Sam Chau Duong

On the Continuous Control of the Acrobot via Computational Intelligence

The focus of this work is the continuous control of the Acrobot under limited-torque condition. By utilizing neural network (NN) and genetic algorithm (GA), a global controller is constructed in order to handle both swing-up and balancing control stages of the Acrobot without the need of different control strategies for the two processes. Based on given control timings, two different evaluation functions are introduced, one being continuous evaluation and the other multi-point based evaluation. In order to improve the system performance, an enhanced GA is proposed which recovers the diversity of population when it tends to be lost by applying an adaptive mutation operator based on a convergence index that reflects the diversity of population in GA. To verify the system performance, numerical simulations are implemented with different timing constraints. Comparisons between the proposed GA with the conventional method as well as between the two evaluation schemes are also provided. Simulation results show that the proposed GA has good performance and the neurocontrol system is able to control the Acrobot effectively by either one of the two evaluation schemes.

A switch controller design for the acrobot using neural network and genetic algorithm

This paper presents an intelligent control method for the Acrobot with applications of neural network (NN) and genetic algorithm (GA). A switch controller is proposed where a neurocontroller (NC) optimized by GA is used for the swing-up stage and a linear quadratic regulator (LQR) is applied to the balancing stage. In order to analyze the characteristics of the proposed control system, we investigate the performance of the controller with different swing-up timing constraints. Simulation results show that the method has advantages in that it can provide smooth control process and allow us to flexibly define the swing-up time in advance.

Particle swarm optimization with genetic recombination: a hybrid evolutionary algorithm

This article presents a hybrid evolutionary algorithm (HEA) based on particle swarm optimization (PSO) and a real-coded genetic algorithm (GA). In the HEA, PSO is used to update the solution, and a genetic recombination operator is added to produce offspring individuals based on the parents, which are selected in proportion to their relative fitness. Through the recombination, new offspring enter the population, and individuals with poor fitness are eliminated. The performance of the proposed hybrid algorithm is compared with those of the original PSO and GA, and the impact of the recombination probability on the performance of the HEA is also analyzed. Various simulations of multivariable functions and neural network optimizations are carried out, showing that the proposed approach gives a superior performance to the canonical means, as well as a good balance between exploration and exploitation.

Neurocontroller with a genetic algorithm for nonholonomic systems: flying robot and four-wheel vehicle examples

This article considers intelligent control for a class of nonholonomic systems using a neurocontroller (NC) and a genetic algorithm (GA). First, we introduce the design of the NC with use of the GA, and then we apply the NC to control two typical examples of nonholonomic systems: a hopping robot in the flight phase and a four-wheel vehicle. In order to verify the effectiveness of the control system, the performance of the NC is investigated and also compared to that of the so-called direct gradient descent control (DGDC) approach, which is able to utilize a GA with the same examples in the comparison. Simulations show that the NC could achieve a competitive performance and control the nonholonomic systems effectively. Furthermore, the use of the NN and GA provide a straightforward solution for the problem without the need of the chained form conversion.

Particle Swarm Optimization of a Recurrent Neural Network Control for an Underactuated Rotary Crane with Particle Filter Based State Estimation

This paper addresses the control problem of an underactuated rotary crane system by using a recurrent neural network (RNN) and a particle filter (PF) based state estimation. The RNN is used as a state feedback controller which is designed by a constricted particle swarm optimization (PSO). As the study also considers the problem with assuming that the velocities of the system are not obtained, PF is utilized to estimate the latent states. Simulations show that the RNN could provide a superior evolutionary performance and less computational cost compared to a feed forward NN and that the PF is effective in estimating the unobserved states.

Intelligent control of a three-DOF planar underactuated manipulator

Recently, computational intelligence has been applied extensively in control engineering, especially for systems that cannot easily be controlled by conventional means. In this article, attention is paid to the control of a three-DOF planar underactuated manipulator, also known as the three-link gymnastic robot, by utilizing a neural network (NN) and a genetic algorithm (GA). In an attempt to make the problem more analogous to human gymnastics, constraints are applied to the joint angles. With different swing-up timings, the performance of the proposed controller is investigated and control simulations are performed. Numerical simulations show that the neurocontroller is able to control the system effectively within the constraints and given timings.

Fruit maturity detection using neural network and an odor sensor: Toward a quick detection

Maturity detection is very important for fruit farmhouses. In a previous study, we revealed a type of odor sensor that responds to the strength of the fruits smell as well as to the fruits maturities. The smell data consists of a dead time and a step response of a first-order lag element. We focus on the step response of first-order lag element, which is a form that rises exponentially to a constant value. This paper presents a quick detection method of fruit maturity in a few seconds of the rising signal of the odor sensor. Using neural network, the method performs without waiting for the sensor to fully reach up to a constant value. First, a neural network is trained for sample data with two kinds of maturities: fully ripe and immature. By testing the neural network with untrained data, we confirmed that the network is able to detect the fully-ripened, middle-ripened, and unripe fruits.

Online neuroadaptive control of a rotary crane system

This paper is concerned with the control of a rotary crane system which is perturbed by a strong and sudden disturbance. Since the payload of the crane system is affected strongly by inertia, it is hardly stabilized quickly, particularly when there exists disturbance. An adaptive adjustment of the controller against the disturbance is thus needed to maintain the desired performance. The problem becomes more challenging when using evolutionary algorithms based techniques as they are usually computationally demanding. In this study, an online control method using neural network (NN) and genetic algorithm (GA) is proposed where a state is predicted and then used as a new initial condition for GA to perform re-designing the controller. Simulations show that the method works effectively to regulate the perturbed system to the desired state.