﻿ 一种基于历史轨迹数据的实时公交到站预测算法 A Real-Time Bus Arrival Time Prediction Algorithm Based on Historical GPS

Open Journal of Transportation Technologies
Vol.07 No.04(2018), Article ID:25767,7 pages
10.12677/OJTT.2018.74028

A Real-Time Bus Arrival Time Prediction Algorithm Based on Historical GPS

Jiandong Qiu1, Shaoxuan Dong2, Qiang Li2

1Shenzhen Urban Transport Planning Center Co., Ltd., Shenzhen Guangdong

2Traffic Information Engineering & Technology Research Center of Guangdong Province, Shenzhen Guangdong

Received: Jun. 18th, 2018; accepted: Jun. 29th, 2018; published: Jul. 6th, 2018

ABSTRACT

Real-time bus arrival time prediction is an important part of smart transit service. Reliable predicting results can help passengers to reduce waiting time and plan trip reasonably. Common algorithms have many problems, such as many kinds of data needed, complex models, single application scenarios and high complexity of algorithms, this paper presents a real-time bus arrival time prediction algorithm based on historical GPS, providing passengers with remaining travel time, number of stations and distance. And an Android application accessing real-time GPS data in Shenzhen City is developed to verify the accuracy and simplicity of the proposed algorithm to meet the needs of large-scale application.

Keywords:Transit, Real-Time Arrival Time Prediction, GPS

1深圳市城市交通规划设计研究中心有限公司，广东 深圳

2广东省交通信息工程技术研究中心，广东 深圳

Copyright © 2018 by authors and Hans Publishers Inc.

1. 研究背景

2. 算法思路

Figure 1. Line interruption

2.1. 历史数据处理

1) 数据清洗：对历史轨迹数据中位置异常和重复、无效记录进行清洗，将部分有信息缺失的记录根据其他记录进行补齐；

2) 轨迹匹配：根据轨迹点位置和方向角信息判断车辆运行方向，并匹配至公交线路最邻近点上(图2)；

3) 到站时间插值：搜索站点邻近轨迹点，对历史到站时间进行插值。为了简化估计过程，本文假设相邻轨迹点间车辆匀速运行，从而根据距离插值得到站点的到达时间。如图3所示，A、B为相邻两个轨迹点，记录时间分别为 ${t}_{A}$${t}_{B}$ ，中间经过站点x，则到达x的时间

${t}_{x}={t}_{A}+\left({t}_{B}-{t}_{A}\right)\ast {d}_{Ax}/{d}_{AB}$

4) 路段行程时间：计算物理区间内行程时间，并根据工作日、非工作日，以1个小时为粒度求得各时间段的平均行程时间 ${t}_{AB}^{i}$ ，i为第i个时间段(图4)。

2.2. 实时预测

${t}_{{P}^{\prime },i+1}={t}_{i,i+1}\ast {d}_{{P}^{\prime },i+1}/{d}_{i,i+1}$

${T}_{n}={t}_{{P}^{\prime },i+1}+{t}_{i,i+1}+\cdots +{t}_{n-1,n}$

3. 算法应用

(a) 匹配前 (b) 匹配后

Figure 2. Schematic diagram of trace matching

Figure 3. Arrival time interpolation diagram

Figure 4. Historical data processing

Figure 5. Real-time location acquisition

Figure 6. User interface

3.1. 数据表设计

3.2. 算法效率分析

3.3. 实测验证

4. 总结与展望

1) 对于行程时间预测的精细度有待进一步提高，加入车辆停靠时间、信号灯导致的延迟等因素 [5] ；

2) 对于异常突发事件如事故、台风天气导致的车辆延迟情况，需要在算法中加入异常处理模块。

Table 1. Data display example

Figure 7. Algorithm efficiency analysis

A Real-Time Bus Arrival Time Prediction Algorithm Based on Historical GPS[J]. 交通技术, 2018, 07(04): 222-228. https://doi.org/10.12677/OJTT.2018.74028

1. 1. 沈旗. 基于浮动公交车的交通信息实时检测和公交车到站时间预测研究[D]: [硕士学位论文]. 杭州: 浙江大学, 2007.

2. 2. 向红艳, 彭学文. 公交到站时间预测研究现状与发展趋势[J]. 交通信息与安全, 2014, 32(4): 57-61.

3. 3. 王锐, 杨宏业, 张海鹏. 基于浮动公交车的SVM实时交通状态识别研究[J]. 内蒙古工业大学学报, 2016, 35(1): 18-23.

4. 4. 胡华, 高云峰, 刘志钢. 基于AVL数据的公交到站时间实时预测模型[J]. 重庆交通大学学报(自然科学版), 2012, 31(5): 1014-1017.

5. 5. 孙棣华, 赖云波, 廖孝勇, 等. 公交浮动车辆到站时间实时预测模型[J]. 交通运输工程学报, 2011(2): 84-89.