车－路环境自适应的重型汽车动力系统控制方法研究

项目名称： 车－路环境自适应的重型汽车动力系统控制方法研究

项目编号： No.51275291

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 机械、仪表工业

项目作者： 杨林

作者单位： 上海交通大学

项目金额： 80万元

中文摘要： 为推动汽车节能减排，国内外主要从动力系统、轻量化等方面着手。车-路环境与动力系统间呈强耦合的闭环系统，对油耗、排放等有巨大反作用力，但在现有控制方法中都没有考虑，属共性关键问题。本项目以重型车动力系统为例，通过感知推理、传感监测和在线估计相结合的方法，进行车-路环境信息识别与预报；基于样本训练和参数学习，建立车辆油耗、排放等性能的评价和预测模型；进而运用学习控制、切换系统和预测控制等理论和方法，进行发动机功率多段智能控制，并在预控制时间/距离窗口内按油耗与排放等多目标优化及与驾驶协调控制的要求，研究车-路环境自适应的动力系统自学习控制及控制参数自学习标定方法，自动考虑该反作用力，为实现系统的自适应控制提供理论基础；并研制控制系统样机进行仿真和试验研究，为成果应用奠定基础。本研究尚未见报道，是汽车动力系统控制新方法、新方向，将为汽车低成本节能减排提供新思路、新途径，应用前景广阔，极具研究价值

中文关键词： 重型汽车动力系统；车－路环境识别与预测；机器自学习；全局近似最优；自适应控制

英文摘要： In order to save fuel consumption and reduce emissions of vehicles, many researchers have mainly focused their efforts on the technologies of vehicular powertrain and lightweight. We noticed that, the relationship between vehicle status, road environment and powertrain is a strongly coupled closed-loop system, which has a very important impact on the fuel consumption and emissions. However up to the present, this very important impact has not been effectively involved in the vehicle powertain control system. So, we hope to solve this problem in this project. In his project, the heavy-duty vehicle powertrain will be taken as an example, with a combined method of perception reasoning, sensors monitoring and on-line estimation, the driver behavior, vehicle parameters and road environment signals will be identified and predicted; through the algorithms of samples training and parameter learning, the evaluation and predictive models for fuel consumption and emissions will be established; further, based on the theory and methods from learning control, switch system and predictive control, the engine-power-multi-segment based intelligent control strategy will be studied; the methods for vehicle-road environment self-learning control and control parameters self-learning calibration will be researched according to the de

英文关键词： heavy-duty vehicular powertrain；vehicle-road-environment identification and predic；machine self-learning；Global approximate optimal；adaptive control