E. Dauce and M. Quoy (2000) "Random recurrent neural networks for autonomous system design." Proceedings of the 6th International Conference on Simulation of Adaptive Behavior (SAB'2000), pp. 31-40.
==================================================================================================



     Bicho, E. and Schoner, G (1997) The dynamics approach to autonomous robotics demonstrated on a low-level vehicle plat form. Robotics and autonomous System, 21, pp 23-35 => no

     Schoner, G. Dose M, and Engels C (1995). "Dynamics of behavior: theory and applications for autonomous robot architectures." Robotics and autonomous System, 16 pp 213-245. => no




"As animal carries out a cognitive act, that is, one can observe global spation-temporal patterns of activation emerging from background activity. Such patterns have a very shor life, e.g. of the order of tenths of milliseconds, and their extinction leads to the emergence of new patterns."

"As soon as the gain parameter g is high enough, every random network tends to produce a complex spation-temporal pattern of activation."

Neurons whose mean potential is large positive or negative have almost constant output signals. Such neurons are called inactive or quiescent. Only neurons whose potentials oscillate around zero have their signal amplified by the transfer function. such neurons are called active neurons. ... For usual parameter values, active neurons represent about 30% of the whole population."

"We have seen that a majority of neurons in our system have an almost constant activation value (silent or saturated). Such neurons are not involved in the learning process."

First the dynamics is iterated without changing the weights, until the system reaches its stationry dynamics. Second, the learnign rule is iterated while the robot isw moving. ... The learning process is lasting 20 time steps (one time step corresponds to one movement.)

The main drawback is the need to tell the system when to stop learning. ... we have chosen to stop after a fixed number of iterations.