Using Reinforcement Learning to Combine Green Light Optimised Speed Advisory or Equivilent Systems and Adaptive Traffic Lights

随着道路使用者数量的增加，道路规划和建设成本高昂、耗时且对环境有害，许多机构都在寻求越来越智能的基础设施，如新的自适应交通控制系统和绿灯优化速度咨询系统。然而，使用当前的自适应交通控制系统，绿灯优化速度咨询系统会导致效率低下，因为准确的未来信号计划并不总是可用的，从而导致不准确的速度咨询。本文的目的是利用机器学习领域的技术来训练人工智能来控制未来的信号计划，并决定发送给司机的细节。本文详细介绍了用Python构建的测试平台的构建，并用于演示每个系统的优点以及使用两者的低效率。

With the number of road users increasing and road planning and building being expensive, time consuming andenvironmentally damaging, many institutions are looking towards increasingly smarter infrastructure like new adaptivetraffic control systems and green light optimisation speed advisory systems. However, using current adaptive trafficcontrol systems green light optimisation speed advisory systems leads to inefficiency as accurate future signal plans arenot always available leading to inaccurate speed advisories. The aim of this thesis will be to employ techniques fromthe eld of machine learning to train an AI to control future signal plans as well as decide on the details to send todrivers. This thesis details the building of test bed that has been built in Python and used to demonstrate theadvantages of each system as well as the inefficiencies of using both.