DESIGN OF PID & ARTIFICIAL NEURAL NETWORKS BASED CONTROLLER FOR INVERTED PENDULUM

DESIGN OF
PID & ARTIFICIAL NEURAL NETWORKS BASED CONTROLLER FOR
INVERTED PENDULUM

ABSTRACT

Control Systems are the means by which any quantity of interest in a machine, mechanism or other equipment is maintained or altered in accordance with a desired manner. Control systems have played a major role in the rapid industrial development.

In the present dissertation an attempt has been made to design a controller for inverted pendulum system. The inverted pendulum problem is a classical control problem that has been extensively studied by many researchers and seems to be well understood. Also, the problem is representative of many typical dynamical plant control problems. Thus, designing the controller for this inherently unstable system will allow us to approach and solve a number of other, similar control tasks. The complexity of the task of inverted pendulum balancing is significant enough to make the problem interesting.

A number of control techniques are available to balance an inverted pendulum, like classical techniques and Modern technique. In classical techniques we find the solution of our problem using frequency response method, root locus method etc., in modern techniques we find solution using pole-placement method.

In the present thesis an attempt has been made to find the solution of our problem using three ways as explained bellow,

1. Design of compensator using frequency response method.

Here we found that pendulum can stabilized but cart can moves in negative direction.

2. Design of PID controller for Inverted Pendulum.

Here also get the same result, Pendulum can be stabilized but cart can move in negative direction.

3. Design of controller by using Artificial Neural Network(ANN).

In recent years, ANNs are becoming more popular among engineers because of their various advantages over other methods. Here we use Decoupled Neural Network Reference compensation Techniques to control a two degree-of-freedom inverted pendulum. Neural Networks are used as an auxiliary controllers to help the PD controller for the system to minimize the errors of angles and position of each axis. The Neural Network might be used directly as a controller, but this approach has several drawbacks like,

1. During the training period, the control system is not operational.

2. It cannot eliminate unpredictable disturbances.

3. This approach bears a less direct connection to the design methods for traditional controller.

To avoid these problems, in the proposed scheme, the conventional PD controller is combined with feed forward Neural Network.

Two separate Neural Networks are used for controlling each axis of Inverted Pendulum. Here Neural Network is used to control both angle and position of Inverted Pendulum. Two types of Neural Networks are used for this purpose; one, two layer feed-forward structure with a tangent hyperbolic activation function and second, Radial Basis Networks.