Multiple sensor data fusion in robotic prosthetic eye system

To ensure the robustness of the robotic eye sensor system, multiple sensors are used to detect the eye movement. However, some sensors may fail and provide faulty data. In the paper, multivariate statistical techniques are used to deal with sensor data monitoring, and faulty sensor detection and isolation. In addition, principal component analysis is used to monitor the sensor data and detect the sensor failure, and an incidence matrix is used to isolate the faulty sensor. We also study LMS and minimum variance methods for recovering the faulty sensor data. Simulation studies are included.