The Olfactory Perception System Of Robot Based On Multi-Sensor Fusion

With the development of technology, the application of robots will become more and more common in the production and human life. The robots are required to be more intelligent, not only required to accomplish a given task, but also to perception surrounding environment and react accordingly to achieve complex missions. The robot with olfactory function can complete lots of serious dangerous work instead of human, such as searching for poison gas into the gas market sources, finding leakage points of pipeline and so on. The research of robots'olfactory has important meanings and abroad application foreground.This thesis is supported by a National Program 863 item--"Multi-perception humanlike head system facing on limit circumstance". By learning the human perception of the outside gas, it studied the olfactory perception system of robots, completed the quantitative detection by fusing information of gas sensors, established foundation of gas source localization. The main contents are as follows:First, Three major functional modules of the typical artificial olfactory perception system: sensors array, signal preprocessing and pattern recognition, are introduced. After the studies of general ways of the three major functional modules, the program of olfactory perception system hardware platform is completed. Sensors array is confirmed, the meatus structure is designed based on the principle of humanoid nose, data information is collected and the communication with the computer is completed by SCM.Secondly, a BP neural network about the quantitative detection of unknown gases, which based on the correlative theories, are designed, train and check it up by sample signals. Test networks that has different numbers of crytic layer cells, choose the appropriate number of crytic layer cells by error quantities. The result proves that the BP neural network can complete multi-sensor information fusion preferably and realizes the quantitative detection of unknown gases.Finally, a advanced way, which optimize the sensors array, are designed by the combination of genetic algorithm and BP neural network. Simultaneously, program the process by the use of the genetic algorithm and neural network tool chests of MATLAB, obtain the optimized array based on the result analyse, reduce the system scale, enhance the measure precision.