论文部分内容阅读
提出了一种基于状态空间神经网络(SSNN)和拓展卡尔曼滤波(EKF)的混合式行程时间预测模型.以往的研究表明,状态空间神经网络能够较好地处理复杂的非线性时空问题.然而,状态空间神经网络需要大量的历史数据作为离线训练之用.其不足之处在于,首先是需要花费大量的时间和精力去收集、准备数据,以及训练神经网络.其次,输入输出随着时间不断增加,训练过程需要不断的从新重复.为了提高状态空间神经网络的有效性,扩展卡尔曼滤波代替了传统的方法来对神经网络进行训练.荷兰的一条城市道路被选择为模型验证的试验路段.通过与另外两个预测模型之间的对比验证,该模型的预测能力能够达到满意的有效性、准确性和鲁棒性.
A hybrid travel time prediction model based on state space neural network (SSNN) and extended Kalman filter (EKF) is proposed in this paper.Previous researches show that state space neural network can deal with complex nonlinear time-space problems better.However, , The state space neural network requires a lot of historical data for offline training.The disadvantage is that firstly it takes a lot of time and energy to collect, prepare data and train neural networks.Secondly, input and output over time In order to improve the effectiveness of neural network in state space, extended Kalman filter instead of the traditional method to train the neural network.An urban road in the Netherlands was selected as a model validation test section. Through the comparison with the other two prediction models, the predictive ability of the model can reach satisfactory validity, accuracy and robustness.