基于多目标蚁群算法的共享单车调度优化方法

薛晴婉; 瞿麦青; 彭怀军; 姚运梅; 郭伟伟; 谭墍元; 王云

doi:10.3963/j.jssn.1674-4861.2024.02.013

基于多目标蚁群算法的共享单车调度优化方法

doi: 10.3963/j.jssn.1674-4861.2024.02.013

1.
北方工业大学电气与控制工程学院北京 100144
2.
亿雅捷交通系统（北京）有限公司北京 100101
3.
北京交通大学交通运输学院北京 100044

基金项目:

国家自然科学基金项目 72371020

北京市教委科技一般项目 KM202310009009

北京京投卓越科技发展有限公司科研课题 BIITT-2022-09

详细信息

作者简介:
薛晴婉（1991—），博士，讲师. 研究方向：交通规划、交通安全. E-mail：qwxue@ncut.edu.cn

通讯作者:
王云（1991—），博士，副教授. 研究方向：公共交通规划、运筹优化. E-mail：wang.yun@bjtu.edu.cn

中图分类号: U121
计量
- 文章访问数: 43
- HTML全文浏览量: 29
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2023-09-08
- 网络出版日期: 2024-09-14

A Scheduling Optimization Method of Shared Bicycles Based on a Multi-objective Ant Colony Algorithm

1.
School of Electrical and Control Engineering, North China University of Technology, Beijing 100044, China
2.
BII Transit System (Beijing) Co., Ltd, Beijing 100101, China
3.
School of Traffic and Transportation, Beijing Jiaotong University, Beijing 100044, China

摘要

摘要: 共享单车作为公共交通接驳、“最后一公里”出行的重要交通工具，存在供需时空不匹配的问题，需要利用调度车实现共享单车的再平衡。针对部分现有共享单车调度方法存在的优化目标单一、调度点只能被访问1次、未考虑连续调度衔接等问题，建立了以总需求不满足度最小和调度成本最小为目标的多目标优化模型。该模型考虑高峰小时调度点需求远大于调度车容量的情况，允许多辆调度车多时段连续调度，且允许调度车重复访问调度点。设计了多目标蚁群算法进行求解，引入非支配排序方法，将解集划分为不同的非支配层级，取最高层级的解，形成1组同时考虑2个目标的Pareto最优解。该算法引入了最大-最小蚂蚁系统，改进了状态转移概率规则和信息素更新规则，使其能够适用于求解多目标优化问题。算例结果表明，该模型能够在保证较低调度成本的同时，减少需求损失，算例调度后的总需求不满足度由不进行调度时的26.48%降低到17.86%。将不同算例规模下多目标蚁群算法与贪心算法求解结果进行比较，多目标蚁群算法在多时段连续调度问题上具有优势，能够统筹安排每辆调度车在每个调度周期的行驶路径和在各调度点的到达时间和共享单车装卸数量。多目标蚁群算法所求得的解的质量优于贪心算法，较大规模算例求解得到的调度成本和总需求不满足度比贪心算法分别降低了62%和23%。
- 城市交通 /
- 共享单车调度 /
- 多目标优化 /
- 蚁群算法
Abstract: As a crucial mode for facilitating public transportation connections and addressing the "last mile" problem, shared bicycles confront the challenge of supply and demand imbalances. To solve this issue, deploying vehicles for scheduling purposes becomes an essential step in rebalancing the shared bicycles. In order to address the issues of current shared bicycle scheduling methods including single optimization objective, limited visits to scheduling sites, and insufficient consideration of continuous scheduling connections, a multi-objective optimization model is developed in this paper to minimize both total demand dissatisfaction and scheduling costs. This model considers the situation that the demand at the scheduling site surpasses the capacity of the scheduling vehicle during peak hours. Consequently, it enables the scheduling vehicle to make multiple trips to the site and allows to conduct continuous scheduling in multiple periods of time for multiple vehicles. A multi-objective ant colony algorithm is designed to solve this model by integrating the technique of non-dominated sorting to classify the solution set into various levels of non-dominance. The solution at the highest level is then utilized to create a Pareto-optimal solution, which considers two objectives concurrently. This algorithm introduces a new ant system incorporating maximum-minimum criteria, modifies the state transition probability rule and pheromone update rule to enhance their efficacy to deal with the multi-objective optimization problem. In order to verify the feasibility of the model and algorithm, a case study is carried out. The results show that the model is confirmed to be effective in decreasing demand loss while ensuring the lower scheduling costs. Specifically, the total demand dissatisfaction degree is reduced from 26.48% to 17.86%. Comparing the results of the multi-objective ant colony algorithm and greedy algorithm under various example sizes, the multi-objective ant colony algorithm shows a clear superiority in continuous scheduling of multiple periods of time. Specifically, it is capable of organizing the driving path of each scheduling vehicle in each scheduling cycle, as well as the arrival time and the loading and unloading numbers of shared bicycles at each scheduling site. Meanwhile, compared with greedy algorithm, the multi-objective ant colony algorithm shows a clear superiority in the quality of the solutions, and the scheduling costs and total demand dissatisfaction degree obtained in a large-scale case are reduced by 62% and 23%, respectively.
- urban traffic /
- shared bicycles scheduling /
- multi-objective optimization /
- ant colony algorithm

HTML全文

图 1 共享单车调度示意图

Figure 1. Shared bicycles scheduling schematic diagram

下载: 全尺寸图片幻灯片

图 2 多目标蚁群算法流程图

Figure 2. Diagram chart of multi-objective ant colony algorithm

下载: 全尺寸图片幻灯片

图 3 调度点位置分布

Figure 3. Scheduling center location distribution

下载: 全尺寸图片幻灯片

图 4 不同α的Pareto最优前沿

Figure 4. Pareto optimal front of different α

下载: 全尺寸图片幻灯片

图 5 不同β的Pareto最优前沿

Figure 5. Pareto optimal front of different β

下载: 全尺寸图片幻灯片

图 6 不同ρ的Pareto最优前沿

Figure 6. Pareto optimal front of different ρ

下载: 全尺寸图片幻灯片

图 7 不同种群数量p的Pareto最优前沿

Figure 7. Pareto optimal front of different p

下载: 全尺寸图片幻灯片

图 8 不同种群数量p的迭代收敛曲线

Figure 8. Iterative convergence curve of different p

下载: 全尺寸图片幻灯片

图 9 调度车容量敏感性分析

Figure 9. Sensitivity analysis of scheduling vehicle capacity

下载: 全尺寸图片幻灯片

图 10 单位距离成本敏感性分析

Figure 10. Sensitivity analysis of unit distance cost

下载: 全尺寸图片幻灯片

图 11 调度车数量敏感性分析

Figure 11. Sensitivity analysis of scheduling vehicle quantity

下载: 全尺寸图片幻灯片

图 12 Pareto最优前沿

Figure 12. Pareto optimal front

下载: 全尺寸图片幻灯片

图 13 迭代收敛曲线

Figure 13. Iterative convergence curve

下载: 全尺寸图片幻灯片

图 14 Iterative convergence curve

Figure 14. The scheduling route of the scheduling vehicles in each cycle

下载: 全尺寸图片幻灯片

表 1 相关文献对比

Table 1. Comparison of relevant literature

研究者	静态/动态调度	目标函数	求解算法	是否是多目标优化	是否允许多次访问调度点	多时段调度
Raviv等^[3]	静态	总调度成本最小、总惩罚成本最小	启发式算法	否	是	是
Erdoğan^[4]	静态	路线总成本最小	分离算法	否	是	否
蒋塬锐等^[5]	静态	调度总成本最小	遗传算法	否	是	否
关宏志等^[6]	静态	调度时间成本最小	求解器	否	否	否
于德新^[14]	静态	成本最小、投放率最高	改进遗传算法	是	否	否
曾琼燕等^[9]	动态	总利润最大	模拟退火算法	否	否	是
Zhang^[10]	动态	车辆出行成本最小、用户不满足度最小	启发式算法	否	否	是
Hu等^[11]	动态	整体满意度利润最小、再平衡成本最小	MOEA/D	是	否	是
汪慎文等^[19]	动态	调度量最大	改进蚁群算法	否	是	否
本文	动态	调度成本最小、总需求不满足度最小	多目标蚁群算法	是	是	是

下载: 导出CSV

表 2 模型符号含义

Table 2. Model symbolic meaning

	符号	含义
集合	V	调度点集合
集合	K	调度车集合
索引	i	调度点索引，i ∈ V
	j	调度点索引，j ∈ V
	k	调度车索引，k ∈ K
	t	调度周期索引，将调度计划期划分为m个调度周期，t = 1，2，…，m
参数	T	调度周期的时间长度
	C	调度车容量
	l_ij	从调度点i到调度点j的距离
	I_i(t)	调度点i在周期t的骑入量
	O_i(t)	调度点i在周期t的需求满足量
	v	车辆速度
	g	共享单车的单位装卸时间
	w	单位距离成本
决策变量	x_ij^k(t)	二元变量，如果x_ij^k(t) = 1，则车辆k在周期t从调度点i行驶到调度点j，否则为0
	y_ij^k(t)	当车辆k在周期t从调度点i行驶到调度点j时，车辆k上携带的共享单车数量。如果车辆k不从i行驶到j，则y_ij^k(t) 为零
	z_i^k(t)	车辆k在周期t在调度点i的共享单车装卸数量，z_i^k(t) < 0表示卸车，z_i^k(t) > 0表示装车
辅助变量	S_i(t)	调度点i在周期t的初始供给量
辅助变量	d_i(t)	调度点i在周期t的需求量

下载: 导出CSV

表 3 调度中心位置

Table 3. Scheduling center location

调度中心	经度/（°）	纬度/（°）
0	116.534 107	39.974 810

下载: 导出CSV

表 4 调度点位置和需求

Table 4. Scheduling site location and requirements

调度点	经度/（°）	纬度/（°）	初始存量/辆	周期1		周期2		周期3		周期4
调度点	经度/（°）	纬度/（°）	初始存量/辆	骑入量/辆	需求量/辆	骑入量/辆	需求量/辆	骑入量/辆	需求量/辆	骑入量/辆	需求量/辆
1	116.534 107	39.974 810	200	531	668	657	833	417	367	758	611
2	116.534 568	39.974 349	100	27	91	30	101	50	13	100	39
3	116.539 664	39.973 239	80	18	50	20	56	36	15	73	25
4	116.539 364	39.977 137	10	20	15	43	16	8	25	15	60
5	116.543 248	39.974 374	65	20	66	22	74	20	40	96	29
6	116.544 986	39.976 873	5	30	11	33	12	7	20	13	52
7	116.544 342	39.971 858	5	25	9	27	10	7	30	13	40
8	116.548 559	39.976 018	5	49	18	55	20	20	20	30	71
9	116.548 119	39.973 626	5	40	16	45	18	12	50	23	58
10	116.547 883	39.971 455	80	33	96	36	107	8	20	139	47
11	116.548 955	39.980 639	45	29	65	32	72	20	40	94	42
12	116.550 758	39.978 156	5	71	40	79	44	23	62	45	103
13	116.552 174	39.973 034	76	26	88	29	98	64	10	127	38
14	116.552 314	39.974 859	5	56	55	71	61	10	70	70	92
15	116.554 674	39.977 934	5	80	76	100	80	49	78	100	130
16	116.554 551	39.973 490	25	8	19	8	21	14	35	27	11
17	116.558 569	39.971 348	50	899	722	1 108	896	433	656	865	1 048
18	116.556 938	39.975 352	50	9	28	9	32	30	9	50	12
19	116.558 815	39.977 613	16	18	28	20	28	10	28	20	30
20	116.558 976	39.980 351	57	11	34	12	37	30	8	50	16
21	116.561 648	39.978 748	5	85	27	95	30	18	86	30	123
22	116.560 865	39.982 875	70	16	46	18	51	36	5	60	23

下载: 导出CSV

表 5 各周期调度车的调度路径、装卸量和到达时间

Table 5. The scheduling route, loading and unloading volume and arrival time of the scheduling vehicles in each cycle

车辆编号	周期1			周期2			周期3
车辆编号	路径	装卸车数量/辆	到达时间/s	路径	装卸车数量/辆	到达时间/s	路径	装卸车数量/辆	到达时间/s
车辆1	0	0	0	8	4	0	6	-37	0
	3	12	76	9	-11	155	2	37	1 173
	4	15	530	8	11	520	4	-40	2 340
	5	-27	1 026	11	-20	941
	7	21	1 889	5	-20	1 620
	11	-21	2 576	4	10	2 266
	8	36	3 297	6	30	2 623
车辆2	0	0	0	13	0	0	19	-10	0
	9	29	182	17	30	58	18	28	344
	13	-29	1 144	14	-40	1 025	17	-37	1 238
	12	36	2 045	17	40	2 292	13	37	2 406
	13	-26	3 156	19	-18	3 590	12	-40	3 547

下载: 导出CSV

表 6 不同算例规模下2种算法求解结果比较

Table 6. Comparison of results between two algorithms under different example sizes

指标	算例22		算例36
指标	多目标蚁群算法	贪心算法	多目标蚁群算法	贪心算法
目标函数	14.82	22.58	16.19	31.09
调度成本/元	11.67	25.40	14.92	39.51
总需求不满足度/%	17.96	19.76	17.46	22.67
运行时间/s	81.00	0.13	213.67	0.42

下载: 导出CSV

参考文献(20)

[1]	WANG Z L, CHEN F, XU T K. Interchange between metro and other modes: access distance and catchment area[J]. Journal of Urban Planning and Development, 2016, 142(4): 04016012. doi: 10.1061/(ASCE)UP.1943-5444.0000330
[2]	尹芹, 孟斌, 张丽英. 基于客流特征的北京地铁站点类型识别[J]. 地理科学进展, 2016, 35(1): 126-134. YIN Q, MENG B, ZHANG L Y. Classification of subway stations in Beijing based on passenger flow characteristics[J]. Progress in Geography, 2016, 35(1): 126-134. (in Chinese)
[3]	RAVIV T, TZUR M, FORMA I A. Static repositioning in a bike-sharing system: models and solution approaches[J]. EURO Journal on Transportation and Logistics, 2013, 2(3): 187-229. doi: 10.1007/s13676-012-0017-6
[4]	ERDOĞAN G, BATTARRA M, CALVO R W. An exact algorithm for the static rebalancing problem arising in bicycle sharing systems[J]. European Journal of Operational Research, 2015, 245(3): 667-679. doi: 10.1016/j.ejor.2015.03.043
[5]	蒋塬锐, 贾顺平, 李军. 基于调度池的共享单车调度研究[J]. 交通信息与安全, 2019, 37(5): 124-132. doi: 10.3963/j.issn.1674-4861.2019.05.016 JIANG Y R, JIA S P, LI J. A study of bicycle-sharing scheduling based on scheduling pool[J]. Journal of Transport Information and Safety, 2019, 37(5): 124-132. (in Chinese) doi: 10.3963/j.issn.1674-4861.2019.05.016
[6]	关宏志, 卢笙, 宋茂灿. 共享单车分层调度策略研究[J]. 重庆交通大学学报(自然科学版), 2020, 39(2): 1-7. GUAN H Z, LU S, SONG M C. Hierarchical scheduling strategy for free-floating bike-sharing[J]. Journal of Chongqing Jiaotong University (Natural Science), 2020, 39(2): 1-7. (in Chinese)
[7]	高楹, 宋辞, 舒华, 等. 北京市摩拜共享单车源汇时空特征分析及空间调度[J]. 地球信息科学学报, 2018, 20(8): 1123-1138. GAO Y, SONG C, SHU H, et al. Spatial-temporal characteristics of source and sink points of mobikes in Beijing and its scheduling strategy[J]. Journal of Geo-information Science, 2018, 20(8): 1123-1138. (in Chinese)
[8]	谢青成, 毛嘉莉, 刘婷. 城市共享单车的动态调度策略[J]. 华东师范大学学报(自然科学版), 2019, (6): 88-102. doi: 10.3969/j.issn.1000-5641.2019.06.009 XIE Q C, MAO J L, LIU T. Dynamic scheduling strategy for bicycle-sharing in cities[J]. Journal of East China Normal University (Natural Science), 2019, (6): 88-102. (in Chinese) doi: 10.3969/j.issn.1000-5641.2019.06.009
[9]	曾琼燕, 杨晟. 基于模拟退火算法的共享单车动态调度问题研究[J]. 综合运输, 2023, 45(2): 75-79. ZENG Q, YANG S. Dynamic scheduling of shared bicycles based on simulated annealing algorithm[J]. China Transportation Review, 2023, 45(2): 75-79. (in Chinese)
[10]	ZHANG D, YU C, DESAI J, et al. A time-space network flow approach to dynamic repositioning in bicycle sharing systems[J]. Transportation Research Part B: Methodological, 2017, 103: 188-207. doi: 10.1016/j.trb.2016.12.006
[11]	HU R, ZHANG Z, MA X, et al. Dynamic rebalancing optimization for bike-sharing system using priority-based MOEA/D algorithm[J]. IEEEAccess, 2021, 9: 27067-27084.
[12]	SHUI C S, SZETO W Y. Dynamic green bike repositioning problem-a hybrid rolling horizon artificial bee colony algorithm approach[J]. Transportation Research Part D: Transport and Environment, 2018, 60: 119-136.
[13]	PAL A, ZHANG Y. Free-floating bike sharing: Solving real-life large-scale static rebalancing problems[J]. Transportation Research Part C: Emerging Technologies, 2017, 80: 92-116.
[14]	于德新, 张行, 王薇, 等. 共享单车调度模型及算法研究[J]. 重庆交通大学学报(自然科学版), 2020, 39(7): 1-7. YU D X, ZHANG H, WANG W, et al. Scheduling model and algorithm for shared bicycle[J]. Journal of Chongqing Jiaotong University (Natural Science), 2020, 39(7): 1-7. (in Chinese)
[15]	孙卓, 李一鸣. 考虑多仓库的共享单车重新配置问题研究[J]. 运筹与管理, 2021, 30(1): 121-129. SUN Z, LI Y M. Solving a static bike repositioning problem with multiple depots[J]. Operations Research and Management Science, 2021, 30(1): 121-129. (in Chinese)
[16]	徐国勋, 张伟亮, 李妍峰. 共享单车调配路线优化问题研究[J]. 工业工程与管理, 2019, 24(1): 80-86. XU G X, ZHANG W L, LI Y F. Research on optimization of shared bicycle repositioning routing problem[J]. Industrial Engineering and Management, 2019, 24(1): 80-86. (in Chinese)
[17]	SHI L, ZHANG Y, RUI W N, et al. Study on the bike-sharing inventory rebalancing and vehicle routing for bike-sharing system[J]. Transportation Research Procedia, 2019, 39: 624-633.
[18]	CAGGIANI L, CAMPOREALE R, OTTOMANELLI M, et al. A modeling framework for the dynamic management of free-floating bike-sharing systems[J]. Transportation Research Part C: Emerging Technologies, 2018, 87: 159-182.
[19]	汪慎文, 徐亮, 杨锋, 等. 基于蚁群算法的动态共享单车调度优化[J]. 南昌工程学院学报, 2019, 38(3): 71-76. WANG S W, XU L, YANG F, et al. Optimization of dynamic shared bike scheduling based on ant colony[J]. Journal of Nanchang Institute of Technology, 2019, 38(3): 71-76. (in Chinese)
[20]	ZHAO B L, GUI H X, LI H Z, et al. Cold chain logistics path optimization via improved multi-objective ant colony algorithm[J]. IEEE Access, 2020, (8): 142977-142995.