EVOLVING ADAPTIVE SENSORS IN A SYNTHETIC LISTENER
Peter Beyls · University of Michigan Library Repository · 2005
This paper suggests a method to evolve responsive networked sensor configurations using a genetic algorithm.The system integrates symbolic processing in specialized, adaptive sensor functions with sensor fusion using a distributed, connectionist network.The process of interaction itself is viewed as a complex dynamical system featuring wide ranges of remarkable non-linear behavior.An off-line algorithm is proposed to gradually evolve fit networks and discover interesting dynamic behavior that could not have been anticipated by an explicit programmer.