Neural networks in robotics

The essence of the development of technical sciences is the search for new theoretical and practical solutions. Currently analitical solutions to technical problems are substitited by applications of neural networks. The paper describes the utilisation of neural networks in robotics. Neural networks are applied to robot control systems, especially to sensor data processing and solution of kinematics and dynamics problems. Different types of networks as well as the features of the learning process (e.g. types of patterns used, learning phases) are described. The paper is concluded by the discussion of difficulties in the selection of neural network, learning algorithms and network parametrs.