Research Areas

Multi-Agent Systems

A multi-agent system is made of many agents. An agent is a computer software program that is autonomous and situated in some distributed environments in order to meet its design objectives. Since the agents are faced with different environments, they are designed differently and properly for the given environment. Moreover, the agent is intelligent because it is reactive, proactive, social, flexible, and robust. In a large-scale distributed complex system, the agent’s autonomous and intelligent properties can reduce the complexity by reducing the coupling problems between the subsystems. Furthermore, the proactive, reactive, and robust properties can be well suited for applications in a dynamic and unreliable situation.

Intelligent Distributed Control of Power Plants

The objective of this research is development of an Intelligent Distributed Control System (IDCS) for a large-scale power plant, coupled with complex network of sensor/actuators. To operate a large-scale power plant, the monitoring and control systems are distributed and automated for each subsystem in the power plant. The approach is to use Multi-Agent Systems (MAS), which allows implementation of significantly more sophisticated measures to compensate for the unsecure and nonrobust properties plaguing traditional control systems.

Neural Networks Architectures

The proposed project will focus on an investigation of a mathematical approach to extrapolation, using a combination of system-type neural network architecture and the semigroup theory. The target of the investigation will be a class of distributed parameter systems for which, because of their complexity, lack an analytic description. Although the primary objective is extrapolation, this effort must begin with the development of an analytic description from the given empirical data, and then, proceed to extend that analytic description into an adjoining domain space for which there is neither data nor a model. That is, given a set of empirical data for which there is no analytic description, we first develop an analytic model and then extend that model along a single axis. Semigroup theory provides the basis for the neural network architecture, the neural network operation and also for the extrapolation process.

Research Projects

Multi-Agent Systems

Multi-Agent System A Multi-Agent System (MAS) can be defined as a loosely coupled network (organization) of problem solvers (agents), which interact to solve problems that are beyond the individual capabilities or knowledge of each problem solver (agent). An agent is a computer software program that is autonomous and situated in some distributed environments in order to meet its design objectives. Since the agents are faced with different environments, they are designed differently and properly for the given environment. Moreover, the agent is intelligent because it is reactive, proactive, social, flexible, and robust. In a large-scale distributed complex system, the agent’s autonomous and intelligent properties can reduce the complexity by reducing the coupling problems between the subsystems. Furthermore, the proactive, reactive, and robust properties can be well suited for applications in a dynamic and unreliable situation.
In order to perform the cooperative works in the MAS, it is presented to build multiple hierarchical structures for the multi-agent system organization. The organization has low level, middle level, and high level, and an agent in each level has a specific role in the society so that there is a conceptual idea of supervision for processing the tasks. The high level agents are the task delegation and interface agents, the middle level agents are the mediate and monitoring agents, and the low level agents are intelligent agents. The hierarchical structure that has three levels gives advantages for dynamic organization and autonomous systems. Moreover, the idea of multiple hierarchical structures is well suited for large-scale distributed systems. http://www.ecs.baylor.edu/ece/

Intelligent Distributed Control System for Power Plants

The objective of this research is development of an Intelligent Distributed Control System (IDCS) for a large-scale power plant. The complex network of sensor/actuators and distributed systems inherent in modern large-scale power plants immediately suggests a Multi-Agent System (MAS) as a viable solution. To operate a large-scale power plant, the monitoring and control systems are distributed and automated for each subsystem in the power plant. The approach is to use MAS, which allows implementation of significantly more sophisticated measures to compensate for the unsecure and nonrobust properties plaguing traditional control systems.
The intellectual merit of this research is that MAS is well suited for distributed control of power plants since the resulting cooperative and negotiated solutions provide the ability for distributed autonomy, robust control, and flexible auto configuration. Current power plants use only centralized or loosely decentralized control schemes and require continuous intervention by the operator, while MAS can manage the power plant operation by itself. The broader Impact of this research includes creation of a powerful unified tool for monitoring and control of power plants; an operational environment that is secure, fuel efficient, and provides practical and realistic infrastructure; and application in a broad range of engineering problems including nuclear plants, fuel cell plants, and power grids. With over 32.4% of current generation coming from coal alone, it is important to burn these fuels as efficiently and cleanly as possible. MAS paints a coherent picture for operators and will be useful as a training tool for workforces in energy systems. The MAS-IDCS facility will be utilized in developing courses in Power Systems Control and Computational Intelligence, and will be attractive in recruiting diversified groups of students.
http://www.ecs.baylor.edu/ece/

Development of System-Type Neural Network Architectures for Distributed Parameter Systems Using Algebraic Decomposition

As it stands now, there is no universal method of solving a general extrapolation problem. In fact, there is no consensus of opinion concerning whether the general extrapolation problem is a mathematically well-posed problem. It is a central contention of this research that if the extrapolation problem carries with it a group-like property, extrapolation is possible. It is a second central contention that, under these conditions, the key to extrapolation is to develop an extrapolating agent which acquires a group-like property of its own which becomes one-to-one linked with the group-like property of the object of extrapolation. The proposed neural network architecture consists of dual channels, one of which is the Function Channel and the other being the Semigroup Channel. The Function Channel spans a vector space and the Semigroup Channel selects vectors from within the space. The input is split such that the extrapolating variable becomes the input to the Semigroup Channel and all other input variables become the input to the Function Channel. The Function Channel consists of a parallel arrangement of n RBF networks, each one of which implements one orthonormal vector of an n-dimensional basis set of vectors. The outputs of the orthonormal vectors are linearly summed so that the channel spans an n-dimensional function space. The coefficient vector which determines the linear sum and thereby defines the specific function being implemented is supplied by the Semigroup Channel. The design of the Semigroup Channel is an adaptation of the Diagonal Neural Network (DRNN) or a recurrent architecture.
http://web.ecs.baylor.edu/faculty/lee/