基于逆向强化学习的铁路线路方案优选研究  

Research on Railway Route Scheme Optimization Selection Based on Inverse Reinforcement Learning

作　　者：马青松 朱颖 高天赐 罗圆[4] 何庆[1,2] 王平 MA Qingsong;ZHU Ying;GAO Tianci;LUO Yuan;HE Qing;WANG Ping(School of Civil Engineering,Southwest Jiaotong University,Chengdu 610031,China;MOE Key Laboratory of High-speed Railway Engineering,Southwest Jiaotong University,Chengdu 610031,China;China Railway Croup Limited,Beijing 100039,China;China Railway Eryuan Engineering Group Co.Ltd.,Chengdu 610031,China)

机构地区：[1]西南交通大学土木工程学院,成都610031 [2]西南交通大学高速铁路线路工程教育部重点实验室,成都610031 [3]中国中铁股份有限公司,北京100039 [4]中铁二院工程集团有限责任公司,成都610031

出　　处：《铁道建筑》2023年第7期1-7,共7页Railway Engineering

基　　金：国家自然科学基金(U1934214,51878576)。

摘　　要：铁路线路方案评价及比选多采用组合赋权法,其主观赋权过程计算冗杂。选取具备一定程度普适性的专家案例,采用最大熵逆向强化学习方法从专家案例中学习主观赋权“知识”,得到专家案例隐藏的“奖励”,从而获取可解释性的主观权重。将此主观权重与离差法所得客观权重组合并投入后续TOPSIS(Technique for Order Preference by Similarity to Ideal Solution)评价流程,对线路方案进行最终评价。结合具体实例,建立设计阶段绿色铁路的评价指标体系。结果表明:该方法可以有效计算铁路线路方案评价的量化指标,减小现有赋权方法的计算复杂度,取得较好的评价效果,与真实案例比选结果一致。通过讨论该方法的适用性、局限性及原因,确定该方法在初步评价和泛用性评价中的定位。The evaluation and comparison of railway line schemes often use the combination weighting method,and its subjective weighting process is computationally complex.This paper selected expert cases with a certain degree of universality,and used the maximum entropy reverse reinforcement learning method to learn their subjective weighting“knowledge”from expert cases,so as to obtain the hidden“reward”of expert cases,and thus obtain the interpretable subjective weight.Combining this subjective weight with the objective weight obtained by the dispersion method and inputting it into the subsequent TOPSIS(Technique for Order Preference by Similarity to Ideal Solution)evaluation process could conduct the final evaluation of the line scheme.Based on specific examples,an evaluation index system for green railways during the design phase was established.The results show that this method can effectively calculate the quantitative indicators of railway line scheme evaluation,reduce the computational complexity of existing weighting methods,and achieve good evaluation results,which are consistent with the comparison results of real cases.By discussing the applicability,limitations,and reasons of this method,the positioning of this method in preliminary and general evaluation is determined.

关 键 词：铁路选线 方案决策 评价模型 最大熵逆向强化学习 TOPSIS 绿色铁路 

分 类 号：U212.32[交通运输工程—道路与铁道工程]