Output details

Adaptive Dynamic Control of Quadrupedal Robotic Gaits with Artificial Reaction Networks

Here the ARN model is used to create pattern recognition networks, using an evolutionary algorithm. These are then interfaced to the previously investigated temporal pattern-producing networks. The result is a full robotic control system, capable of recognising environmental stimuli (for example, from a camera) and producing a resultant pattern to control the robots’ legs. The importance of this is two fold: Firstly it demonstrates the chemical signalling networks within cells are capable of such complex mappings (similar to those in neural networks). Secondly, that the units can produce both control and pattern recognition systems - which simple neural networks cannot.