时间和需求不确定下的里程共享动态匹配问题研究

Dynamic Matching of Ride-sharing under Time and Demand Uncertainty

随着里程共享模式逐渐在人群中流行普及,与之相关的匹配方案设计、路径优化等成为国内外学者的研究热点。文章考虑乘客需求不确定性和车辆行程时间不确定性,综合司机收益、系统服务水平和系统碳排放3个目标,构建多目标里程共享动态匹配模型。运用鲁棒优化的方法处理不确定参数,采用传统的线性加权和的方法将多目标规划问题转化为单目标规划。将模型中含有不确定时间参数的约束转化为机会约束,并进一步使用Hoeffding方法将机会约束安全近似为线性约束,最终完成模型求解和算例分析。结果表明,不确定因素的增多会对实际情况中的里程共享服务产生更多的负面影响,系统总收益随之减少,总碳排放随之增加。此外,在不确定因素的条件一定时,随着置信水平和不确定水平的增加,系统总收益减少,总碳排放增加。因此,在实际决策过程中,要在系统收益、系统碳排放和解的鲁棒性之间进行权衡。

With the increasing popularity of riding-sharing(RS),matching scheme design and route optimization have attracted the attention of academic researchers.This study develops a multi-objective ride-sharing dynamic matching model combining three goals which are driver’s revenue,system service level and system carbon emission by considering the uncertainty of travel demand and travel time.Specifically,robust optimization is used to deal with uncertain demand parameters in the model,and the multi-objective programming problem is transformed into single-objective programming by using traditional linear weighted method.Meanwhile,the constraints of the model with uncertain time parameters are transformed into chance constraints,and the Hoeffding method is used to approximate the chance constraint as a linear constraint.The results indicate that increasing uncertainties have negative effects for riding-sharing service,including reduction in total revenue and increase in emission.In addition,total revenue will decrease when carbon emission increases as the increasing of confidence interval level and uncertainty level with a given level of uncertainty.Hence,this study demonstrates that managers should balance system revenue,carbon emissions,and solution robustness in practice to obtain efficient decisions.