A comparison of stochastic optimal control laws for a parallel hybrid vehicle

Paper in proceeding, 2006

This paper studies the properties of stochastic optimal control laws for hybrid vehicles, derived using stochastic dynamic programming with a Markov Chain as a drive behavior model. The focus of the study is on establishing how complex the Markov Chain model needs to be in order to achieve good performance. Moreover, a simple robustness analysis is done in order to examine how a control law, optimized with data from one city, performs when evaluated in a city with a considerably different drive behavior. The results indicate that it is sufficient to use a Markov Chain model with only power demand as the sole variable in the Markov state. Furthermore the simulation results show a remarkable robustness for the fuel consumption of the derived controllers when evaluated on a new drive behavior.