Inverted Pendulum Control Using Soft Computing Techniques

Abstract

The inverted pendulum (IP) problem is a standard problem in control systems. It is widely used as benchmark for testing of different advance control algorithms, such as neural networks, genetic algorithms etc. The inverted pendulum system is an inherently unstable, non-linear system. It belongs to the class of underactuated mechanical systems having fewer control inputs than degrees of freedom. This renders the control task more challenging. newlineThe inverted pendulum control problem is similar to balance a stick on the palm. IP consists of a cart confined to linear motion along a track and a rod attached to the cart by a hinge joint. The cart is equipped with a motor that accepts a voltage and outputs a torque as a function of the voltage applied. Generally, the inverted pendulum system is approximated as a linear system and hence, the model is valid only for small oscillations of the pendulum. newlineIn earlier works, researchers have applied various techniques for inverted pendulum control. Among the various techniques available, soft computing techniques are becoming useful in developing the methodologies for balancing inverted pendulum and understanding the dynamics of non-linear system control. Moreover, quantum inspired techniques have been proved effective to control highly nonlinear complex systems. It is observed that quantum inspired techniques have not been yet implemented for inverted pendulum control. newlineIn this research work, various soft computing techniques based controllers; such as- artificial neural network, generalised neuron model and quantum inspired genetic algorithm based neural techniques, have been developed and implemented for balancing inverted pendulum. The total work is divided into two parts. First part consists of the modeling and simulation of single and double inverted pendulum systems. The development of different neural controllers and their testing on the simulated system is performed.