Sensor networks with transmission random packet dropouts are considered.

Estimation from outputs with random matrices and correlated noises is addressed.

Filters are designed at sensors using its information and those from its neighbours.

The algorithms, based on covariances, are derived using an innovation approach.