Glentek GA4555P linear amplifier, which in turn move the system actuators to implement control. To perform experiments related to pointing, a slewing mechanism, already designed, will be fabricated and added to the structure.

In the next two years, the SPACE Laboratory will develop an accurate finite-element model including the suspension system. Using this model, the principal modes related to the motion of the secondary mirror, its structure, and its interaction with then-primary dynamics will be carefully studied using software that integrates optics. A simulation model for control performance evaluation will be developed for comparison with experimental test results.

The segmented reflector telescope under consideration consists of a large number of structural components (reflector panels, supporting and peripheral structures), as well as sensors and actuators leading to mathematical models that involve hundreds of states. Consequently, the design of control laws based on conventional methodologies becomes exceedingly difficult. Decentralized control appears to be a viable approach in circumventing the difficulties related to the dimensionality problem. Because of the nature of the structure, the research will employ decentralization techniques for the development of control laws to accomplish precision pointing and vibration control. It will implement decomposition techniques that will result in physical or mathematical decentralization of the structure into lower-order subsystems. The design of the control laws will demonstrate that: a) the flexible structure may consist of a considerable number of components that are weakly interconnected in terms of dynamics and actuation; and b) failure in one component should be accommodated by the overall structure and should not degrade the overall performance. Control laws for precision pointing will be derived for each isolated subsystem including the secondary mirror. These conditions and the properties of the interconnecting patterns will be further utilized to derive control laws for precision pointing of the overall structure. Several control laws including adaptive and neural control will be used to achieve precision pointing.

Adaptive control is an advanced technique with learning capability that has been successfully applied to a large number of applications in aerospace, robotics, and the chemical and automotive industries, as well as in the area of flexible structures. It processes input-output information in order to learn about changes in the dynamics of the system that may arise because of certain failures, degradation of equipment, etc., and to adjust itself to accommodate them. In their work in the SPACE Laboratory, the PIs have developed an effective adaptive control scheme and applied it to the SPACE structure to accomplish shape control. This URC research effort will utilize a similar approach to develop adaptive control techniques to achieve precision pointing. In the design process, the weak interconnections between the various components will be neglected. An adaptive controller will be designed for each isolated subsystem and for the secondary mirror. This approach will lead to a decentralized adaptive control scheme that will meet the overall performance requirements despite the presence of the unmodeled interconnections.

The Investigators will also utilize neural controllers in conjunction with traditional controllers to achieve precision pointing. Neural networks, with their inherent capability to approximate complex nonlinear functions to within arbitrary accuracy, have received considerable attention in the areas of dynamic system identification and control. These networks may be utilized to obtain non-parametric representations of a system from input/output data without prior knowledge of its dynamic characteristics. Neural-network architectures can also be utilized for control of dynamic systems either as fixed-gain controllers with synaptic weights pre-tuned using the identified plant model, or trained on-line in a direct adaptive control setting.