Randomized algorithms for synthesis of switching rules for multimodal systems

In this paper, we consider the design of globally asymptotically stabilizing state-dependent switching rules for multimodal systems, first restricting attention to linear time-invariant (LTI) systems with only two states for the switch, and then generalizing the results to multimodal LTI systems and to nonlinear systems. In all cases, the systems considered do not allow the construction of a single quadratic Lyapunov function and, hence, fall in the class of problems that require multiple Lyapunov functions and thus are nonconvex. To address the challenge of nonconvexity , we introduce probabilistic algorithms, and prove their probability-one convergence under a new notion of convergence. Then, to reduce complexity, we develop modified versions of the algorithm. We also present a class of more general nonconvex problems to which this approach can be applied. The results are illustrated using two- and three-dimensional systems with multiple switch states.     

Acrobot control by learning the switching of multiple controllers

Reinforcement learning (RL) has been applied to constructing controllers for nonlinear systems in recent years. Since RL methods do not require an exact dynamics model of the controlled object, they have a higher flexibility and potential for adaptation to uncertain or nonstationary environments than methods based on traditional control theory. If the target system has a continuous state space whose dynamic characteristics are nonlinear, however, RL methods often suffer from unstable learning processes. For this reason, it is difficult to apply RL methods to control tasks in the real world. In order to overcome the disadvantage of RL methods, we propose an RL scheme combining multiple controllers, each of which is constructed based on traditional control theory. We then apply it to a swinging-up and stabilizing task of an acrobot with a limited torque, which is a typical but difficult task in the field of nonlinear control theory. Our simulation result showed that our method was able to realize stable learning and to achieve fairly good control.This work was presented, in part, at the 9th International Symposium on Artificial Life and Robotics, Oita, Japan, January 28–30, 2004     

Address‐While‐Display technique combined with gas‐discharge AND logic for driving AC‐PDPs

A hybrid AWD/AND drive technique has been developed in which an Address‐While‐Display (AWD) scheme is combined with an AND logic characteristic that gas discharges demonstrate. The AWD technique enables AC‐PDPs to be driven at high luminance, while the AND logic reduces the number of scan drivers by an order of magnitude. A detailed analysis of the addressing operation has been made. The hybrid drive utilizes the AND logic in two ways: (1) a combination of two voltage pulses and (2) a combination of a voltage pulse and discharge‐priming particles. It was found that the addressing operation requires the establishment of a discharge between the scan and data electrodes, and also between the scan and display electrodes.     

On-line EM algorithm for the normalized gaussian network

A normalized gaussian network (NGnet) (Moody & Darken, 1989) is a network of local linear regression units. The model softly partitions the input space by normalized gaussian functions, and each local unit linearly approximates the output within the partition. In this article, we propose a new on-line EMalgorithm for the NGnet, which is derived from the batch EMalgorithm (Xu, Jordan, &Hinton 1995), by introducing a discount factor. We show that the on-line EM algorithm is equivalent to the batch EM algorithm if a specific scheduling of the discount factor is employed. In addition, we show that the on-line EM algorithm can be considered as a stochastic approximation method to find the maximum likelihood estimator. A new regularization method is proposed in order to deal with a singular input distribution. In order to manage dynamic environments, where the input-output distribution of data changes over time, unit manipulation mechanisms such as unit production, unit deletion, and unit division are also introduced based on probabilistic interpretation. Experimental results show that our approach is suitable for function approximation problems in dynamic environments. We also apply our on-line EM algorithm to robot dynamics problems and compare our algorithm with the mixtures-of-experts family.     

Fatigue crack propagation in copper and Cu−Al single crystals

The effect of crystallographic orientation, temperature, and stacking fault energy on the rate of fatigue crack propagation was studied in polycrystalline copper, copper single crystals, and Cu−Al single crystals at room temperature and at liquid nitrogen temperature. A stress intensity factor was used to normalize the crack propagation data. It was found that dislocation cross slip plays a critical role on the rate of fatigue crack propagation. Existing mathematical crack propagation formulae could not explain the data on single crystals. A new fatigue crack propagation model to explain the observed results is proposed. H. ISHII, formerly Research Assistant at Materials Science Department, Northwestern University, Evanston, Ill. This paper is based on a thesis submitted by H. ISHII in partial fulfillment of the requirements of the degree of Doctor of Philosophy at Northwestern University.     

Absolute linearity measurements on HgCdTe detectors in the infrared region

The nonlinearity characteristics of photoconductive and photovoltaic HgCdTe detectors were experimentally investigated in the infrared wavelength region by use of the National Physical Laboratory detector linearity measurement facility. The nonlinearity of photoconductive HgCdTe detectors was shown to be a function of irradiance rather than the total radiant power incident on the detector. Photoconductive HgCdTe detectors supplied by different vendors were shown to have similar linearity characteristics for wavelengths around 10 microm. However, the nonlinearity of response of a photovoltaic HgCdTe detector was shown to be significantly lower than the corresponding value for photoconductive HgCdTe detectors at the same wavelength.     

The application of the simple fuel pin transient and melting model to thermal-hydraulics in LMFBR subassembly

A simple conduction model with phase change has been developed for the transient analysis of a reactor fuel pin based on lumped parameter techniques. The purpose of this analysis is to provide a simple useful tool to obtain the general information about fuel and clad leaning into the cooling transients and melting. Such a simple fuel and clad thermal transient model is particularly useful to multichannel analysis where conventional conduction computer codes require considerable computing time and storage space. At the present, this formulation is being employed for the analysis of sodium thermohydraulics, sodium voiding, and melting of cladding and fuel in a subassembly of a fast reactor core. A detailed analysis of the predicted coolant, fuel and cladding thermal transients leading into sodium voiding and fuel pin melting has been made in comparison with the results of various in-pile experiments and with the predictions from the existing more complicated codes.