Computationally tractable learning-based nonlinear tube MPC using difference of convex neural network dynamic approximation

Learn dynamical systems as a difference of convex functions (DC) using a Recurrent Neural Network (RNN) architecture with DC structure. Example on the coupled tank problem.

Learn dynamical systems as a difference of convex functions (DC) using a feedforward Neural Network (NN) architecture with DC structure. The resulting model learns the dynamics f in DC form as follows: f= f1 -f2 where f1 and f2 are convex functions. The DC structure of the network allows to independently express f1 and f2 as two input convex NN.