Proceedings of the American Control Conference

Tracking or rejection of unknown exogenous signals with known generating dynamics is of major concern in feedback control design. Linear feedback control based on the internal model principle achieves asymptotic performance for linear systems with linear exogenous signal dynamics. This paper presents a control design based on the internal model principle to track or reject nonlinear exogenous signal dynamics for nonlinear systems. Necessary condition to achieve asymptotic disturbance rejection based on the proposed control structure is first derived. It is shown that the necessary condition becomes sufficient for linear systems with linear disturbance dynamics. Inspired by the unique structure of the necessary condition, sufficient conditions are then proposed. Simulations of a nonlinear plant with chaotic disturbance show the effectiveness of the proposed scheme.

Deals with a stochastic stability concept for discrete-time Markovian jump linear systems. The random jump parameter is associated to changes between the system operation modes due to failures or repairs, which can be well described by an underlying finite-state Markov chain. In the model studied, a fixed number of failures or repairs is accepted, after which, the system is brought to a halt for maintenance or for replacement. The usual concepts of stochastic stability are related to pure infinite horizon problems, and are not appropriate in this scenario. A new stability concept is introduced, named stochastic /spl tau/-stability that is tailored to the present setting. Necessary and sufficient conditions to ensure the stochastic /spl tau/-stability are provided, and the almost sure stability concept associated with this class of processes is also addressed. The paper also develops an equivalence among second order concepts that parallels the results for infinite horizon problems.

The modem market demands individual products and short delivery times. Many companies redesign their administrative and automation systems to enhance flexibility and quick response to market. This demands large-scale integration of business systems with production systems. The nature of these two classes of systems is very different. Business systems are transaction based and operate primarily on aggregated values in non-real time. Automation systems are real time based operating on single point values. The integration of the two types of systems implies they can be viewed as one large scaled distributed company control system where, often, the human acts as the controller. The innovative aspect is to combine the control related knowledge with the real, and always non-ideal, world. One result of this innovative view of a company is that the future demands to manufacturing companies very much consist of closing as many loops as possible without the necessity of a human being. The loops are not traditional machine control loops but rather those that run the company by combining business and production processes. In these loops the human being should interchange the role of a direct controller with that of a right-to-intervention.

When building smaller, less expensive spacecraft, there is a need for intelligent fault tolerance vs. increased hardware redundancy. If fault tolerance can be achieved using existing navigation sensors, cost and vehicle complexity can be reduced. A maximum-likelihood-based approach to thruster fault detection and identification (FDI) for spacecraft is developed here and applied in simulation to the X-38 space vehicle. The system uses only gyro signals to detect and identify hard, abrupt, single- and multiple-jet on- and off-failures. Faults are detected within one second and identified within one to five seconds.

In this paper we develop data compression techniques for subspace methods to extend their ability to work with long data records. Satisfactory results are obtained when a subspace algorithm is used to identify the system based on averaged measurements with periodic input signals. We also discuss pitfalls of using multitonal input signals in combination with the proposed scheme.

Proposes a protocol-based multiple aircraft conflict resolution for a finite information horizon, in which the communication range of an aircraft is finite. A protocol for multiple aircraft conflict resolution for an infinite information horizon is presented and then this protocol is extended to a finite information horizon problem using graph theory. Communication topology among aircraft is important for the finite information horizon problem and is modeled by a graph called the communication graph. We develop a protocol for multiple-aircraft conflict resolution based on the communication graph and show the safety of the protocol. Finally, we validate our protocol through simulations with a dynamic aircraft model.

The extended Kalman filter (EKF) is considered for parameter identification of Hammerstein/Wiener nonlinear systems. The EKFs for parameter identification of both combined Hammerstein/Wiener as well as for pure Hammerstein and pure Wiener models are formulated. In order to enable efficient estimation of unknown nonlinearities linear parametrizations with linear static mappings and basis function expansions are proposed and the EKFs for these cases are established. The efficiency and performance of the approach is demonstrated by means of a computer simulation of a Hammerstein and a Wiener model.

We investigate the efficient implementation of continuous time output feedback nonlinear model predictive control (NMPC) using a moving horizon observer (MHE) for state estimation. For this purpose we briefly review some aspects of MHE and some results on the computationally efficient implementation of nonlinear model predictive control. For the MHE we outline how an efficient dynamic optimization strategy for NMPC, the so called real-time iteration scheme, can be suitably adapted leading to a computationally efficient MHE implementation. The scheme is based on the multiple shooting method, a well known robust technique for nonlinear dynamic estimation problems. Application of the scheme allows us to use the MHE based output feedback NMPC controller in practice for reasonably sized problems. We exemplify the performance and computational demand of the proposed scheme by a simulation study considering the control of a high-purity distillation column.

In hierarchical systems, the upper levels in the hierarchy are obtained as a result of aggregation on the states and control variables of the bottom levels. A considerable interest in case Of the hierarchical systems is how the various system properties viz. controllability, observability and stability propagate from the bottom levels to the upper levels. In the present paper, we analyze the conditions under which such propagation is possible.

This paper considers a discrete-time linear time-invariant feedback control system where an unstable plant is actuated by a quantized controller and the measurements are disturbed by an additive wideband noise in the feedback loop. The quantized control may cause the discrete-time system to be modeled as variable structure under the existence and convergence conditions of a quasi-sliding mode. The control guides the system to the origin via a quasi-sliding mode, however, it may not constrain the system to the origin. Linear stabilizing controllers near the origin are investigated, and in the steady state the performance of the stabilized system is examined via a quadratic performance index. Numerical examples are presented and discussed as an illustration.