I am trying to train an LSTM model on Matlab to forecast the position of a vehicle when driving around a roundabout. My main concern right now is that my dataset consists of 4 features (X position, Y position and 2 other) and I referred to Are RNN or LSTM appropriate Neural Networks approaches for multivariate time-series regression?
and understood the method of sliding for multistep predictions, although intuitively this seems like the right way to do it, my model generates huge errors for some reason - prediction of tn is fairly alright but tn+1 onwards there is an issue similar to this.
Additionally, I have about 200 samples in my dataset ( i.e driving path of 200 different vehicles in the roundabout) how do I go about training a model for all these time-series patterns?
As of now, I am trying to make a model to work on 1 time-series sample and later build on it.
Any suggestion and proposals are appreciated!
