A Newton/GMRES Approach to Predictive Ecological Adaptive Cruise Control of a Plug-in Hybrid Electric Vehicle in Car-following Scenarios*. Issue 21 (2016)

• Record Type: Journal Article
• Authors: Tajeddin, Sadegh; Vajedi, Mahyar; Azad, Nasser L.
• Is Part Of: IFAC-PapersOnLine. Volume 49:Issue 21(2016)
• Page Start: 59
• Page End: 65
• Publication Date: 2016
• Subjects: Nonlinear Model Predictive Control -- Ecological Adaptive Cruise Controller -- Newton/GMRES Optimization -- Intelligent Transportation Systems; Automatic control -- Periodicals; 629.805
• DOI: 10.1016/j.ifacol.2016.10.511
• Languages: English
• ISSNs: 2405-8963

Abstract

Abstract: Exceptional compound structure of Hybrid Electric Vehicles (HEVs) enables scholars to employ intelligent control strategies that could improve their ecological characteristics. The complex powertrain dynamics of HEV, demands a powerful optimal control strategy that could handle complicated, nonlinear, mutli-objective problems with constraints. Nonlinear Model Predictive Control (NMPC) is a powerful tool that could provide such demands and also enables the ability to consider future predictions of the vehicle's environment in the control process. However, NMPC is very computationally expensive and to run in real-time, demands a fast accurate solver. GMRES-based optimization methods are promising fast solvers for NMPC that could perfectly handle such huge computational burden. In this study, we use GMRES-based NMPC to design an Ecological Adaptive Cruise Controller (E-ACC) for a Plug-in HEV, namely the Toyota Plug-in Prius. The designed E-ACC utilizes future trip information and an on-board vehicle radar to optimise energy cost of the trip while maintaining safety and comfort. Also, an automatic code generator will be introduced that generates the fast Newton/GMRES optimizer, used at the heart of our controller. The proposed E-ACC is then evaluated by a car following scenario in two different driving cycles while the effect of important adaptive gains and weighting factors are investigated. Our simulation result shows up to 3.4% improvement in energy cost compared to a classic PID controller.