面向自动驾驶测试场景生成的博弈神经网络算法

doi:10.3778/j.issn.1002-8331.2307-0320

摘要/Abstract

摘要： 为提高自动驾驶车辆虚拟测试场景的可解释性和高风险场景覆盖度，提出了一种将博弈论与神经网络相结合的虚拟测试场景生成算法SIG-GAN（social interactive gaming-generative adversarial network）。以高速匝道合流场景为例，通过捕捉匝道汇入车辆与主车道行驶车辆的交互特征，构建汇入交互博弈模型。利用汇入数据获得车辆优先通行概率来计算博弈策略的纳什均衡求解，并融入S-GAN神经网络模型中进行轨迹生成。同时引入PICT（pairwise independent combinatorial testing）模型对观测区域交互车辆的真实轨迹进行组合测试，结合SIG-GAN算法生成大量具有现实博弈交互行为的高风险交互轨迹。通过与LSTM、S-LSTM、S-GAN等轨迹生成算法进行对比实验，结果显示：（1）模型较其他算法在3.2?s与4.8?s时域下，生成轨迹ADE平均下降25.30%、18.98%、7.02%，FDE平均下降17.33%、16.06%、7.65%，生成精度更为准确；（2）通过组合测试后生成轨迹数量为原始的150倍，覆盖度更高。生成轨迹与原轨迹的碰撞时间（time to collision，TTC）分别集中在1.057?7?s、3.513?5?s，场景风险程度更大，对自动驾驶汽车的虚拟场景强化测试具有实际意义。

关键词: 自动驾驶汽车, 场景生成, 博弈论, 生成对抗网络

Abstract: In order to improve the interpretability of virtual test scenarios for autonomous vehicles and the coverage of high-risk scenarios, a virtual test scenario generation algorithm combining game theory and neural network SIG-GAN (social interactive gaming-generative adversarial network) is proposed. Taking the high-speed ramp merging scenario as an example, a converging interaction game model is constructed by capturing the interaction characteristics of ramp converging vehicles and vehicles traveling in the main lane. The converging data are used to obtain the vehicle priority probability to calculate the Nash equilibrium solution of the game strategy, and are integrated into the S-GAN neural network model for trajectory generation. At the same time, PICT (pairwise independent combinatorial testing) model is introduced to combine the real trajectories of interacting vehicles in the observation area, which is combined with SIG-GAN algorithm to generate a large number of high-risk interaction trajectories with realistic game interaction behavior. Through the comparison experiment with LSTM, S-LSTM, S-GAN and other trajectory generation algorithms, the results show that: (1) The model generates trajectories with an average decrease of 25.30%, 18.98%, 7.02% in ADE and 17.33%, 16.06%, 7.65% in FDE compared with other algorithms in the time domains of 3.2 s and 4.8 s, and generates trajectories more accurately. (2) The number of generated trajectories after the combination test is 150 times of the original trajectories, with higher coverage. The TTC (time to collision) of the generated trajectory and the original trajectory is concentrated in 1.057 7 s and 3.513 5 s respectively, with a greater degree of scene risk, which is of practical significance for the virtual scene enhancement test of autonomous vehicles.

Key words: autonomous vehicle, scene generation, game theory, generative adversarial network

李文礼, 李超, 张祎楠, 宋越, 胡雄. 面向自动驾驶测试场景生成的博弈神经网络算法[J]. 计算机工程与应用, 2024, 60(22): 335-346.

LI Wenli, LI Chao, ZHANG Yinan, SONG Yue, HU Xiong. Game Neural Network Algorithm for Generating Autonomous Driving Test Scenarios[J]. Computer Engineering and Applications, 2024, 60(22): 335-346.

参考文献

[1] 中国电动汽车百人会, 腾讯自动驾驶, 中汽数据有限公司, 等. 中国自动驾驶仿真技术蓝皮书2020[R]. China EV100, 2020.
China EV100, Tencent Autopilot, CAC Data, et al. Blue book of autonomous driving simulation technology in China 2020[R]. China EV100, 2020.
[2] 蒋拯民, 党少博, 李慧云, 等. 自动驾驶汽车场景测试研究进展综述[J]. 汽车技术, 2022(8): 10-22.
JIANG Z M, DANG S B, LI H Y, et al. A survey on the research progress of scenario-based testing for autonomous vehicles[J]. Automotive Technology, 2022(8): 10-22.
[3] 舒红, 袁康, 修海林, 等. 自动驾驶汽车基础测试场景构建研究[J]. 中国公路学报, 2019, 32(11): 245-254.
SHU H, YUAN K, XIU H L, et al. Construction of basic test scenarios of automated vehicles[J]. China Journal of Highway and Transport, 2019, 32(11): 245-254.
[4] LI S S, WANG W S, MO Z B, et al. Cluster naturalistic driving encounters using deep unsupervised learning[C]//Proceedings of the 2018 IEEE Intelligent Vehicles Symposium, Changshu, 2018: 1354-1359.
[5] 郭景华, 李克强, 王进, 等. 基于危险场景聚类分析的前车随机运动状态预测研究[J]. 汽车工程, 2020, 42(7): 847-853.
GUO J H, LI K Q, WANG J, et al. Study on prediction of preceding vehicles stochastic motion based on risk scenarios clustering analysis[J]. Automotive Engineering, 2020, 42(7): 847-853.
[6] 朱宇, 赵祥模, 徐志刚, 等. 基于蒙特卡洛模拟的无人车高速公路变道虚拟测试场景自动生成算法[J]. 中国公路学报, 2022, 35(3): 89-100.
ZHU Y, ZHAO X M, XU Z G, et al. Automatic generation algorithm of lane-change virtual test scenario on highways for automated vehicles using Monte Carlo simulation[J]. China Journal of Highway and Transport, 2022, 35(3): 89-100.
[7] YAN X, FENG S, SUN H, et al. Distributionally consistent simulation of naturalistic driving environment for autonomous vehicle testing[J]. arXiv:2101.02828, 2021.
[8] JUNIETZ P, WACHENFELD W, KLONECKI K, et al. Evaluation of different approaches to address safety validation of automated driving[C]//Proceedings of the 2018 21st International Conference on Intelligent Transportation Systems, Maui, 2018: 491-496.
[9] TUNCALI C E, PAVLIC T P, FAINEKOS G. Utilizing S-TaLiRo as an automatic test generation framework for autonomous vehicles[C]//Proceedings of the 2016 IEEE 19th International Conference on Intelligent Transportation Systems, Rio de Janeiro, 2016: 1470-1475.
[10] 周文帅, 朱宇, 赵祥模, 等. 面向高速公路车辆切入场景的自动驾驶测试用例生成方法[J]. 汽车技术, 2021(1): 11-18.
ZHOU W S, ZHU Y, ZHAO X M, et al. Vehicle cut-in test case generation methods for testing of autonomous driving on highway[J]. Automotive Technology, 2021(1): 11-18.
[11] KOREN M, ALSAIF S, LEE R, et al. Adaptive stress testing for autonomous vehicles[C]//Proceedings of the 2018 IEEE Intelligent Vehicles Symposium, Changshu, 2018: 1-7.
[12] 李江坤, 邓伟文, 任秉韬, 等. 基于场景动力学和强化学习的自动驾驶边缘测试场景生成方法[J]. 汽车工程, 2022, 44(7): 976-986.
LI J K, DENG W W, REN B T, et al. Automatic driving edge scene generation method based on scene dynamics and reinforcement learning[J]. Automotive Engineering, 2022, 44(7): 976-986.
[13] 李晓佳, 赵国生, 汪洋, 等. 面向CNN和RNN改进的物联网入侵检测模型[J]. 计算机工程与应用, 2023, 59(14): 242-250.
LI X J, ZHAO G S, WANG Y, et al. Improved Internet of things intrusion detection model for CNN and RNN[J]. Computer Engineering and Applications, 2023, 59(14): 242-250.
[14] 黄思远, 赵宇海, 梁燚铭. 融合图嵌入和注意力机制的代码搜索[J]. 计算机科学与探索, 2022, 16(4): 844-854.
HUANG S Y, ZHAO Y H, LIANG Y M. Code search combining graph embedding and attention mechanism[J]. Journal of Frontiers of Computer Science and Technology, 2022, 16(4): 844-854.
[15] 李文礼, 韩迪, 石晓辉, 等. 基于时-空注意力机制的车辆轨迹预测[J]. 中国公路学报, 2023, 36(1): 226-239.
LI W L, HAN D, SHI X H, et al. Vehicle trajectory prediction based on spatial-temporal attention mechanism[J]. China Journal of Highway and Transport, 2023, 36(1): 226-239.
[16] ALAHI A, GOEL K, RAMANATHAN V, et al. Social LSTM: human trajectory prediction in crowded spaces[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition, New York, 2016: 961-971.
[17] GUPTA A, JOHNSON J, LI F F, et al. Social GAN: socially acceptable trajectories with generative adversarial networks[C]//Proceedings of the 2018 IEEE Conference on Computer Vision and Pattern Recognition, New York, 2018: 2255-2264.
[18] KRAJEWSKI R, MOERS T, NERGER D, et al. Data-driven maneuver modeling using generative adversarial networks and variational autoencoders for safety validation of highly automated vehicles[C]//Proceedings of the 2018 21st IEEE International Conference on Intelligent Transportation Systems, New York, 2018: 2383-2390.
[19] SHEN D, YI Q, LI L X, et al. Test scenarios development and data collection methods for the evaluation of vehicle road departure prevention systems[J]. IEEE Transactions on Intelligent Vehicles, 2019, 4(3): 337-352.
[20] 王靓喆. 基于博弈论的智能网联自动驾驶车辆换道行为研究[D]. 长春: 吉林大学, 2022.
WANG L Z. Research on lane-changing behavior of intelligent autonomous connected-vehicles based on game theory[D]. Changchun: Jilin University, 2022.
[21] ARBIS D, DIXIT V V. Game theoretic model for lane changing: incorporating conflict risks[J]. Accident Analysis & Prevention, 2019, 125: 158-164.
[22] 王军, 曹雷, 陈希亮, 等. 纯策略纳什均衡的博弈强化学习[J]. 计算机工程与应用, 2022, 58(15): 78-86.
WANG J, CAO L, CHEN X L, et al. Game reinforcement learning of pure strategy Nash equilibrium[J]. Computer Engineering and Applications, 2022, 58(15): 78-86.