Trajectory optimization using SNOPT. With constraints expressed using Neural Net.
SNOPT python wrapper can be downloaded from https://github.com/snopt/snopt-python.
The neural network constraint is implemented using pytorch.
optimal_control.py contains the class SQP, which calls SNOPT to solve a simple trajectory optimization with state constraints.
It is tested with a simple 2D point mass dynamic example, with circle constraint.