基于浮动车轨迹的城市交通拥堵评估与预测

发布时间：2018-04-24 13:50

本文选题：浮动车轨迹 + 模糊综合评价　；参考：《大连理工大学》2014年硕士论文

【摘要】：交通拥堵问题,严重影响市民日常生活,在一定程度上限制了社会、经济稳定发展。缓解交通拥堵,尤其是常发性拥堵,已迫在眉睫。缓解交通拥堵的重要前提是交通拥堵评估与预测,但是现有方法在准确性、实时性和稳定性三方面的性能不能满足交通需求。为了提高拥堵评估与预测的性能,本文提出基于浮动车轨迹的拥堵评估与预测方法。对于评估方面,本文提出一种基于多指标权重适应变化的模糊综合评价法。它通过对指标赋权和多指标模糊矩阵变换进行综合评价,根据拥堵交通流特性给指标适应赋权,比以往固定权重法,能够提高评估准确率和实时性能。对于预测方面,本文提出一种基于优化SVM (Support Vector Machine)的拥堵预测方法,它包括VP(Volume Predict)模块、SP (Speed Predict)模块、优化模块以及拥堵状态划分模块。VP模块和SP模块用于预测交通量和平均速度,优化模块对VP模块和SP模块中SVM的惩罚系数以及多个核函数参数进行优化,拥堵状态划分模块将预测的交通流参数转化为市民所认知的拥堵状态。它核心算法是粒子群优化算法PSO (Particle Swarm Optimization)和SVM, PSO计算复杂度低,结合SVM不同核函数有不同的预测精度和拟合能力,能在最短时间内找到固定最优解,可满足预测准确率、实时性、稳定性。最后,本文对提出的交通拥堵评估和预测方法进行仿真验证。实验内容分为两部分：本文提出的评估方法与指标固定赋权的模糊综合评价法的对比和本文提出的预测优化方法与经典优化算法的对比。实验证明本文提出的评估与预测方法在准确率、实时性和稳定性上均具有优势。
[Abstract]:The problem of traffic jam seriously affects the daily life of citizens and restricts the stable development of society and economy to a certain extent. It is urgent to ease traffic congestion, especially frequent congestion. The important premise of alleviating traffic congestion is traffic congestion evaluation and prediction, but the performance of existing methods in accuracy, real-time and stability can not meet the traffic demand. In order to improve the performance of congestion assessment and prediction, a new method based on floating vehicle trajectory is proposed in this paper. In the aspect of evaluation, a fuzzy comprehensive evaluation method based on multi-index weight adaptation is proposed in this paper. Through the comprehensive evaluation of index weighting and multi-index fuzzy matrix transformation, the index can be weighted according to the traffic congestion characteristics. Compared with the previous fixed weight method, it can improve the accuracy and real-time performance of the evaluation. In the aspect of prediction, a congestion prediction method based on optimized SVM support Vector Machine is proposed. It includes VP(Volume predictor module, optimization module and congestion partition module. VP module and SP module are used to predict traffic volume and average speed. Optimization module optimizes the penalty coefficient of SVM and several kernel function parameters in VP module and SP module. The congestion state partition module converts the predicted traffic flow parameters into the congestion state known by citizens. Its core algorithms are particle swarm optimization (PSO) and particle Swarm optimization (SVM). PSO has low computational complexity. Combining with different kernel functions of SVM, it has different prediction accuracy and fitting ability, and can find the fixed optimal solution in the shortest time. Stability. Finally, the proposed traffic congestion assessment and prediction methods are verified by simulation. The experiment is divided into two parts: the comparison between the evaluation method proposed in this paper and the fuzzy comprehensive evaluation method with fixed index weighting, and the comparison between the proposed prediction optimization method and the classical optimization algorithm. Experiments show that the proposed evaluation and prediction method has advantages in accuracy, real-time and stability.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U491.265

【参考文献】