Evolutionary neural network-based sensor self-calibration scheme using IEEE 1451 and wireless sensor networks

Abstract

Plug-and-play sensor self-calibrating technology is presented in this paper. The solution involves the evolution and tuning of a neural network (NN), through a genetic algorithm (GA). The former is utilized to interface to a sensor, on-board a robotic sensor agent. Multiple NN interfaces can be utilized for multiple sensors, hence providing for a parallel and scalable system. The system introduces the "sense remotely, actuate immediately" concept, along with an analysis of a completely pervasive and sentient environment, in which sensors provide the user with real-time and wireless sensory information, while actuators provision for the user's response to the filtered data, streaming from the plethora of intelligent sensors.