Method for servoing a robot in a three dimensional space and robot moving in such a space

摘要

The process uses a preliminary learning stage (1) during which a neural network is input with data corresponding to a number of visual images of a control mark from which are calculated the inverse of the Jacobean matrix, each corresponding to the variation of perception of the robot due to its changing position in three dimensional space. Having trained the neural network to recognise the spatial position of the robot from the visual images, the network forms an element of closed loop control during the real time operation of the robot.