Showing papers by "Zeng-Guang Hou published in 2010"

Foundation and Classification of Nonconventional Neural Units and Paradigm of Nonsynaptic Neural Interaction

Abstract: This chapter introduces basic types of nonconventional neural units and focuses their mathematical notation and classification. Namely, the notation and classification of higher-order nonlinear neural units, time-delay dynamic neural units, and time-delay higher-order nonlinear neural units is introduced. The classification of nonconventional neural units is founded first according to nonlinearity of aggregating function, second according to the dynamic order, third according to time-delay implementation within neural units. Introduction into the simplified parallel of the higher-order nonlinear aggregating function of higher-order neural units revealing both the synaptic and nonsynaptic neural interaction is made; thus, a new parallel between the mathematical notation of nonconventional neural units and the neural signal processing of biological neurons and is drawn. Based on the mathematical notation of neural input inter-correlations of higher-order neural units, it is shown that higher-order polynomial aggregating DOI: 10.4018/978-1-60566-902-1.ch027

Compressive sensing approach based mapping and localization for mobile robot in an indoor wireless sensor network

Abstract: Reliable navigation in the Wireless Sensor Networks (WSNs) always require mobile robot do localization and environment map building processes, which depends heavily on estimating the position of the features within the entire surroundings, that means as a sensor receiving platform, the robot needs to detect and process information transmitted from sensors as much as possible, in order to perform tasks. However, some large and complex deployed wireless sensor network environments, in which sensor information are relatively sparse compared with the number of sensor sources, usually make the robot hard to receive enough crucial information. To make robot know its position and construct the environment map with minimal sensing information. We propose a novel navigation algorithm based on RF wireless sensor networks to simultaneous localization and mapping (SLAM) approach, thus, a new framework that allows a team of robots to build a map of the parameter of interest with a small number of measurements is presented. By using the recent results in the area of compressive sensing, we show how the robots can build a map with limited number of sensing measurements. The proposed algorithm is conceptually simple and easy to implement. Simulation and experimental results show that good result can be achieved using the proposed method.

Higher Order Neural Networks: Fundamental Theory and Applications

Abstract: The human brain has more than 10 billion neurons, which have complicated interconnections, and these neurons constitute a large-scale signal processing and memory network. The mathematical study of a single neural model and its various extensions is the first step in the design of a complex neural network for solving a variety of problems in the fields of signal processing, pattern recognition, control of complex processes, neurovision systems, and other decision making processes. Neural network solutions for these problems can be directly used for computer science and engineering applications. ABSTRACT

System and method for robot trajectory generation with continuous accelerations

Abstract: A method for robot trajectory generation with continuous acceleration, Receiving a user's motion command through a motion command interface, and sending the user's motion command to Cartesian trajectory generator; Converting the user's command to a trajectory path points of robot end effector in Cartesian space; Transforming the trajectory path points of robot end effector in Cartesian space into a robot trajectory path points in a joint space; Calculating positions, velocities and accelerations of robot joints in each motion servo cycle; Comparing the positions, velocities and accelerations of the robot joints generated by a joint Trajectory Interpolator with a velocity's limit value and an acceleration's limit value of each robot joint stored in a robot parameter database respectively.

Adaptive Neural Network Control of FES Cycling

Abstract: FES cycling is a safe and easy way for the rehabilitation of spinal cord injury (SCI) patients. In order to design an control system for FES cycling, this paper presents a control strategy of the cycling induced by FES. The control system is developed based on artificial neural networks and consists of two layers: the outer layer controls the FES cycling model dynamics and generates desired torque; the inner layer controls multi-muscle to generate the torque that tracks the desired torque. And the distribution of multi-channel FES stimulation intensities is optimized based on the energy and muscle fatigue minimization principles. The simulation results show that the control system designed in this paper is stable and robust to muscle fatigue. Finally, some remarks are given on the clinical experiments of this control strategy.

Tracking ground targets using an autonomous helicopter with a vision system

Abstract: An autonomous helicopter with a vision system onboard to track ground targets can accomplish very important tasks like inspection and surveillance. The autonomous helicopter target tracking system is a interdisciplinary complex system. A typical autonomous helicopter target tracking system using vision sensors usually includes five major component: a navigation system, a flight controller, a active vision system, a target state estimator and a helicopter guidance law. In this paper, we focus on target state estimator and helicopter guidance law. It is assumed that the target moves on a flat ground, so the relative position between the target and the helicopter can be measured by a monocular vision system. Kalman filter is used to estimate the state of the target based on the vision measurements. A new guidance law inspired by velocity pursuit guidance and car-following control is proposed to guide the helicopter to follow the moving target. The tracking system is tested in a virtual reality environment and the results shows that the performance of the system is satisfied.

An adaptive RBF neural network control strategy for lower limb rehabilitation robot

Abstract: This paper proposed an adaptive control strategy based on RBF (radial basis function) neural network and PD Computed-Torque algorithmfor precise tracking of a predefined trajectory. This control strategy can not only give a small tracking error, but also have a good robustness to themodeling errors of the robot dynamics equation and also to the system friction. With this control algorithm, the robot can work in assist-as-needed mode by detecting the human active joint torque. At last, a simulation result using matlab simulink is given to illustrate the effectiveness of our control strategy.

Multiple kernel learning with ICA: Local discriminative image descriptors for recognition

Abstract: Local image features have been proven to be a powerful way to describe pattern of interest, both from single objects and complex scenes. While learning from images represented by local features is challenging, recent publications and developments in object recognition has shown that significant performance achievements can be achieved by carefully combining multi-level, coarse-to-fine, sparsely distributed feature encodings [10], and kernel based learning methods, which defines a generalized similarity measure among data using multiple kernel functions instead of a single one, also known as multiple kernel learning (MKL). In this paper we show that the Kernel ICA descriptors based MKL supervised learning approach perform better than other descriptors for object recognition, since the ICA-based representation is localized. In low-level feature extraction, ICA produces independent image bases that emphasize edge information in the image data. In high-level classification, MKL classifies the ICA features as discriminative components. We demonstrate our algorithm on different databases for recognition tasks, showing that the proposed method is accurate and more efficient than current approaches.

Adaptive neural network control of a 5 DOF robot manipulator

Abstract: In this paper, a robust neural network based controller is proposed to steer the joint angles of rigid-link robot manipulators to track the desired trajectories asymptotically. The developed control scheme makes use of a two-layer neural network to learn the behaviors of unknown dynamics of robot. Both the estimation error and external disturbances can be effectively counteracted by employing smooth robust compensators. It is proved that arbitrarily small tracking errors could be achieved by selecting proper design parameters. In the control method derived here, there is no preliminary off-line learning phase, which is time-consuming, for better estimation of unknown nonlinear smooth function. The input weights are chosen randomly in [−1, 1] and they are fixed in the whole simulation, and the adjustable output weights of neural network are simply initialized to be zero. The weight tuning algorithm for tunable parameters guarantees both closed loop stability and bounded weights. In the simulation, besides MATLAB, a famous multi-body dynamics analysis software in the world called ADAMS is employed. The combined simulation of ADAMS and MATLAB is able to produce realistic results of the closed loop system behaviors. The co-simulation results validate the effectiveness of the proposed approach.

Dynamic programming field based environment learning and path planning for mobile robots

Abstract: a new approach called dynamic programming field for modeling the robot environments is presented and it's beneficial to the path planning. The dynamic programming field, which is approximated by Neuro-Dynamic Programming, records environmental information through a neural network and can be used to compute the approximate optimal cost between any two points. Based on the dynamic programming field, the path planning combines the optimization of global planning and the real-time of local planning. Simulation results show the validity of the proposed approach.

Topological localisation based on monocular vision and unsupervised learning

Abstract: Global localisation is a very fundamental and challenging problem in robotics. This paper presents a new method for mobile robots to recognise scenes with the use of a single camera and natural landmarks. In a learning step, the robot is manually guided on a path. A video sequence is acquired with a font-looking camera. To reduce the perceptual alias of features easily confused, we propose a modified visual feature descriptor which combines colour information and local structure. A location features vocabulary model is built for each individual location by an unsupervised learning algorithm. In the course of travelling, the robot uses each detected interest point to vote for the most likely location. In the case of perceptual aliasing caused by dynamic change or visual similarity, a Bayesian filter is used to increase the robustness of location recognition. Experiments are conducted to prove that application of the proposed feature can largely reduce wrong matches and performance of proposed method is reliable.

Local vs. global: Indoor multi-robot simultaneous localization and mapping in wireless sensor networks

Abstract: The multi-robot SLAM or cooperative SLAM (CSLAM) by multiple robots, both local based and global based, in a RF wireless sensor network, is investigated in detail in this paper. We propose a new multi-robot SLAM approach, the localization is done by using trilateration localization method, and we use cooperative compressive sensing to build map with sparse information acquired by robots. We make an evaluation between the current baseline algorithms and our proposed method. The comparison result implies that particle filter is more robust while our approach could be more feasible dealing with the environment where either little prior knowledge about the sensed field is available.

Image category learning and classification via optimal linear combination of multiple partially matching kernels

Abstract: Multiple kernel learning (MKL) aims at simultaneously optimizing kernel weights while training the support vector machine (SVM) to get satisfactory classification or regression results. Recent publications and developments based on SVM have shown that by using MKL one can enhance interpretability of the decision function and improve classifier performance, which motivates researchers to explore the use of homogeneous model obtained as linear combination of various types of kernels. In this paper, we show that MKL problems can be solved efficiently by modified projection gradient method and applied for image categorization and object detection. The kernel is defined as a linear combination of feature histogram function that can measure the degree of similarity of partial correspondence between feature sets for discriminative classification, which allows recognition robust to within-class variation, pose changes, and articulation. We evaluate our proposed framework on the ETH-80 dataset for several multi-level image encodings for supervised and unsupervised object recognition and report competitive results.

A design of multi-axis motion controller for welding robot based on DSP

Abstract: To satisfy the motion control for wielding robot, a design of multi-axis motion controller based on DSP is proposed in this paper. The hardware structure and the software structure are discussed in detail. The motion controller consists of three DSP control cards. There are two control units which can control two servo motors in one DSP control card. So the controller can control six servo motors at the same time. The motion controller can perform high accuracy and high velocity interpolation. It communicates with the upper computer through dual-port RAM. More axis controlling can be achieved by adding more DSP control cards with bus card. The absolute position is detected by motion controller. A variety of control mode can be achieved includes position control, velocity control and torque control. The interpolation algorithm is presented. The experiment data indicates that the motion controller has the advantages of good real-time performance and highly machining accuracy.

Model based control of a rehabilitation robot for lower extremities

Abstract: This paper mainly focuses on the trajectory tracking control of a lower extremity rehabilitation robot during passive training process of patients. Firstly, a mathematical model of the rehabilitation robot is introduced by using Lagrangian analysis. Then, a model based computed-torque control scheme is designed to control the constrained four-link robot (with patient's foot fixed on robot's end-effector) to track a predefined trajectory. Simulation results are provided to illustrate the effectiveness of the proposed model based computed-torque algorithm. In the simulation, a multi-body dynamics and motion software named ADAMS is used. The combined simulation of ADAMS and MATLAB is able to produce more realistic results of this complex integrated system.

A simulation study of an FES based rehabilitation control system

Abstract: This paper studies the use of functional electrical stimulation (FES) method with movement in patients with paraplegia (SCI) for rehabilitation. When FES is applied to the muscles in patients with certain frequency and amplitude of electrical stimulation, the stimulated muscles generate contraction strength. The use of FES can effectively prevent muscle atrophy in patients with paraplegia, and produce good rehabilitation results. This paper is focused on the use of multi-channel FES to lower extremity and the control of multi-muscle to generate rehabilitation movements. Because of the complexity of human motion, this paper has established a three-dimensional musculoskeletal model of the human body. Based on this model, the control of multi-channel FES to achieve cycling has been simulated. The controller designed in this paper is divided into two layers: the outer layer is based on fuzzy control method and it generates desired torque which is needed for cycling movement; the inner layer is a composite controller based on feedforward and PID control, and this layer control of multi-channel FES stimulate muscles to produce desired torque for tracking purposes. The controller of the inner layer uses tracking differentiator to obtain derivative information for the control system. Finally, the simulation results provided shows the effectiveness of the proposed method.

Formation control for multiple mobile robots based on the Spiking Neural Network

Abstract: In this paper, a Spiking Neural Network (SNN) based controller is designed to fulfill the task of formation control of multiple mobile robots. The neural network contains three layers with different neuron model for different layer: the input layer encodes the inputs including sensor and task-related information by leaky integrate-and-fire (LIF) neurons, the hidden layer uses the approximate coincidence detection coding to fuse the information from the input layer and the spike response model (SRM model) is applied to the output layer to fire spikes to drive the motors. By the leader-follower strategy and the SNN controller, the multiple mobile robots system can realize the formation control. The validity of this controller is testified by the simulations.

Correlation detection with firing rate estimation based on temporal coincidence coding

Abstract: In this paper, the firing rate of the neuron based on temporal coincidence coding is estimated for correlation detection. Two cases are considered: the independent inputs are stochastic spike trains that are modeled by the homogeneous Poisson process or the renew process. The situation that the inputs are correlated is also considered and the conditions for the neuron to detect the correlation among inputs are discussed. The results are demonstrated by the simulations.

A motion controller for a pan-tilt camera on an autonomous helicopter

Abstract: In this paper, a motion controller for a pan-tilt camera mounted on an autonomous helicopter is presented. The motion planner is designed according to the kinematics of the pan-tilt camera. However, the solution of the inverse kinematics of the pan-tilt camera is not unique. To deal with this situation, a decision maker is designed. The decision maker makes use of a cost function to decide which solution should be chose. And the error signal is defined as the difference between pan-tilt joint angles and the desired joint angles. The dynamics of the error system is derived, and proportional controllers are designed according to the error dynamics to control the pan and tilt joints respectively. This system is validated in a virtual reality environment, and the simulation results show that this system can track the target successfully.

Design of an H ∞ loop shaping velocity controller for a small unmanned helicopter

Abstract: In this paper, a velocity controller based on H ∞ loop shaping for an unmanned helicopter is presented. H ∞ loop shaping controller guaranteed the stability margin of the closed-loop system above desired level. The performance of the controller is improved using gain-scheduling of key parameters. The simulation results show that the controller is robust and has good performance.