Two neural architectures applied in the environment rebuilding for a mobile robot

This paper presents two neural networks architectures (i) to recognize and rebuild the environment surrounding a mobile robot, (ii) to generate the robot reflexive behaviors. The first one classifies in a single step the whole set of environmental objects. The second architecture sorts the environment into classes and operates a second processing to highlight objects. A third neural network creates the reflexive behaviors that steer the robot in the navigational phases of its moving. The three nets work with information coming from a 2-D laser telemeter supplying the distance measurements between the obstacles and the robot. After the description of the mobile autonomous robot behavioral architecture, the neural networks are presented for every architecture, as well as a complete set of results and the training method.