Supervisory MPC for Powertrain Control

Tags: Automotive

General Motors and ODYS have developed a supervisory control system for powertrain with Continuously Variable Transmission (CVT) based on real-time MPC, with the goal of optimizing driver requested axle torque tracking and improving fuel economy and drivability. This controller has been scheduled for production by GM starting in 2018.

The MPC supervisory controller takes advantage of dynamical models of powertrain, engine and transmission to coordinate desired engine torque and desired CVT ratio in real time, by solving a constrained optimization problem at each sampling step based on estimates of axle torque and engine fuel rate.

Compared to more classical controls, this MPC approach enables real-time optimization of fuel economy and improves coordination of powertrain actuators to satisfy system requirements, while maintaining robustness with respect to measurement noise, ambient conditions, and part-to-part variations.

This control strategy has been implemented in production by taking advantage of the MPC software environment developed for the engine control project carried out by ODYS and GM. Such systematic, model-based framework opens up further opportunities for adaptation of the design to different powertrain architectures.

The project is described in the paper “Supervisory Model Predictive Control of a Powertrain with a Continuously Variable Transmission”, and it has been presented at the SAE WCX World Congress Experience, held in Detroit, MI, in April 2018.