Research

1 - Evolving Model-Based Fuzzy Control

Unknown nonstationary processes require modeling and control design to be done in real time using streams of data collected from the process. The purpose is to stabilize the closed-loop system under changes of the operating conditions and process parameters. We have introduced evolving model-based fuzzy control approaches as a step toward the development of a general framework for online modeling and control of unknown nonstationary processes with no human intervention. Incremental learning algorithms have been proposed to develop and adapt the structure and parameters of the process model and controller based on information extracted from uncertain data streams. State feedback control laws and closed loop stability are obtained from the solution of relaxed linear matrix inequalities derived from different Lyapunov functions. Bounded control inputs, fuzzy granular data, and parallel distributed compensation have also been taken into account in the design of evolving control systems. Fuzzy granular computation provides a way to handle data uncertainty and facilitates incorporation of domain knowledge. Although evolving granular approaches (such as that shown in Fig. 1) are oriented to control systems whose dynamics is complex and unknown, in Fig. 2 we show the stabilization of the well-known Lorenz chaos. In general, evolving fuzzy systems can be used in model-based and model-free control schemes. In model-free evolving control, there are a variety of open questions related to missing data due to sensor faults (robust fault-tolerant control), outlier detection, feature selection, filtering of redundant data, and systematic methods for robustness analysis. We have attempted to develop a methodology and learning algorithms to handle practical problems found in complex and unknown dynamic systems.

2 - Data Stream Modeling and Weather Prediction

Observing past outcomes of a system to estimate its future behavior is the essence of forecasting and prediction. Time series prediction is based on the idea that the series carry the potential information needed to predict their future behavior. Analyzing data produced by actual phenomena can give good insights into the phenomena themselves and knowledge about the laws underlying the data. In particular, weather predictions (temperature, precipitation, cloudiness, wind direction, atmospheric pressure, snowfall) are useful for people to plan activities, protect property; and to assist decision making in many different sectors such as energy, transportation, aviation, agriculture, and inventory planning. Any system that is sensitive to the state of the atmosphere may benefit from weather forecasts.

We have developed computational methods and models to provide pointwise (singular) and guaranteed (granular) predictions of weather time series. Granular predictions together with singular predictions are important because they convey a real value and a range of possible values that can be linguistically labeled. Prediction bounds can be understood as optimistic and pessimistic values. Particularly, we have used the predictions to estimate energy demand and minimize risks. The energy industry has expressed a need for improved weather forecast accuracy, improved resolution of forecast information, and time and space data analysis for a better understanding of the uncertainty that is inherent in the forecasts.

4 - Early Detection of Parkinson's Disease

Parkinson’s disease is one of the most common age-related neurodegenerative disorders. Early diagnosis is a key to improve patients’ quality of life and to prolong it. Clinically, the disease is characterized by a series of motor features that include resting tremor, rigidity, bradykinesia, and gait impairment with postural instability. The hallmark pathologic features of the disease are highlighted by degeneration of neurons in the substantia nigra pars compacta coupled with Lewy bodies. Early diagnosis consists in identifying non-motor patterns of the disease.

Frequent non-motor symptoms of Parkinson’s disease include REM sleep behavior disorder, mood disturbances, autonomic dysfunction, olfactory dysfunction, and vocal impairment. Particularly, vocal degradation and impaired sense of smell have been pointed out as some of the earliest non-motor indicators of the disease which patients consider major barriers. We have customized our recent proposed methods on evolving artificial intelligence to deal with early recognition of the disease and with the identification of the severity and progress of the disease in known scales.