Heterogeneous neural networks for adaptive behavior in dynamic environments
Advances in neural information processing systems 1
SIGGRAPH '93 Proceedings of the 20th annual conference on Computer graphics and interactive techniques
SIGGRAPH '94 Proceedings of the 21st annual conference on Computer graphics and interactive techniques
| Hi-index | 0.00 |
We developed a simulation model and generated a gait pattern for quadruped artificial life. The model is based on three dimensional physical simulation using a physics engine. Neural networks are used to control each leg, and the genetic algorithm is used to evolve the gait. The generated gait pattern is similar to the gait called "walk" in real-world animals. An analysis is conducted of the developing gait pattern, in addition to the final result. The emergent walk-like gait is similar to a newborn baby crawling on the floor, and it would appear that the artificial life struggles to go straight ahead on the midway of evolution.