In the development of control theory, the validity and the feasibility in practical application need to be proved by a typical object controlled by a controller designed according to the theory. The inverted pendulum is such a controlled device. Adaptive inverse control can be used to design wellperformed adaptive inverse control system with little prior experience and without knowing the mathematic model of the controlled device. However, when it comes to the nonlinear control system, it is unfeasible to apply the linear selfadapting filter to the nonlinear modeling, for adaptive inverse control requires the modeling of forward model and inverse model. In contrast, the neural network is based on selflearning and is capable of approximating arbitrarily nonlinear mapping, so it can control the inverted pendulum. The neural network can obtain forward model and reverse model after network training, which can complete the adaptive inverse control of deeply nonlinear double inverted pendulum.Choosing a training method is very important when using neural network to obtain the model and the inverse model of the nonlinear object. The general BP algorithm based on gradient descent depends on the initial weight, and its convergence rate is relatively slow. Study shows that particle swarm algorithm is a potential algorithm in neural network training, it reserves the collateral global hunting strategy based on genus swarm and its velocitydisplacement model is simple to operate and easy to achieve.Particle swarm optimization is a kind of selfadaptive random algorithm based on group hunting strategy. As a kind of intelligent optimization, it can be used to solve various optimization problems and shows great potential in practice. Now, it has been widely applied in many areas, such as artificial neural network, fuzzy system control and pattern recognition.Based on the above, this paper has done the following contents: 1. Introduces the basic concepts and extended structure of adaptive inverse...
