A Costate Estimation for Pontryagin's Minimum Principle by Machine Learning

摘要

A possibility of costate estimation for Pontryagin's Minimum Principle is discussed in this paper. The optimal initial costate data are collected from various driving cycles by backward simulation. In the simulation, a parallel hybrid electric vehicle model is used. The cycles which are used in data collection are generated from real world driving data. Long Short-Term Memory networks(LSTMs) which are one of machine learning algorithm are used to learn optimal initial costate.