Time series autoencoder

I have a training set from two real-valued time series from two very similar sensors. I would like to use deep cognition to map a one-time series to the other. I can split both the time series in sequences of say 20 secs.
How would you suggest I go about modelling the autoencoder in deep cognition? Has anyone else done something similar? I would be grateful if someone can point me in the right direction.