Search papers, labs, and topics across Lattice.
University of Melbourne
1
0
3
FilterDDP achieves global convergence for constrained optimal control problems by leveraging a novel backward-forward computation method that rivals traditional Newton steps.