Towards a sub-optimal solution of a partially observed state dependent JLQ problem