Sensory data fusion: a team consensus approach

The authors present a new data fusion approach which is based on a recursive team loss function that is capable of bringing sensors to a consensus. In this model a consensus is achieved by allowing all group members to linearly pool their assessments in a recursive manner. Each sensor must first assess its own observations. It is then confronted with the result of observations of the other sensors and revises the result of its own observations in light of the others by making an assessment of each group member's weight. Given that each sensor revises its opinion in this manner, to be consistent each should be update its own assessment in light of the revisions made by the others and the process continued until further revision no longer changes the assessments of any member. A decision theory formulation for the data fusion problem is established and it is shown how the consensus model can be used, under this formulation, to fuse sensory data.<
>