Eco-driving with Learning Model Predictive Control

TL;DR

This paper introduces a Learning Model Predictive Control approach for eco-driving that iteratively enhances fuel efficiency on daily routes while ensuring timely arrivals, validated through simulations.

Contribution

It extends the LMPC framework to include time constraints and reformulates it for computational efficiency in predictive cruise control.

Findings

Improved fuel economy over repeated trips

Guarantees on arrival time

Computationally tractable control formulation

Abstract

We present a predictive cruise controller which iteratively improves the fuel economy of a vehicle traveling along the same route every day. Our approach uses historical data from previous trip iterations to improve vehicle performance while guaranteeing a desired arrival time. The proposed predictive cruise controller is based on the recently developed Learning Model Predictive Control (LMPC) framework, which is extended in this paper to include time constraints. Moreover, we reformulate the modified LMPC with time constraint into a computationally tractable form. Our method is presented in detail, applied to the predictive cruise control problem, and validated through simulations.