I am searching for an approach that is able to predict a complete time series for a given parameter set.

Imagine a robotic arm which has a starting position and a target position. There is a sensor mounted to the top of the arm. For random start and target positions I can record the sensor values as a time series. Now, I would like to predict the sensor values for a previously unknown start and target position.

Currently, the only approaches I can find try to forecast the next values for a given time series. Other approaches aim for generating new sequences for a given set of sequences, which is close to what I would like, but without the possibility to connect the sequences to parameter sets.

Does anyone have an idea?


1 Answer 1


Check the literature on functional data. Your function here comprises the sensor values over a complete path of the robotic arm. From a random sample of such functions you are wanting to predict a new function with specified start and target positions. A good intro is Ramsay and Silverman (2005).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.