基于监督式ADP的汽车智能巡航控制

项目名称： 基于监督式ADP的汽车智能巡航控制

项目编号： No.61273136

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 自动化技术、计算机技术

项目作者： 赵冬斌

作者单位： 中国科学院自动化研究所

项目金额： 80万元

中文摘要： 汽车的自适应巡航控制和起停控制相结合可实现全速度范围内的辅助驾驶，对于提高汽车行驶的安全性、舒适性、节能性具有重要意义。作为一种辅助驾驶系统，其应用普及的关键在于控制效果需要符合驾驶员的特性。因此，提出了开发全速度范围内符合不同驾驶员特性的智能巡航控制(ICC)系统的需求。理论上，将ADP(自适应动态规划)和监督学习相结合，既可提高ADP的学习效率，又使得学习结果具有导师特性和最优性，本质上非常适合求解汽车ICC的问题。本项目拟通过理论、方法、平台、实验紧密结合的方式对汽车ICC问题进行系统研究：提出先进的监督式ADP方法，给出收敛性和系统稳定性的理论分析结果，形成采用人机交互方式实现最优控制的理论和方法；应用所提出的方法，实现汽车的ICC，并在仿真驾驶平台上针对不同场景进行驾驶员在环和硬件在环的实验验证，与驾驶员特性进行比较分析，进而为汽车的安全、辅助驾驶提供一套ICC方法和技术。

中文关键词： 自适应动态规划；监督式自适应动态规划；稳定性；智能巡航控制；驾驶员行为习惯

英文摘要： The full-range Adaptive Cruise Control（ACC） system is a driver asisstance system which considers both the ACC situation in highway systems and the Stop and Go (SG) situation in urban street systems. It has an important significance in improving safety, comfort and fuel economy. The key to its universal application is to meet the requirement of human driving habbits. Therefore, it is urgent to develop an intelligent cruise control (ICC) system in full-range speed to cater for the human-like driving habits. In theory, ADP (Adaptive Dynamic Programming) can be combined with supervised learning, to improve the learning efficiency of ADP and obtain the optimality and supervisory feature, which is deemed as a feasible solution for the automotive ICC problem. The project aims at the systematic research on the theory, methods, platforms and experiments of the automotive ICC problem. It is expected to propose advanced SRL methods, analyze the convergence and stability, and formulate the optimal control theory and methods through human-computer interaction. The proposed method will be implemented in the automotive ICC by the driver-in-loop and hardware-in-loop simulator under different scenarios. Experimental results will be compared and analyzed to experienced drivers, to derive the ICC method and technique for enhancing

英文关键词： Adaptive dynamic programming；Supervised adaptive dynamic programming；Stability；Intelligent cruise control；Human driving habit