An Edge Computing Framework for Powertrain Control System Optimization of Intelligent and Connected Vehicles Based on Curiosity-Driven Deep Reinforcement Learning

摘要

For the ongoing revolution in developing intelligent and connected vehicles (ICVs), there is a lack of research for powertrain control systems using the latest artificial intelligence and vehicle-to-everything technology that have already been widely adopted in the autonomous driving systems. In this context, recent development of deep reinforcement learning (DRL) and one of the latest computing frameworks are coupled to facilitate an onboard-based intelligent powertrain control. Taking the boost control of a diesel engine equipped with variable geometry turbocharger as an example, the results show that the final control behavior indicated by the cumulated rewards is improved by 50.43% and the learning efficiency is improved by 74.29% for the proposed curiosity-driven DRL algorithm, compared with the same structure DRL algorithm with classic random exploration policy. In addition, unlike most of the DRL-based powertrain optimization algorithms, which have only been applied to single-machine architecture, this work manages the proposed DRL algorithm in parallel and, more importantly, from an edge computing perspective. This, in addition to greatly speeding up the algorithm training, can also realize a good balance of control accuracy and generality depending upon the selected training scenario. Moreover, unlike most of the cloud computing frameworks, which require low network latency, the proposed architecture can achieve a similar final control performance even if good network communication is not allowed. Compared with other existing powertrain control methods, the proposed algorithm is able to approximate a global powertrain control optimization autonomously in a connected manner, making it attractive to current ICVs with advanced automated driving and traditional powertrain control.