Spiking neural network based heterogeneous federal filter method for sensor failure in multi-actuator systems

A Spiking neural network based heterogeneous federal filter method is proposed in this paper to deal with sensor failure problems in multi actuator systems. Firstly, an accurate dynamic model of multi-motor system is built, which can provide a high precision estimated state information in viral system state simulator. Secondly, a heterogeneous federal filter is proposed to provide a stable and accurate state feedback whether the sensor is normal or not. This filter collects real-time measured state data from different kinds of sensors and the estimated state information, and gives out an filtered state output for each sensor according the spiking neural network based reliability diagnosis module. By this way, the noise and disturbance can be eliminated. When one or more sensors break down, an estimated accurate state feedback can be obtained. Finally, this heterogeneous federal filter is applied into an four-motor servo system. The experiment results indicate that: the spiking neural network based reliability diagnosis module could quickly monitor sensor status accurately and the heterogeneous federal filter could handle sensor failures with control accuracy remained the normal level.