This website requires JavaScript.

Streaming Traffic Flow Prediction Based on Continuous Reinforcement Learning

Yanan XiaoMinyu LiuZichen ZhangLu JiangMinghao YinJianan Wang
Dec 2022
摘要
Traffic flow prediction is an important part of smart transportation. Thegoal is to predict future traffic conditions based on historical data recordedby sensors and the traffic network. As the city continues to build, parts ofthe transportation network will be added or modified. How to accurately predictexpanding and evolving long-term streaming networks is of great significance.To this end, we propose a new simulation-based criterion that considersteaching autonomous agents to mimic sensor patterns, planning their next visitbased on the sensor's profile (e.g., traffic, speed, occupancy). The datarecorded by the sensor is most accurate when the agent can perfectly simulatethe sensor's activity pattern. We propose to formulate the problem as acontinuous reinforcement learning task, where the agent is the next flow valuepredictor, the action is the next time-series flow value in the sensor, and theenvironment state is a dynamically fused representation of the sensor andtransportation network. Actions taken by the agent change the environment,which in turn forces the agent's mode to update, while the agent furtherexplores changes in the dynamic traffic network, which helps the agent predictits next visit more accurately. Therefore, we develop a strategy in whichsensors and traffic networks update each other and incorporate temporal contextto quantify state representations evolving over time.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答