Real World Robustness for Hybrid Vehicle Optimal Energy Management Strategies Incorporating Drivability Metrics

Abstract

Hybrid vehicle fuel economy and drive quality are coupled through the “energy managementâ€‌ controller that regulates power flow among the various energy sources and sinks. This paper studies energy management controllers designed using shortest path stochastic dynamic programming (SPSDP), a stochastic optimal control design method which can respect constraints on drivetrain activity while minimizing fuel consumption for an assumed distribution of driver power demand. The performance of SPSDP controllers is evaluated through simulation on large numbers of realworld drive cycles and compared to a baseline industrial controller provided by a major auto manufacturer. On realworld driving data, the SPSDPbased controllers yield 10% better fuel economy than the baseline industrial controller, for the same engine and gear activity. The SPSDP controllers are further evaluated for robustness to the drive cycle statistics used in their design. Simplified drivability metrics introduced in previous work are validated on large realworld data sets.