Proceedings of the 2015 Chinese Intelligent Systems Conference

We discuss the robust, finite-time stabilization of fractional-order extended nonholonomic chained form systems for the first time in this article. By applying sliding mode variable structure theory and stability theorem of finite-time control, the three-step switching control scheme is proposed to deal with the presence of system uncertainties and external disturbance, so that the closed-loop system is finite time stable at the origin equilibrium point within any given settling time. Finally, simulation results show the effectiveness of the presented controller.

Electromagnetic band-gap structures are widely used in antenna and microwave element designs due to their unique electromagnetic features which are shown in two aspects: forbidden band gap and in-phase reflection. In this paper, three different antennas are designed and simulated by Ansoft HFSS to know the use of the EBG structure in microstrip antenna engineering. Radiation characteristics such as gain, radiation patterns and s-parameters of these microstrip antennas are performed. The simulated results verify that the gain has been increased noticeably, the radiation pattern has been improved and the sidelobe and backlobe levels have been reduced by using the mushroom-like EBG structures. Especially, the microstrip antenna over the EBG structures not only shows the best electromagnetic characteristics but also maintains its small dimensions.

In this paper, a terminal sliding mode (TSM) control method for the boost converter is studied. First of all, the nonlinear model of boost converter is established. Then a change of coordinate is employed such that the nonlinear model will be transformed into a linear system. On this basis, a TSM composite controller including the state feedback control and the disturbance feedforward compensation is given. Under the composite controller, the reference voltage can be tracked in a finite time. The effectiveness of the paper is shown with a simulation example.

To realize the precision of cross-section shaping, iterative learning control (ILC) is studied in the trajectory tracking strategy of roadheader cutting head in this paper. The dynamic model of roadheader cutting arm is established, and then an appropriate PD-type iterative learning controller is designed. The simulation results show that the tracking error decreases with the increasing number of iterations, and the tracking curve approaches to the desired target trajectory gradually. The control method is effective for trajectory tracking, which meets the requirement of accuracy in the actual site. Moreover, the method lays a theoretical foundation for automatic cross-section shaping research.

The containment problem of second-order multi-agent systems with varying delays is investigated. Supposing system topology is dynamic changed and jointly-connected, the control algorithm of double-integrator systems with multiple leaders is proposed. The stability of the control algorithm is analyzed on Lyapunov-Krasovskii method. Finally, a simulation example is provided to prove the effectiveness of the conclusion.

This paper presents a simple yet effective visual tracking method to attack the challenge when the target object undergoes partial or even full occlusion. First, a fixed number of image patches are sampled as the template set around current object location. In the detection stage, candidate image patches are sampled as the candidate set around the object location in the previous frame. Second, both the template set and candidate set patches are divided into sub-regions and features can be efficiently extracted via random projections. The confidence score for a specific candidate patch is computed through compressive features’ low-rank regulation with the template set patches. The lowest confidence score in the current frame indicates the new object location. The encouraging experimental results show that our proposed method outperforms several state-of-the-art algorithms, especially when the target object suffers partial or even full occlusion.

Coal moisture control process is a critical process in energy saving for pollution reduction and improving production efficiency and the quality of coke. The RBF artificial neural network approach for modeling is used to achieve precise control of coal moisture control system and against their strong coupling nonlinear systems with time-delay characteristics. The bionic BFO (Bacterial Foraging Optimization) is used to the fitness to optimize the RBF Neural network parameters. In order to achieve better results the RBF Neural network performance is optimized by these bionic BFO. This method provides a theoretical basis for accurate control of coal moisture process. The reduction of energy and pollution with improving the quality of coke is established.

Image super-resolution focuses on achieving the high-resolution version of single or multiple low-resolution images. In this paper, a novel super-resolution approach based on morphological component analysis (MCA) and dictionary learning is proposed in this paper. The approach can recover each hierarchical structure well for the reconstructed image. It is integrated mainly by the dictionary learning step and high-resolution image reconstruction step. In the first step, the high-resolution and low-resolution dictionary pairs are trained based on MCA and sparse representation. In the second step, the high-resolution image is reconstructed by the fusion between the high-resolution cartoon part and texture part. The cartoon is acquired by MCA from the interpolated source image. The texture is recovered by the dictionary pairs. Experiments show that the desired super-resolution results can be achieved by the approach based on MCA and dictionary learning.

Scene classification aims at grouping images into semantic categories. In this article, a new scene classification method is proposed. It consists of regularized auto-encoder-based feature learning step and SVM-based classification step. In the first step, the regularized auto-encoder, imposed with the maximum scatter difference (MSD) criterion and sparse constraint, is trained to extract features of the source images. In the second step, a multi-class SVM classifier is employed to classify those features. To evaluate the proposed approach, experiments based on 8-category sport events (LF data set) are conducted. Results prove that the introduced approach significantly improves the performance of the current popular scene classification methods.

To solve the problem that it is difficult to determine the initial location of the fruit fly in Fruit Fly Optimization Algorithm (FOA), an improved FOA, Adaptive Fruit Fly Optimization Algorithm (AFOA), is proposed in this paper. According to the ranges of variables to be optimized, AFOA can set the initial location of the fruit fly automatically and adjust the step value adaptively during iteration. Finally, the proposed algorithm is applied to Himmelblau’s non-linear optimization problem and time series prediction using Echo State Network (ESN). The experimental results imply that AFOA is effective and also show better ability in adaptation and optimization than traditional FOA, Particle Swarm Optimization (PSO) and Genetic Algorithm (GA).

To address the problems that the minimum squared error (MSE) algorithm is short of sufficient robustness and there are often not enough training samples for face recognition (FR), a mirror image-based robust MSE algorithm which uses mirror face as new training samples is proposed. The solution vector of MSE classification model needs to transform both original training samples and its mirror images into its class labels. Owing to mirror image reflect the change of pose and expression of original face images, the solution of classification model has some robustness. In addition, two classifier schemes are proposed and the second classifier scheme outperforms the first classifier scheme computationally. The experimental results on FERET, Extended YaleB and ORL databases indicate that the proposed approach achieves better robustness on images whose pose and expression are changed than the traditional MSE algorithm.

Flocking problem is one of the most critical problems of multi-agent systems. In this paper, a modified artificial potential function is proposed, which guarantees the stability of their inter-agent distances and ensures the smooth collision avoidance among neighboring agents. The prediction control algorithm is designed that can accelerate the convergence of speed for the second-order multi-agent systems. Mathematical analysis and simulation examples show the effectiveness of the theory.

In the bi-directional flow-rack (BFR) automated storage and retrieval systems (AS/RS), bins slope to opposite directions to make unit-loads be retrieved from half bins and be stored to the other half on the same working face. For random storage policy, a batching-greedy heuristic (BGH) has been proposed to generate dual-command (DC) operations in BFR AS/RS. In this paper, a novel DC operation generation rule specially designed for the BFR AS/RS is introduced to BGH, of which the effectiveness and efficiency are evaluated by simulation experiments.

All along, the immune system has been a hotspot and difficulty in the field of biological research. Traditional experimental immunology can observe the overall reaction of the immune system, but would be difficulty on the some details research, such as the recognition principle between antigen and antibody. In this paper, we use the binary string to express the gene of antigen and antibody, and use binary string matching to express the immune recognition process. Then, we would use the Agent computer model and the computer simulation method to study the microscopic properties of the immune system and some important details. The results of our study provide powerful basis to establish accurate perfect model of the immune system. Using the simulation model of immune responses to influenza virus, include the interactions between cells, some basic rules of the immune system are obtained.

In biological experiments, it has been impossible that we just use experimental apparatus to deal with the complex problems in immune cells. And the traditional mathematics and the physics model have some limitations, like lacking of microcosmic performance description of unit cells [1]. In this article, we do detail design after analyzing the requirements of the immune system. Then, combining with the related data of influenza virus, we use the Android platform application development to simulate the system. Android platform’s simple style of page, the application of interactive interface and the easy management can bring us different experiences. With the help of the computer program simulation, the experimental result is consistent with the model of immune response in the immune system.

In order to improve the accuracy and reliability of nonlinear state estimation problems with unknown but bounded noises in power system, an extend set-membership filter for state estimation in power systems is present in this article. The method is based on three sampling sine wave relational model. It overcomes the poor robustness, divergence and weak traceability of kalman filter, avoids the complex calculation process of traditional extend set-membership filter. Compared with kalman filter, the simulation results show that extend set-membership filter algorithm can track signals faster and more accurately.

In this paper, an anti-disturbance tracking control scheme is proposed for T-S fuzzy models subject to parametric uncertainties and unknown disturbances. Different with those previous results, exogenous disturbances are also described by T-S disturbance models. Under this framework, a composite observer is constructed to estimate the system state and the disturbances simultaneously. Meanwhile, by integrating the PI-type control algorithm with the estimates of the state and the disturbance, a feedback control input is designed to ensure the system stability and the convergence of the tracking error to zero as well as satisfactory disturbance estimation and attenuation performance.

Micro fluid dispensing technology is widely applied in electronics packaging, micro electromechanical system assembly, and biotechnology experiments, in which pl or nl amount of fluid materials (such as solder paste, adhesive, and DNA solution) are delivered controllably for the purpose of conducting, bonding, sealing, etc. This paper reviewed the latest developments as well as advantages and limits of three kinds of micro dispensing technology, which are needle nozzle type, integrated nozzle type and pin transfer type, classified according to the configuration of the nozzle unit. The measuring methods for the micro droplets are also briefly introduced in the article. Our work of dispensing less than 3 pl adhesive in an microassembly task is briefly introduced, and the trends and challenges of micro fluid dispensing are also discussed.

It is, as the state space explosion problem indicates, not uncommon that tremendous complexity and size of a system would annoyingly quiver the performance of discrete state-based modeling formalisms. The past few years, however, have inspiringly witnessed a brand new PEPA-based strategy offering a feasible solution against such disturbing puzzle. Via PEPA, a family of ordinary differential equations (ODEs) is figured out as continuous state space approximation. This paper establishes some significant properties for the fluid approximation of a PEPA model, including the existence, uniqueness, boundedness and convergence of the derived ODEs solution.

Focusing on the four-joint robotic fish, based on the fish body wave motion curve equation, deduced a mathematical model of four joint rotation of robotic fish. By setting two different parameters of amplitude coefficient and polarization coefficient, using simulation software MATLAB for different motion curves. For dynamics modeling of four-joint robotic fish provide theoretical basis. Hydrodynamic analysis is done too, the simulation software FLUNT is using for the underwater robotic fish prototype’s theoretical analysis. The simulation results have validated its effectiveness, reliability, and scalability and embodied out in the proposed prototype mechanical structure model.

In this study, the electroencephalogram (EEG) signals-analysis experiments were made to classify seizures patients. Principal component analysis (PCA) and kernel principal component analysis (KPCA) were used for the data compression with the (EEG) signals. Classifiers based on support vector machine (SVM)-PCA and SVM-KPCA were designed. The classification performances of four kinds of kernel function were also compared using the same dataset. The results showed that using SVM-KPCA had higher recognition performance than SVM-PCA. Experimental results showed that the algorithm using SVM-KPCA with Gaussian-kernel had better recognition performance than the other three methods.

In a networked control system, the time-delay has random and nonlinear characteristics, make the stability of system is hard to ensure. It need the controller in system can accurately predict time-delay. So the precise time-delay prediction of networked control system is an important factor in ensuring the stability of the control system. The echo state networks has good predictive ability on nonlinear time series, it is suitable for predict the time-delay. But parameters of echo state networks learning algorithm has a great influence on the prediction accuracy. An improved genetic algorithm is proposed for parameters optimization of echo state networks. The simulation results show that the prediction accuracy of the predictive method in this paper is higher than the conventional predictive model such as auto regressive and moving average (ARMA) model, least square support vector machine (LSSVM) model and Elman neural network.

With the rapid development of modern computer technology, the communication between man and machine is becoming more frequent. A large amount of data is needed to input when people use Microsoft Excel software in the fields of account, office, finance, hospital, etc. This paper presents a recognition method of naming speech as input for Excel table. In this system, we use the characteristics of the name as a basic speech unit by using the Mel Frequency Coefficients (MFCC) as the feature parameters. Moreover, we use the Hidden Markov model (HMM) as the basis to train the acoustic models of this environment. The HMM can ease the mismatch caused by the identification of the test environment and training environment, which can improve the recognition rate further. Finally, experiments show that this system has good recognition and input function. This study establishes the foundation for future development of the method of application system.

This paper discusses the decoupling control problem of manipulators and proposes a novel control method. This method does not need an accurate model of the robots and only needs input and output information of the robot. The system will be first transformed into a certain integral form by an extensive state observer (ESO) which is used to accurately estimate the systems various states and disturbances inside and outside. Then the decoupling control is investigated by using feed forward compensation. In the end, PD control with gravity compensation and active disturbance rejection control are provided to illustrate effectiveness of the controllers in the simulation.

Production process system is a dynamic process, so whether the dynamic process’ sensor is faulted or not is determined through the method of various sensor data acquisition and analysis. The double water tank data processing and fault diagnosis model was established according to the basic method of principal component analysis theory and its application research in the field of fault diagnosis. The test data was input into the model, so whether there was a failure was determined by comparing thresholds, and which sensor and what kind of fault are determined. The effectiveness was proved by the simulation result.

This paper aims to improve the steel liquid level control quality via iterative learning control (ILC) with extended error information. The ILC is one kind of type P iterative learning control, and besides the forgetting factor and the on-off switching action, error information was further extended on account of introduction of the just past and the second past cycles error signals. Results demonstrated that, the control quality can still be improved even under the model uncertainties, periodic bulging disturbances, the measuring noises, as well as the input signal error, the state error and the output error can be guaranteed to be ultimately bounded. Simulation results were provided to clarify the suggested idea further.

Quantization is an essential but often ignored part of the realization of compressive sampling (CS), and the analysis of quantization noise arise from CS is incomplete and not sufficient until now. The quantization noise is generated from quantizing CS values by a uniform quantizer under ideal and noise conditions. And also, the auto correlation function and power spectrum have been derived. It is concluded that the quantization noise is always uncorrelated with the input signals, the quantization noise is white and the spectrum is white noise spectrum. On this basis, we analyze the reconstruction error introduced by quantization noise quantitatively and give the upper and lower bounds of reconstruction error. Simulation results validate the validity of the analysis for further.

An attitude controller is proposed via employing backstepping control technique, and being represented by modified Rodriguez parameters. A general dynamic attitude model of satellites is deduced, along with a general model of actuator dynamics which can describe presumably all actuators for space application. External disturbances and actuator constraints are explicitly addressed. The control performance is proved in the numerical simulation experiences at last.

Distributed consensus problem is investigated for Lipschitz nonlinear multi-agent systems (MASs). Under the assumption that the states of the multiple agents are unmeasured, nonlinear observer for each agent is designed. Based on these observers, a distributed protocol is proposed, in which the coupling weights between adjacent agents are time-varying and can automatically change according to the designed adaptive law. Lyapunov-Krasovskii functional is constructed to analyses the consensus problem of the MASs under the proposed distributed adaptive protocol. By using free-weighting matrix approach, sufficient conditions that can ensure consensus are given. Finally, numerical example is presented to illustrate our result.

The asphalt gas concentration control system is characterized by its long-time delay, large inertia object and varying control parameter, too complicated to be applied in traditional PID control. In this paper, we use fuzzy-PID controller to control the asphalt gas concentration control system. The fuzzy-PID controller is designed based on fuzzy adaptive PID control principle; the PID parameters are adjusted according to error $$ e $$e and error change rate $$ ec $$ec. Fuzzy-PID controller has more excellent performances than the traditional PID controller in the gas concentration control from the simulation results comparisons.

Active disturbance rejection control is a unique control approach which could provide nice performance and need little knowledge of physical processes/plants. In order to analyze the stability of linear active disturbance rejection control (LADRC) by a direct and simple way, virtual equivalent system (VES) technique is adopted. By VES, global asymptotically stable with known process/plant dynamics and bounded input and bounded output stable with unknown process/plant dynamics are analyzed. The stability of LADRC for general single input single output nonlinear systems subject to dynamical and external uncertainties is analyzed from a brand-new viewpoint, which may be also helpful for stability analysis of other LADRC based system.

A single camera cannot provide a broad perspective and provide the details of monitoring targets at the same time. This paper design a dual camera system based on a wide-angle camera used to provide a wide-angle visual field and a close-up camera used to provide the details of moving targets. The real-time video data are firstly acquired based on DirectShow. Secondly, a Gaussian mixture model and Kalman filter are used to detect and track the moving targets, respectively. In order to realize the collaborative relationship among the dual camera, the conversion model among coordinate systems is established. Finally, the PID algorithm is used to driver the close-up camera to make the target locate in the center of its visual field. Experimental results demonstrate the effectiveness of the proposed method.

The inductances’ nonlinear characteristic in the assistant motor can affect the steady-state and dynamic performance of motor, therefore it also relates with assist torque of Electric Powering Steering (EPS) system closely. This paper constructs an accurately optimized model based on the theory of fractional calculus. This reference model takes the advantage of fractional orders which can improve robustness to the nonlinear of mechanical system. Based on the reference model, the fractional order controller $$ PI^{\lambda } D^{\mu } $$PIλDμ is designed. Several experiments show the proposed control strategy is effective and stable, furthermore the EPS system using the control strategy has a good anti-interference performance in the presence of nonlinear factors.

When facing with the mid-air interactive tasks in a wide range, the laser spot motion sensing technique can be as a information input mode of human-computer interaction(HCI). This paper explores the use of optical mouse sensors for building a laser spot tracking system, which can be used as a motion-based HCI device. Our work is focused on the characterization of the laser speckle sensing by optical mouse sensors. Based on the laser speckle displacement measurement capability of optical mouse sensors, we demonstrate that the low-cost optical mouse sensor can be used to record the motion of laser spot as a precise, fast and compact sensing method. To make a prototype system for demonstration, we describe a kind of deployment method for building optical mouse sensor array and propose a weighted method to fuse the raw data of multi-optical mouse sensor array. In experimental testing, the ISO standard tests for HCI input device were used for evaluating the efficiency of HCI input by optical mouse sensor-based laser spot tracking. A paradigm that using laser pointer and our laser spot tracking system to complete the dynamic hand gesture recognition task is also given in this paper.

This paper is devoted to the event-triggered consensus control for a general linear multi-agent network system with external disturbances. First a controlled output is defined and a model transformation is conducted to transform the consensus problem into an $$H_{\infty }$$H∞ problem. Then a distributed event-based controller is proposed so that the system can reach the consensus results by only using the agent’s own information and its neighbors’. The final conclusion is given in the form of a matrix inequality. Finally a simulation is introduced to verify the theoretical conclusions.

This paper considers an optimal consensus problem of second-order leader-follower multi-agent systems by using inverse optimality and LMI method. For a given control input, under the condition that the communication topology among followers is undirected connected, a positive definite matrix in a linear quadratic performance index function is obtained, which makes the linear quadratic performance index function to obtain the minimum value. Meanwhile, through theory analysis, we prove that the coefficient matrix of the given control input is the optimal feedback gain matrix. The simulation results show the effectiveness of our conclusions.

In this paper an activity recognition algorithm based on multi-feature fuzzy cluster is designed to find out more details of the activities so as to achieve an accurate classification among them. Firstly, it is proved that distribution of feature vectors vary from activity to activity. And then, a multi-feature extraction algorithm is designed to extract the feature vectors of each activity which makes up a standard activity class. Finally, an activity recognition algorithm based on similarity measurement is brought up and the misjudgment rate turns out to be acceptable, which proves that this algorithm is accurate and highly feasible.

Many methods of processing epileptic EEG signals are concentrated in the classification, and most of them use the wavelet transform and SVM classification algorithm. Although these algorithms acquire the high accuracy, it is still unable to provide a good explanation of quantitative difference and physical meaning between epileptic EEG and normal EEG. This paper presents a new hybrid algorithm (CWT-SVM-HMM) to classify epileptic EEG signal. By the results of classification of HMM, we can track back abnormal signal frequency sources, through the analysis of the sources of seizures during different frequency band, we can get a seizure of accurate quantitative analysis according to clinical feature of interictal epileptiform discharges.

Wind diesel hybrid system is an important new energy supply mode, but the output power of wind turbine generator (WTG) is fluctuated depending on weather conditions. Especially, the wind-diesel hybrid system with multiple WTGs for remote areas and islands operation, it will inevitably lead to the fluctuated output power to produce the large frequency deviation. So it’s crucial to coordinate the WTGs for providing high quality of electricity. Based on the designed disturbance observer, the coordination control strategy for the multi-WTGs wind-diesel hybrid system is proposed. The load variation is allocated to the WTGs output power reference by using the observer. Here, the constructed controller is compared with the traditional method for every WTG system with only PID control. The simulation results show that of frequency deviations are reduced and output power of every WTGs are controlled effectively.

At present, the application of 3D models is becoming more widely, which makes it very important and crucial to retrieve 3D model effectively. With the method based on content to search for those 3D models, rather than textual annotations, it is very important. For this purpose, this paper presented an effective sketch-based 3D model shape retrieval approach. The algorithm compares multi-view rendering of 3D models with the 2D sketch, and the feature vector is defined as a combination of global feature and local feature. We perform experiments, the results of which show a significant increase in precision for 3D model retrieval.

A new autonomous navigation scheme for planetary landing is presented. The navigation system contains an inertial measurement unit (IMU) and a stereo camera which can measure unit directional vectors and range information from the camera to detected landmarks. The lander’s motion is estimated by a algorithm known as vision-aided inertial navigation (VAIN). The algorithm uses the unit directional vectors and range measurements of features tracked in two sequential images and the lander’s corresponding poses derived from the IMU and it does not require any a priori terrain information. An augmented implicit extended Kalman filter (IEKF) tightly integrates measurements from the stereo camera and the IMU to produce an accurate estimation of the lander’s pose and velocity and to correct the IMU constant biases. The results of a numerical simulation show that the proposed VAIN method can vastly improve the navigation accuracy of the INS and satisfy the requirements of future planetary exploration missions.

For unidirectional communication topology of multi-agent systems, this paper studies its strong structural controllability. When the topology of agents is a pabud graph, we prove that the multi-agent systems can be strongly structurally controllable by selecting only one agent as leader. When the topology is partitioned to disjoint basic controllable graphs, the system can be strongly structurally controllable via selecting corresponding number of agents as leaders. A method to select leaders is presented to ensure the strong structural controllability of multi-agent systems. Finally, the effectiveness of the proposed method is verified with two examples.

This paper investigates the group consensus problem for a network consisting of Euler-Lagrange systems under directed topology with acyclic partition via neural network strategy. The neural network based controller achieves group consensus for uncertain networked Euler-Lagrange systems. By exploiting thoroughly the specific structure of the network topology, the stable analysis of the group consensus problem for such uncertain networked systems is also provided. Furthermore, a necessary and sufficient condition for ensuring that the systems reach group consensus is established. Finally, examples and simulations are given to show the effectiveness of the presented theoretical results.

Because the complicated non-linear relation between the coal gas outburst and its affecting factors, it is difficult to establish model with traditional mathematical method. The fuzzy system and the neural network were organically combined to establish evaluation strategy of coalmine gas outburst risk level based on fuzzy neural network. This paper made use of fuzzy mathematics to express and deal with the imprecise data and fuzzy information, and utilized self-adaptive neural networks system to solve the problems. Simulation results show that the model is reliable and precise and outburst risk level can be accurately predicted with proposed method and the mean error is small.

In this paper, a new adaptive neural dynamic surface control technique with an improved prescribed performance function is proposed for the nonlinear two-inertia system. An improved error transformation function is used to ensure the prescribed output tracking performance, and the neural network (NN) is utilized to estimate the unknown disturbance. The dynamic surface technique simplifies the controller by introducing first-order filters to eliminate “explosion of complexity” inherent in backstepping approach. Simulation results demonstrate the control scheme is effective.

In this paper, a new neural network observer based optimal tracking control is presented to attenuate the effect of backlash and other uncertainty for the position tracking of multi-motor servomechanism (MMS). By adopting a continuously differentiable function instead of the non-differential dead-zone model of the backlash, the state space representation of MMS is set up by using a linear part of the differentiable function. Based on the state space representation, the optimal neural network (NN) observer is used to estimate the uncertainties and unmeasured states, which combines with the optimal state feedback to synthesis the actual control law. Finally, Lyapunov theory is utilized to certify that the tracking error, the observed error and neural network weights are all semi-globally uniformly ultimately bounded (SGUUB). Simulation results validate the effectiveness of this method.

The applications of fractal have been involved in a wide range of areas. As a fractal brunch—fractional Brownian motion (fBm) has also been used and applied in numerous physical sciences. The fBm technique overcome the traditional non-memory random walk-Brownian motion, it could be used on modelling particle tracking dispersion in ocean surface. The technique gives more accurate simulation on tracking particle diffusion and dispersion. Here we use a developed fBm model (FBMINC) as a diffusion process, simulate an idealized coastal bay surface trajectories and particle cloud dispersions using fBm particle tracking techniques. Compared to the traditional Brownian motion particle tracking model, the newly developed fBm particle tracking model produces patterns more close to the reality. Due to its flexibility, the fBm particle tracking model can be widely used in pollutant dispersion on difference size of water bodies.