Information coding with neural ensembles for a mobile robot

For robot navigation (obstacle avoidance) we propose to use special neural network, because of its large information capacity for non correlated data. We prove this feature in contrast for correlated data in the robot task. This information is generated by a simulator and coded into neural ensembles. The coding method allows different parameters with their numeric values to be stored; it also provides similarity for close values and eliminates it in other case. The developed system combines the quality of the neural network as associative memory and the coding method to permit learning from some specific situations. So we prove the system introducing only the situation information and retrieving the appropriate maneuver for it.