| 基于GA-TD3算法的交叉路口决策模型 |
| |
| 引用本文： | 江安旎,杜煜,原颖,张昊,赵世昕. 基于GA-TD3算法的交叉路口决策模型[J]. 计算机应用研究, 2024, 41(7) |
| |
| 作者姓名： | 江安旎 杜煜 原颖 张昊 赵世昕 |
| |
| 作者单位： | 北京联合大学 智慧城市学院,北京联合大学 机器人学院,北京联合大学,北京联合大学,北京联合大学 |
| |
| 基金项目： | 国家自然科学基金(面上)项目(52072213);北京市教育委员会科研计划资助项目(KM202311417006);北京市朝阳区科技局项目资助项目(纵20200028) |
| |
| 摘 要： | 为了解决交叉路口场景下无人驾驶决策模型成功率低，模型不稳定，车辆通行效率低的问题，从两个方面对TD3算法作出改进，提出了基于GA-TD3算法的交叉路口决策模型。首先引入记忆模块，使用GRU神经网络来提升决策模型的成功率；其次在状态空间引入社会注意力机制，更加关注与社会车辆的交互行为，保证模型稳定性的同时提升车辆的通行效率。采用CARLA仿真器进行20 000回合的模型训练后，TD3算法通过路口的成功率为92.4%，GA-TD3算法的成功率为97.6%，且车辆的通行时间缩短了3.36 s。GA-TD3算法模型在学习效率和通行效率上均有所提升，从而缓解城市中的交通压力，提高驾驶效率。
|
| 关 键 词： | 深度强化学习 无人驾驶决策 交叉路口 循环神经网络 注意力机制 |
| 收稿时间： | 2023-10-12 |
| 修稿时间： | 2024-06-04 |
Intersection decision model based on GA-TD3 algorithm |
| |
| jianganni,Duyu,yuanying,zhanghao and zhaoshixin. Intersection decision model based on GA-TD3 algorithm[J]. Application Research of Computers, 2024, 41(7) |
| |
| Authors: | jianganni Duyu yuanying zhanghao zhaoshixin |
| |
| Affiliation: | Beijing Union University,,,, |
| |
| Abstract: | Addressing problems such as low success rates, instability, and inefficient traffic flow in autonomous decision-making models at intersections, this study proposed enhancements to the TD3 algorithm through the GA-TD3(GRU attention twin delayed deep deterministic policy gradient) algorithm. Firstly, it introduced a memory module which using GRU neural network to improve the success rate of the decision model. Secondly, it introduced a social attention mechanism in the state space to focus on interactions with social vehicles. This mechanism ensured the stability of the model while improving the traffic efficiency of vehicles. After 20 000 rounds of training in the CARLA simulator, the TD3 algorithm achieves a success rate of 92.4%, while the success rate of the GA-TD3 algorithm is 97.6%. Additionally, the vehicle''s travel time is shortened by 3.36. GA-TD3 algorithm improves both learning efficiency and traffic efficiency, which can alleviate traffic pressure in urban scenes and improve driving efficiency. |
| |
| Keywords: | deep reinforcement learning autonomous driving decision intersection scenario recurrent neural network attention mechanism |
|
| 点击此处可从《计算机应用研究》浏览原始摘要信息 |
|
点击此处可从《计算机应用研究》下载免费的PDF全文 |
|
