Aiming at the problem that the existing meta-heuristic algorithm has slow convergence speed and large error in the process of optimizing the neural network to predict vehicle speed, a vehicle speed prediction method for expressway off-ramp based on IWOA-BiLSTM was proposed. Firstly, the Circle chaotic map was used to replace the randomly generated initial population in the whale optimization algorithm to increase the diversity and quality of the population.Secondly, the elite opposition-based learning strategy was used to improve the diversity of the individual 's preferred position and reduce the risk of the algorithm falling into local optimum and premature convergence.Finally, the cosine function was used to change the adaptive convergence factor and introduce the inertia weight strategy. On the premise of retaining the advantages of the whale optimization algorithm, the global search and local development capabilities of the algorithm were balanced. The simulation results show that, compared with the existing meta-heuristic algorithm and vehicle speed prediction model, the IWOA algorithm has significantly improved in terms of optimization accuracy, convergence rate and prediction accuracy.
基于浅层神经网络结构的车速预测方法,虽然其预测精度优于数理统计参数模型,但由于其仅能提取车速数据中基本的非线性关联特征,难以深度挖掘表征交通流数据的时空关联特征和演化规律。循环神经网络和门控循环单元等深度神经网络更适合车速预测这种非线性动态复杂的系统[4]。结合卷积神经网络和长短期记忆网络的深度学习模型(Convolutional neural network-long short term memory network,CNN-LSTM)[8]、基于模糊规则的鲸鱼优化双向长短时神经网络(Bidirectional long short-term memory,LSTM)[9]、鲸鱼优化Bi-LSTM[10],以及基于自适应噪声完全集成经验模式分解和LSTM的车速组合预测模型[11],因其不仅有效解决了数据样本训练过程中因内部梯度受时间步长影响导致梯度爆炸和消失的长程依赖问题[12],也避免了LSTM过分依靠经验得到超参数,导致模型预测精度和泛化能力不足的问题。
现有研究主要通过数理统计和深度学习方法探析车速的时空分布特征并对其进行预测,忽视了驾驶行为出现频数对车速预测的影响,以及现有元启发式算法存在易陷入局部最优和收敛不足的问题。无免费午餐(No free lunch,NFL)定理表明:没有任何一种元启发式算法拥有解决所有复杂工程应用优化问题的能力[13]。因此,本文以快速路出口匝道驾驶行为轨迹数据相关性分析为基础,采用Circle混沌映射、精英反向学习、余弦函数改变自适应收敛因子和引入惯性权重等混合策略改进鲸鱼优化算法,构建了基于改进鲸鱼优化算法优化双向长短期记忆网络(Improved whale optimization algorithm for optimizing bidirectional long short term memory networks,IWOA-BiLSTM)的快速路出口匝道车速预测模型,避免了现有模型忽视驾驶行为出现频数对车速预测的影响,以及现有元启发式算法易陷入局部最优和收敛不足,导致LSTM超参数寻优求解精度不足和运行时间较长造成车速预测应用研究的局限性。
HouL, PeiY, HeQ. A car following model in the context of heterogeneous traffic flow involving multilane following behavior[J].Physica A: Statistical Mechanics and its Applications, 2023, 632: No.129307.
ZhuJie-yu, MaYan-li. Real-time assessment method of multi-vehicle interaction risk in confluence area[J].Journal of Jilin University (Engineering Edition), 2022, 52(7): 1574-1581.
TedjopurnomoD A, BaoZ, ZhengB, et al. A survey on modern deep neural network for traffic prediction: trends, methods and challenges[J].IEEE Transactions on Knowledge and Data Engineering, 2020, 34(4): 1544-1561.
[6]
LeeS, LeeY I, ChoB. Short‐term travel speed prediction models in car navigation systems[J]. Journal of Advanced Transportation, 2006, 40(2): 122-139.
[7]
ZhangZ, WangY, ChenP, et al. Probe data-driven travel time forecasting for urban expressways by matching similar spatiotemporal traffic patterns[J]. Transportation Research Part C: Emerging Technologies, 2017, 85: 476-493.
[8]
ChenP, DingC, LuG, et al. Short-term traffic states forecasting considering spatial-temporal impact on an urban expressway[J]. Transportation Research Record, 2016, 2594(1): 61-72.
[9]
TangK, ChenS, CaoY, et al. Short-Term travel speed prediction for urban expressways: hybrid convolutional neural network models[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 23(3): 1829-1840.
[10]
GeorgeS, SantraA K. An improved long short‐term memory networks with Takagi‐Sugeno fuzzy for traffic speed prediction considering abnormal traffic situation[J]. Computational Intelligence, 2020, 36(3): 964-993.
[11]
ManneS, LydiaE L, PustokhinaI V, et al. An intelligent energy management and traffic predictive model for autonomous vehicle systems[J]. Soft Computing, 2021, 25(18): 11941-11951.
[12]
JiaX, ZhouW, YangH, et al. Short-term traffic travel time forecasting using ensemble approach based on long short‐term memory networks[J]. IET Intelligent Transport Systems, 2023, 17: 1262-1273.
[13]
ShuaiC, WangW C, XuG, et al. Short-term traffic flow prediction of expressway considering spatial influences[J]. Journal of Transportation Engineering, Part A: Systems, 2022, 148(6): No.04022026.
[14]
WolpertD H, MacreadyW G. No free lunch theorems for optimization[J].IEEE Transactions on Evolutionary Computation, 1997, 1(1): 67-82.
[15]
YuanJ, Abdel-AtyM, GongY, et al. Real-time crash risk prediction using long short-term memory recurrent neural network[J]. Transportation research record, 2019, 2673(4): 314-326.
[16]
MirjaliliS, LewisA. The whale optimization algorithm[J]. Advances in Engineering Software, 2016, 95: 51-67.
[17]
KaurG, AroraS. Chaotic whale optimization algorithm[J]. Journal of Computational Design and Engineering, 2018, 5(3): 275-284.
[18]
SeifZ, AhmadiM B. An opposition-based algorithm for function optimization[J]. Engineering Applications of Artificial Intelligence, 2015, 37: 293-306.
LiuYan-ran, MengQing-yu, GuoHong-yan, et al. Cycle car trajectory prediction with high-precision map fusion under the attention mode of map[J].Journal of Jilin University ( Engineering and Technology Edition ), 2023, 53(3): 792-801.
ZhangBing, ZhouDan-dan, SunJian, et al. Bus arrival time prediction model based on bidirectional long short-term memory network[J]. Transportation System Engineering and Information, 2023, 23(2): 148-160.