Sensor uncertainty management for an encapsulated logical device architecture, Part II: A control policy for sensor uncertainty

A procedure to perform data fusion inside a low-level control loop has been developed and implemented on a one degree of freedom manipulator. This procedure uses sensory data provided by low-level and non-dedicated high-level sensors, at different rates. Fusion of the multiple feedback signals generates a signal with a smaller uncertainty (i.e., noise) level. The performance of the control scheme is directly related to the quality and relevance of the feedback signal, since it is the only source of information for the control algorithm. In this control policy, data fusion is performed with the data coming from the different sensors, once these one have been time-correlated using Kalman filters. Also, in order to stabilize the fused feedback signal when there is no data available from the slower sensors, a Kalman filter is used to observe and generate a prediction of the fused measurement signal, which can then be used by the data fusion process. The slow sensor processing delay compensation and fused measurement stabilization are independent of the fusion process. Therefore, any data fusion process can be used with this procedure, as long as the process respects the real-time constraint of the low-level control loop.