摘要
提出一种用于汽车排放试验中驾驶机器人对车速跟踪控制的新方法。该控制方法基于神经网络并结合强化学习的自适应能力,通过神经网络的在线学习对车速进行跟踪控制。利用试验汽车所获得的数据,首先开发出用于车速控制的神经网络模型。然后基于强化学习神经网络结构设计神经网络控制器以取得车速跟踪的自适应控制。在仿真研究中,使用神经网络车速控制模型替代实际汽车来训练初始控制器,并用开发与训练好的自学习神经网络控制器用于汽车车速跟踪控制。结果表明,所开发的神经网络控制器具有良好的车速跟踪性能,控制效果明显。
A new approach for tracking vehicle speeds by robotic driver during emission testing is presented. Based on neural network and combined with adaptive capability of reinforcement learning, it can execute velocity tracking control through on-line learning of neural network. Using the data obtained from a test vehicle, a neural network model of automotive for velocity tracking is developed at first. A neural network controller is designed based on reinforcement learning neural network framework to achieve adaptive control of velocity tracking. During simulation study, the velocity control neural network model is used to train primary controller rather than the actual test vehicle, and the developed and well-trained self-learning neural network controller is applied to velocity tracking control. Results show that the developed neural network controller has good performance of velocity tracking, and control efficacy is obvious.
出处
《测控技术》
CSCD
2007年第7期36-38,共3页
Measurement & Control Technology
关键词
强化学习
神经网络
车速跟踪
驾驶机器人
reinforcement learning
neural network
velocity tracking
robotic driver
