Results 1 -
3 of
3
Investigations on the robustness of an evolved learning mechanism for a robot arm
- In Proc. of the 8th Int. Conf. on Intelligent Autonomous Systems (IAS-8
, 2004
"... ..."
The Transfer Problem from Simulation to the Real World in Artificial Evolution
"... In evolutionary robotics artificial neural networks are often used as controllers. As the process to evolve such controllers in the real world is time-consuming, one usually uses simulators to speed up this process. By doing so a new problem arises: The evolved controllers in simulation show often n ..."
Abstract
- Add to MetaCart
In evolutionary robotics artificial neural networks are often used as controllers. As the process to evolve such controllers in the real world is time-consuming, one usually uses simulators to speed up this process. By doing so a new problem arises: The evolved controllers in simulation show often not the same fitness as those in the real world. In order to close this reality gap we propose to evolve networks able to change with experience allowing to adapt to unforseen perturbations. This paper reports on experiments with a robot arm for which a controller was evolved able to adapt to changes of the robot’s morphology. The controller was not specified but grown using a developmental approach in which the cells were controlled by artificial genomes. 1
