| R. Der, U. Steinmetz, and F. Pasemann. Homeokinesis - a new principle to back up evolution with learning. In Computational Intelligence for Modelling, Control, and Automation, volume 55 of Concurrent Systems Engineering Series, pages 43--47, Amsterdam, 1999. IOS Press. |
....emergence of situated behaviors in a stable and reproducible manner. Our idea is to develop a measure for the situatedness of a behavior which may serve as an objective function for the adaptation of the controller. We have developed these ideas before in the general framework of homeokinesis, cf. [3] which is the dynamical pendant of homeostasis as introduced by Cannon [2] but in the present paper we want to concentrate on the practical realization. 2 Situated behavior Let us call x R the state of a system where in a concrete situation x is the vector of sensor readings of a robot. In the ....
....walls this means that it steers with two sensors, choosing the distance to the wall such that the sensor closest to the wall is at maximum value and the next one is at half maximum. This is a nontrivial effect and can be understood in terms of minimizing the influence of the noise as explained in [3]. The wall following behavior remains stable as long as there are no dramatic changes in the environment. This could be corroborated in a long time experiment where the robot was running nearly three days completely autonomously in a simple maze. It should be mentioned that during all experiments ....
R. Der, U. Steinmetz, and F. Pasemann. Homeokinesis - a new principle to back up evolution with learning. In M. Mohammadian, editor, Computational Intelligence for Modelling, Control, and Automation, volume 55 of Concurrent Systems Engineering Series, pages 43 47. IOS Press, 1999. URL--http://www.informatik.uni-leipzig.de/der/Veroeff/wienfin3.ps.
....of feedforward networks. 11 For recurrent networks, and using a behavior based approach to neurocontrollers, there is no universal learning rule to apply. Using only the internal states of a neural network, we are trying to optimize a given recurrent network structure by using ideas outlined in [3]. Furthermore, equivalents to other additional features of evolutionary algorithms like e.g. crossing over are not yet implemented in the ENS 3 algorithm. ....
Der, R., Steinmetz, U., and Pasemann, F. (1999), Homeokinesis - A new principle to back up evolution with learning, in: Proceedings of the International Conference on Computational Intelligence for Modelling Control and Automation (CIMCA'99), Vienna, 17-19 February 1999.
....of feedforward networks. 11 For recurrent networks, and using a behavior based approach to neurocontrollers, there is no universal learning rule to apply. Using only the internal states of a neural network, we are trying to optimize a given recurrent network structure by using ideas outlined in [3]. Furthermore, equivalents to other additional features of evolutionary algorithms like e.g. crossing over are not yet implemented in the ENS 3 algorithm. ....
Der, R., Steinmetz, U., and Pasemann, F. (1999), Homeokinesis - A new principle to back up evolution with learning, in: Proceedings of the International Conference on Computational Intelligence for Modelling Control and Automation (CIMCA'99), Vienna, 17-19 February 1999.
No context found.
R. Der, U. Steinmetz, and F. Pasemann. Homeokinesis - a new principle to back up evolution with learning. In Computational Intelligence for Modelling, Control, and Automation, volume 55 of Concurrent Systems Engineering Series, pages 43--47, Amsterdam, 1999. IOS Press.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC