Thaddeus T. Shannon

Intelligent supply chain management using adaptive critic learning

A set of neural networks is employed to develop control policies that are better than fixed, theoretically optimal policies, when applied to a combined physical inventory and distribution system in a nonstationary demand environment. Specifically, we show that model-based adaptive critic approximate dynamic programming techniques can be used with systems characterized by discrete valued states and controls. The control policies embodied by the trained neural networks outperformed the best, fixed policies (found by either linear programming or genetic algorithms) in a high-penalty cost environment with time-varying demand.

Adaptive critic based approximate dynamic programming for tuning fuzzy controllers

We show the applicability of the dual heuristic programming (DHP) method of approximate dynamic programming to parameter tuning of a fuzzy control system. DHP and related techniques have been developed in the neurocontrol context but can be equally productive when used with fuzzy controllers or neuro-fuzzy hybrids. We demonstrate this technique on a highly nonlinear 2nd order plant proposed by Sanner and Slotine (1992). Throughout our example application, we take advantage of the Takagi-Sugeno model framework to initialize our tunable parameters with reasonable problem specific values, a practice difficult to perform when applying DHP to neurocontrol.

Application considerations for the DHP methodology

Six questions are posed that are likely to be of interest to one considering applying the dual heuristic programming (DHP) methodology. Tentative answers to these questions are provided in the last section of the paper. As a basis for these, the paper first describes observations made from a series of explorations into various aspects of using the DHP method to design the controller for the pole-cart benchmark problem. Parameters of the explorations included training gains, plant parameter variations, fidelity of plant model, controller sampling rates, network architectures, training strategies, and generalization tests.

Qualitative models for adaptive critic neurocontrol

We demonstrate the use of qualitative models in the dual heuristic programming (DHP) method of training neurocontrollers. Two fuzzy approaches to developing qualitative models are explored: a priori application of problem specific knowledge, and estimation of a first order TSK fuzzy model. These approaches are demonstrated respectively on the cart-pole system and a nonlinear multiple-input-multiple-output plant proposed by Narendra. In both cases we find that a simplified model based on a Fuzzy framework enables better performance to be obtained as compared to use of non-fuzzy models of equivalent complexity. In both cases we use models that, while poor as one-step predictors, achieve effectiveness in the DHP training context equivalent to that of exact analytic models.

Defect screening using independent component analysis on I/sub DDQ/

An I/sub DDQ/ Statistical Post-Processing/spl trade/ (SPP) outlier screen is presented based on the computation of statistically independent sources of variation in the I/sub DDQ/ measurements. I/sub DDQ/ measurements from die passing all other tests are modeled using sources of variation extracted by independent component analysis (ICA). Outliers are separated from the sample population based on residuals computed using these sources and a nearest neighbor spatial signature. An algorithm is presented for applying the proposed technique in production. The screen is demonstrated with 0.18/spl mu/m and 0.11/spl mu/m volume data and shown to effectively identify the outliers at the 0.1 /spl mu/m technology node.

A comparison of training algorithms for DHP adaptive critic neurocontrol

A variety of alternate training strategies for implementing the dual heuristic programming (DHP) method of approximate dynamic programming in the neurocontrol context are explored. The DHP method of controller training has been successfully demonstrated by a number of authors on a variety of control problems in recent years, but no unified view of the implementation details of the method has yet emerged. A number of options are described for sequencing the training of the controller and critic networks in DHP implementations. Results are given about their relative efficiency and the quality of the resulting controllers for two benchmark control problems.

Dual heuristic programming for fuzzy control

Overview material for the special session (tuning fuzzy controllers using adaptive critic based approximate dynamic programming) is provided. The dual heuristic programming (DHP) method of approximate dynamic programming is described and used to design a fuzzy control system. DHP and related techniques have been developed in the neurocontrol context but can be equally productive when used with fuzzy controllers or neuro-fuzzy hybrids. This technique is demonstrated by designing a temperature controller for a simple water bath system. In this example, we take advantage of the TSK model framework to initialize the tunable parameters of our plant model with reasonable problem specific values.

Partial, noisy and qualitative models for adaptive critic based neurocontrol

The roles of plant models in adaptive critic methods for approximate dynamic programming are considered, with primary focus given to the dynamic heuristic programming (DHP) methodology. For complete system identification, partial, approximate, and qualitative models of plant dynamics are considered. Such models are found to be sufficient for successful controller design. As classification is in general easier than regression, the results for qualitative models suggest an avenue for simplifying ongoing system identification in adaptive control applications.

Adaptive critic based design of a fuzzy motor speed controller

We show the applicability of the dual heuristic programming (DHP) method of approximate dynamic programming to the design of a fuzzy control system. The DHP and related techniques have been developed in the neurocontrol context but can be equally productive when used with fuzzy controllers or neuro-fuzzy hybrids. We demonstrate this technique on a speed controller for a brushless motor. In our example, we take advantage of the Takagi-Sugeno model framework to initialize the tunable parameters of our plant model with reasonable problem specific values, a practice difficult to perform when applying DHP to neurocontrol.

Modeling respiration from blood pressure waveform signals: an independent component approach

We describe a method for extracting the additive effect of respiration from arterial blood pressure and central venous pressure waveform signals. Our method estimates a finite impulse response (FIR) separating filter using an independent component approach, analogous to minimizing the coherence of the separated components. We compare the extracted respiration component with the impedance based respiration signal and with a respiration estimate extracted using an LMS-optimal bandpass filter of comparable order.