Research on Creating Environment Model for Mobile Robot Using Improved ART-2 Neural Network

This paper used improved ART-2 network to help robot identify environment model. In application of ART-2 neural network, it is found that there is a characteristic of pattern drifting when processing the gradually-change single-direction patterns, in one hand robot depends on the characteristic when moving in static environment, on the other hand the characteristic may threaten the safety of robot when moving in dynamic environment. To solve the problem, an improved ART-2 neural network is brought forward which makes the robot suitable for the application in static environment as well as in dynamic environment safely. The approach is implemented and tested on CLIMBER robot