Data-Driven Predictive Control of CVT System for Improving Energy Efficiency of Autonomous Vehicles

In this paper, the predictive data-driven control of Continuously Variable Transmission (CVT) systems is investigated to improve the energy efficiency of autonomous vehicles (AVs). Algorithms in AVs are expected to autonomously adapt and optimize themselves with different drive cycle conditions using the collected data during driving operation. We propose a Data Driven Control (DDC) framework for the speed ratio of CVTs in AVs which can learn the optimum pattern from archived real data. It includes a data archiving cycle in which the required DDC input/output signals of the vehicle are recorded. A DDC framework then runs in the driving cycle which includes a rolling optimization problem. More specifically, we use a predictive DDC algorithm to learn the optimal CVT speed ratio during the driving cycle instead of using the traditional pre-defined mapping. Using this framework, the CVT controller can adapt itself to any drive cycle by appropriately collecting data during the archiving cycle. The proposed control framework shows a satisfactory performance in HWFET and NEDC drive cycles. It reduces energy consumption by 1–5% compared to the pre-defined mappings.