S. Goschin, E. Franti, M. Dascalu, and M. Pietraroiu (Romania)
Evolutionary robotics, simulation, neural networks and genetic algorithms
The purpose of this paper is to present a method of combining neural networks and genetic algorithms to create an efficient control system for a simulated autonomous robot in a 3D environment. The experiments concern the emergence of behaviours like obstacle avoidance, target hitting and shortest path finding using a simple yet robust architecture. The evolutionary process takes place inside a dedicated, flexible framework that allows further development and testing.
Important Links:
Go Back
