Xiaoming Xu

Design neural networks based fuzzy logic

Abstract The integration of fuzzy logic and neural networks has given birth to an emerging technology field, fuzzy neural networks. The fuzzy neural networks have the potential to capture the benefits of the two fascinating fields into a single capsule. In this paper three approaches of implementing and constructing the fuzzy neural networks, neural networks based fuzzy logic, have been discussed and an improved genetic algorithm, which is a special neural networks learning algorithm, is proposed.

Urban traffic multi-agent system based on RMM and Bayesian learning

Addresses multi-agent coordination in urban traffic control to coordinate the signals of adjacent intersections for minimizing the waiting car queue in the urban traffic network. For the purpose of this case study, we adopt a multi-agent coordination, which uses the recursive modeling method (RMM) that enables an agent to select his rational action by examining with other agents by modeling their decision making in a distributed multi-agent environment. Bayesian learning is used in conjunction with RMM for belief update. As a result, an agent can determine which models of the other agents are correct, and keep his knowledge up to date. We describe how decision making using RMM and Bayesian learning is applied to the urban traffic control domain to settle a multi-agent traffic control system and show experimental results.

Urban Intelligent Traffic System Based on Multi-Agent

This paper describes multi-agent coordination in urban traffic control, focusing on coordinating the signal of adjacent intersections in order to minimize the waiting vehicles queue in the network. With multi-agent coordination, the Recursive Modeling Method (RMM) is used to select rational actions by examining the decision of other agents in making a distributed multi-agent environment. The paper describes how decision making using RMM and Bayesian learning can be applied to urban traffic control.

Analytical design for linear continuous integrator/time delay systems

Abstract Smith predictor has been proved to be a useful scheme for stable processes with time delay. In this paper an analytical design method of Smith predictor is presented for integrator/time delay processes based on the theory of complex function. The design procedure incorporates a two-stage procedure. In the first stage, the controller that stabilizes the closed loop system is parameterized. In the second stage, high enough order Pade approximation is introduced to approximate the time delay, and convert the design problem of system with time delay to that of system without time delay. The optimal Smith predictor is then derived analytically. Examples are given to illustrate the method.

Drying process modelling based on improved backpropagation algorithm

Abstract An improved backpropagation algorithm to simultaneously determine the optimal learning rate and momentum coefficient is applied to model the impingement stream drying process. Simulation results show that for those new-style high-efficient dryers which can not be modelled based on the motion as well as heat and mass transfer of products during drying, the neural network model given in this paper is capable of predicting the behavior of the dryers exactly and rapidly.

HEURISTIC DECENTRALIZED CONTROL OF MULTI ARMS COORDINATED SYSTEMS

Abstract Controlling multi robot arms systems is very difficult due to the complexity and nonlinearity. In this paper, we adopt the team decision model to decompose the original problem into N (the number of arms) smaller one. For the information insufficiency due to decentralized control structure, heuristic control strategy is used and a suboptimal control is achieved in sense of heuristic optimization. A simple example for simulation study is included to illustrated the applicability.