SEARCH · Engineering Papers
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
In this paper a hierarchical control structure using a fuzzy system for coordination of the control actions is studied. The architecture involves two levels of control: a coordination level and an execution level. Numerical experiments will be utilized to illustrate the behavior of the controller when it is applied to a nonlinear plant.
The Multiple Mirror Telescope (MMT) under development at the University of Arizona takes a new approach in adaptive optics placing a large (0.65 m) force-actuated, thin facesheet deformable mirror at the secondary of an astronomical telescope, thus reducing the effects of emissivity which are important in IR astronomy. However, The large size of the mirror and low stiffness actuators used drive the natural frequencies of the mirror down into the bandwidth of the atmospheric distortion. Conventional adaptive optics takes a quasi-static approach to controlling the, deformable mirror. However, flexibility within the control bandwidth calls for a new approach to adaptive optics. Dynamic influence functions are used to characterize the influence of each actuator on the surface of the deformable mirror. A linearized model of atmospheric distortion is combined with dynamic influence functions to produce a dynamic reconstructor. This dynamic reconstructor is recognized as an optimal control problem. Solving the optimal control problem for a system with hundreds of actuators and sensors is formidable. Exploiting the circularly symmetric geometry of the mirror, and a suitable model of atmospheric distortion, the control problem is divided into a number of smaller decoupled control problems using circulant matrix theory. A hierarchic control scheme which seeks to emulate the quasi-static control approach that is generally used in adaptive optics is compared to the proposed dynamic reconstruction technique. Although dynamic reconstruction requires somewhat more computational power to implement, it achieves better performance with less power usage, and is less sensitive than the hierarchic technique.
The theory of intelligent machines has been recently reformulated to incorporate new architectures that are using neural and Petri nets. The analytic functions of an intelligent machine are implemented by intelligent controls, using entropy as a measure. The resulting hierarchical control structure is based on the principle of increasing precision with decreasing intelligence. Each of the three levels of the intelligent control is using different architectures, in order to satisfy the requirements of the principle: the organization level is moduled after a Boltzmann machine for abstract reasoning, task planning and decision making; the coordination level is composed of a number of Petri net transducers supervised, for command exchange, by a dispatcher, which also serves as an interface to the organization level; the execution level, include the sensory, planning for navigation and control hardware which interacts one-to-one with the appropriate coordinators, while a VME bus provides a channel for database exchange among the several devices. This system is currently implemented on a robotic transporter, designed for space construction at the CIRSSE laboratories at the Rensselaer Polytechnic Institute. The progress of its development is reported.
In this study artificial neural networks and fuzzy logic are used to control the jumping behavior of a three-link uniped robot. The biped locomotion control problem is an increment of the uniped locomotion control. Study of legged locomotion dynamics indicates that a hierarchical controller is required to control the behavior of a legged robot. A structured control strategy is suggested which includes navigator, motion planner, biped coordinator and uniped controllers. A three-link uniped robot simulation is developed to be used as the plant. Neurocontrollers were trained both online and offline. In the case of on-line training, a reinforcement learning technique was used to train the neurocontroller to make the robot jump to a specified height. After several hundred iterations of training, the plant output achieved an accuracy of 7.4%. However, when jump distance and body angular momentum were also included in the control objectives, training time became impractically long. In the case of off-line training, a three-layered backpropagation (BP) network was first used with three inputs, three outputs and 15 to 40 hidden nodes. Pre-generated data were presented to the network with a learning rate as low as 0.003 in order to reach convergence. The low learning rate required for convergence resulted in a very slow training process which took weeks to learn 460 examples. After training, performance of the neurocontroller was rather poor. Consequently, the BP network was replaced by a Cerebeller Model Articulation Controller (CMAC) network. Subsequent experiments described in this document show that the CMAC network is more suitable to the solution of uniped locomotion control problems in terms of both learning efficiency and performance. A new approach is introduced in this report, viz., a self-organizing multiagent cerebeller model for fuzzy-neural control of uniped locomotion is suggested to improve training efficiency. This is currently being evaluated for a possible patent by NASA, Johnson Space Center. An alternative modular approach is also developed which uses separate controllers for each stage of the running stride. A self-organizing fuzzy-neural controller controls the height, distance and angular momentum of the stride. A CMAC-based controller controls the movement of the leg from the time the foot leaves the ground to the time of landing. Because the leg joints are controlled at each time step during flight, movement is smooth and obstacles can be avoided. Initial results indicate that this approach can yield fast, accurate results.
A two-level hierarchical control strategy is proposed for large flexible space structures. The lower level consists of a set of local controllers. The higher level is a stabilizing compensator to account for any instabilities caused by controller-structure interaction with unmodeled dynamics. The advantage of this hierarchical strategy is that the lower level can be designed to meet the performance requirements, and the higher level can be designed independently to produce overall stability.
The problem of control systems synthesis is considered for controlling the rigid body attitude and elastic motion of a large deployable space based antenna. Two methods for control systems synthesis are considered. The first method utilizes the stability and robustness properties of the controller consisting of torque actuators and collocated attitude and rate sensors. The second method is based on the linear quadratic Gaussian (LQG) control theory. A combination of the two methods, which results in a two level hierarchical control system, is also briefly discussed. The performance of the controllers is analyzed by computing the variances of pointing errors, feed misalignment errors and surface contour errors in the presence of sensor and actuator noise.
Control systems synthesis is considered for controlling the rigid body attitude and elastic motion of a large deployable space-based antenna. Two methods for control systems synthesis are considered. The first method utilizes the stability and robustness properties of the controller consisting of torque actuators and collocated attitude and rate sensors. The second method is based on the linear-quadratic-Gaussian control theory. A combination of the two methods, which results in a two level hierarchical control system, is also briefly discussed. The performance of the controllers is analyzed by computing the variances of pointing errors, feed misalignment errors and surface contour errors in the presence of sensor and actuator noise.
The problem of control systems synthesis is considered for controlling the rigid-body attitude and elastic motion of a large deployable space-based antenna. Two methods for control systems synthesis are considered. The first method utilizes the stability and robustness properties of the controller consisting of torque actuators and collocated attitude and rate sensors. The second method is based on the linear-quadratic-Gaussian (LQG) control theory. A combination of the two methods, which results in a two-level hierarchical control system, is also briefly discussed. The performance of the controllers is analyzed by computing the variances of pointing errors, feed misalignment errors and surface contour errors in the presence of sensor and actuator noise.
An experimental and analytical study was undertaken to investigate human interaction with a simple multiloop manual control system in which the human's activity was systematically varied by changing the level of automation. The system simulated was the longitudinal dynamics of a hovering helicopter. The automation-systems-stabilized vehicle responses from attitude to velocity to position and also provided for display automation in the form of a flight director. The control-loop structure resulting from the task definition can be considered a simple stereotype of a hierarchical control system. The experimental study was complemented by an analytical modeling effort which utilized simple crossover models of the human operator. It was shown that such models can be extended to the description of multiloop tasks involving preview and precognitive human operator behavior. The existence of time optimal manual control behavior was established for these tasks and the role which internal models may play in establishing human-machine performance was discussed.
Results of a three phase research program into intelligent control systems are presented. The first phase looked at implementing the lowest or direct level of a hierarchical control scheme using a reinforcement learning approach assuming no a priori information about the system under control. The second phase involved the design of an adaptive/optimizing level of the hierarchy and its interaction with the direct control level. The third and final phase of the research was aimed at combining the results of the previous phases with some a priori information about the controlled system.
The Wide-Field Infrared Explorer (WIRE), currently scheduled for launch in September 1998, is the fifth of five spacecraft in the NASA/Goddard Small Explorer (SMEX) series. This paper presents the design of WIRE's Attitude Control System flight software (ACS FSW). WIRE is a momentum-biased, three-axis stabilized stellar pointer which provides high-accuracy pointing and autonomous acquisition for eight to ten stellar targets per orbit. WIRE's short mission life and limited cryogen supply motivate requirements for Sun and Earth avoidance constraints which are designed to prevent catastrophic instrument damage and to minimize the heat load on the cryostat. The FSW implements autonomous fault detection and handling (FDH) to enforce these instrument constraints and to perform several other checks which insure the safety of the spacecraft. The ACS FSW implements modules for sensor data processing, attitude determination, attitude control, guide star acquisition, actuator command generation, command/telemetry processing, and FDH. These software components are integrated with a hierarchical control mode managing module that dictates which software components are currently active. The lowest mode in the hierarchy is the 'safest' one, in the sense that it utilizes a minimal complement of sensors and actuators to keep the spacecraft in a stable configuration (power and pointing constraints are maintained). As higher modes in the hierarchy are achieved, the various software functions are activated by the mode manager, and an increasing level of attitude control accuracy is provided. If FDH detects a constraint violation or other anomaly, it triggers a safing transition to a lower control mode. The WIRE ACS FSW satisfies all target acquisition and pointing accuracy requirements, enforces all pointing constraints, provides the ground with a simple means for reconfiguring the system via table load, and meets all the demands of its real-time embedded environment (16 MHz Intel 80386 processor with 80387 coprocessor running under the VRTX operating system). The mode manager organizes and controls all the software modules used to accomplish these goals, and in particular, the FDH module is tightly coupled with the mode manager.
In this paper, we propose a design of an active vision system for intelligent robot application purposes. The system has the degrees of freedom of pan, tilt, vergence, camera height adjustment, and baseline adjustment with a hierarchical control system structure. Based on this vision system, we discuss two problems involved in the binocular gaze stabilization process: fixation point selection and vergence disparity extraction. A hierarchical approach to determining point of fixation from potential gaze targets using evaluation function representing human visual behavior to outside stimuli is suggested. We also characterize different visual tasks in two cameras for vergence control purposes, and a phase-based method based on binarized images to extract vergence disparity for vergence control is presented. A control algorithm for vergence control is discussed.
The concept of a generic module for management of onboard systems grew out of the structured analysis effort for the Space Station software. Hierarchical specification of subsystems software revealed that nontrivial supervisory elements are required at all levels. The number of supervisors (and subsequent software) required to implement the hierarchical control over onboard functions comprise a large portion of the Space Station software. Thus, a generic knowledge based supervisory module significantly reduces the amount of software developed. This module, the Generic Supervisor, depends on its knowledge of control to provide direction for subordinates and feedback to superiors within a specific subsystem area. The Generic Supervisor provides an adaptable and maintainable control system. A portion of the Space Station Environmental Control and Life Support System (ECLSS) was implemented as a hierarchy of supervisors. This prototype implementation demonstrates the feasibility of a generic knowledge based supervisor, and its facility to meet complex mission requirements.
This report contains the development of a digital outer-loop three dimensional radio navigation (3-D RNAV) flight control system for a small commercial jet transport. The outer-loop control system is designed using optimal stochastic limited state feedback techniques. Options investigated using the optimal limited state feedback approach include integrated versus hierarchical control loop designs, 20 samples per second versus 5 samples per second outer-loop operation and alternative Type 1 integration command errors. Command generator tracking techniques used in the digital control design enable the jet transport to automatically track arbitrary curved flight paths generated by waypoints. The performance of the design is demonstrated using detailed nonlinear aircraft simulations in the terminal area, frequency domain multi-input sigma plots, frequency domain single-input Bode plots and closed-loop poles. The response of the system to a severe wind shear during a landing approach is also presented.
The basic components that make up the goal of building autonomous intelligent systems are discussed, and ongoing work at the NASA Ames Research Center is described. It is noted that a clear progression of systems can be seen through research settings (both within and external to NASA) to Space Station testbeds to systems which actually fly on the Space Station. The starting point for the discussion is a 'truly' autonomous Space Station intelligent system, responsible for a major portion of Space Station control. Attention is given to research in fiscal 1987, including reasoning under uncertainty, machine learning, causal modeling and simulation, knowledge from design through operations, advanced planning work, validation methodologies, and hierarchical control of and distributed cooperation among multiple knowledge-based systems.
The basic components that make up the goal of building autonomous intelligent systems are discussed, and ongoing work at the NASA Ames Research Center is described. It is noted that a clear progression of systems can be seen through research settings (both within and external to NASA) to Space Station testbeds to systems which actually fly on the Space Station. The starting point for the discussion is a truly autonomous Space Station intelligent system, responsible for a major portion of Space Station control. Attention is given to research in fiscal 1987, including reasoning under uncertainty, machine learning, causal modeling and simulation, knowledge from design through operations, advanced planning work, validation methodologies, and hierarchical control of and distributed cooperation among multiple knowledge-based systems.
In the area of control, several problems can arise during the evolution of the space station. These incude: 1) the use of multiple or articulated flexible bodies; 2) the need for distributed cntrol for maneuvering and maintaining altitude; 3) hierarchical control to automate and manage control systems; 4) structual control (from the standpoint of appendage stamping; isolation, and possible figure control); 5) cntrol position and orientation for component modules during construction (an evolutionary requirement); 6) control during docking and undocking operations; and 7) the normal requirements for stablity and control during systems operations. In addition, there are a number of key technology concerns, such as significant landing modes which tend to be closely spaced and distributed widely, distributed sensors and actuators which are collecated, and the wide distribution of landing modes that must be reduced from a dimensional standpoint.
Adaptive optics systems with Shack-Hartmann wavefront sensors require reconstruction of the atmospheric phase error from slope measurements, with every sensor in the array being used in the computation of each actuator command. This fully populated reconstruction matrix can result in a significant computational burden for adaptive optics systems with large numbers of actuators. A method for generating sparse wavefront reconstruction matrices for adaptive optics isproposed. The method exploits the relevance of nearby slope measurements for control of an individual actuator, and relies upon the limited extent of the influence function for a zonal deformable mirror. Relying only on nearby sensor information can significantly reduce the calculation time for wavefront reconstruction. In addition, a hierarchic controller is proposed to recover some of the global wavefront information. The performance of these sparse wavefront reconstruction matrices was evaluated in simulation, and tested on the Palomar Adaptive Optics System. This paper will present some initial results from the simulations and experiments.