The invention discloses a local traffic optimization method based on reinforcement learning and a generative adversarial network, and the method comprises the steps: building a training model, employing a generative adversarial network to autonomously improve the accuracy of the model, and predicting the traffic flow data at a specified moment through training the real traffic flow data detected at a certain intersection; the real traffic flow data and the virtual traffic flow data are trained and output by adopting Q learning to form a Q value table, an optimal local traffic optimization strategy is obtained by adopting a reward function, and the traffic signal lamp period adjustment efficiency is greatly improved by utilizing the advantage of interactive learning of reinforcement learning, so that the traffic signal lamp period adjustment efficiency is improved; and whether the congestion condition is relieved or not is verified by adjusting the current congestion level and the traffic light time ratio of a certain intersection, the optimal traffic light time ratio is obtained through repeated and continuous optimization, finite-time optimal training of Q learning is achieved through the heuristic self-game thought of the generative adversarial network, local traffic optimization is achieved, and finally, an optimal adjustment scheme is obtained, so that the local traffic optimization capability is improved.
一种基于强化学习与生成式对抗网络的局部交通优化方法,包括建立训练模型,采用生成对抗网络自主提升模型的准确率,通过训练某个路口检测到的真实车流量数据预测指定时刻的车流量数据;采用Q学习对真实车流量数据和虚拟车流量数据进行训练输出动作形成Q值表,采用奖赏函数,得到最佳局部交通优化策略,利用强化学习交互式学习的优点,大大提升了交通信号灯周期调整的效率,由某个路口的当前拥堵级别和红绿灯信号灯时间比调整来验证拥堵情况是否有所缓解,以此往复不断优化得到最佳的红绿灯时间比,再利用生成式对抗网络的启发自博弈思想实现对Q学习的有限时间最佳训练,实现局部交通优化,最终得到最优调整方案,从而提升局部交通优化能力。
Local traffic optimization method based on reinforcement learning and generative adversarial network
基于强化学习与生成式对抗网络的局部交通优化方法
2021-06-18
Patent
Electronic Resource
Chinese
Traffic flow prediction method based on generative adversarial network
European Patent Office | 2021
|Generative Adversarial Network-Based Regional Epitaxial Traffic Flow Prediction
Springer Verlag | 2019
|Urban traffic flow prediction method based on generative adversarial network
European Patent Office | 2020
|Missing traffic state data filling method based on generative adversarial network
European Patent Office | 2023
|Elsevier | 2025
|