歐洲智慧列車自組織模式研究項(xiàng)目SORTED MOBILITY介紹

徐紀(jì)康 整理

SORTED MOBILITY是“面向分布式出行演進(jìn)的自組織軌道交通”（Self-Organized Rail Traffic for the Evolution of Decentralized MOBILITY）的縮寫。該項(xiàng)目提出了一種整體性方法，用于城市及城際區(qū)域公共交通運(yùn)營(yíng)的自組織管理，尤其將軌道交通作為出行網(wǎng)絡(luò)核心進(jìn)行重點(diǎn)關(guān)注。在這一方法框架下，智能列車將以自組織模式運(yùn)行。軌道交通系統(tǒng)將具備更強(qiáng)的韌性，能夠根據(jù)出行需求變化及突發(fā)干擾情況，實(shí)現(xiàn)對(duì)動(dòng)態(tài)環(huán)境的自適應(yīng)調(diào)整。

1 鐵路系統(tǒng)運(yùn)營(yíng)痛點(diǎn)

突發(fā)事件影響：鐵路系統(tǒng)在運(yùn)營(yíng)中易因突發(fā)事件導(dǎo)致性能下降，進(jìn)而引發(fā)沖突—即兩列及以上列車若按計(jì)劃速度運(yùn)行，會(huì)同時(shí)占用同一軌道段。

沖突連鎖反應(yīng)：沖突發(fā)生時(shí)，列車需減速或停車以保證安全間距，最終導(dǎo)致延誤傳播。

2 傳統(tǒng)管理模式局限

采用集中式，主要依賴調(diào)度員的個(gè)人經(jīng)驗(yàn)進(jìn)行重調(diào)度決策。

3 SORTED MOBILITY項(xiàng)目的研究目標(biāo)

為解決鐵路系統(tǒng)中突發(fā)事件引發(fā)的沖突及延誤傳播問(wèn)題，歐洲SORTED MOBILITY項(xiàng)目提出了一種基于自組織的列車重調(diào)度與重路由范式，替代傳統(tǒng)集中式鐵路管理模式。

該模式以列車作為智能體，通過(guò) 鄰域選擇、假設(shè)生成、共識(shí)達(dá)成、合并 四個(gè)模塊的生成實(shí)時(shí)交通計(jì)劃，并以法國(guó)巴黎-勒阿弗爾線一段80公里線路為案例進(jìn)行概念驗(yàn)證，結(jié)果顯示共識(shí)解決方案能較好地逼近全局最優(yōu)。

該項(xiàng)目提出“范式轉(zhuǎn)移”—用多智能體（智慧列車）的自組織決策替代傳統(tǒng)集中式管理。

自組織決策通過(guò)列車（智能體）序貫執(zhí)行以下四個(gè)模塊實(shí)現(xiàn)，最終生成實(shí)時(shí)交通計(jì)劃（RTTP）（包含重路徑與重調(diào)度決策），具體流程如下所示：

4 優(yōu)缺點(diǎn)比較分析

SORTEDMOBILITY項(xiàng)目提出的自組織列車管理模式，相比傳統(tǒng)集中式模式，在決策流程上的核心差異體現(xiàn)在決策主體、靈活性與響應(yīng)性三個(gè)維度：

1.決策主體：傳統(tǒng)集中式模式以“調(diào)度員”為核心，依賴個(gè)人經(jīng)驗(yàn)及集中式算法制定決策；自組織模式以“列車”為智能體，列車自主參與鄰域識(shí)別、決策假設(shè)生成、共識(shí)協(xié)商，具備獨(dú)立決策能力；

2.靈活性：集中式模式受限于統(tǒng)一的優(yōu)化目標(biāo)（如最小化總延誤），難以兼顧個(gè)體列車需求；自組織模式中列車可優(yōu)先評(píng)估自身延誤，鄰域可動(dòng)態(tài)重疊，決策更貼合局部交通實(shí)際；

3.響應(yīng)性：集中式模式需匯總?cè)到y(tǒng)信息后計(jì)算決策，可能存在延遲；自組織模式通過(guò)“每30分鐘更新交通狀態(tài)+分布式共識(shí)”實(shí)現(xiàn)近實(shí)時(shí)調(diào)整，對(duì)擾動(dòng)的響應(yīng)更迅速。

部分英文原文內(nèi)容

Trains of the Future: How Self-Organising Railways Could Revolutionise Transportation

Railway traffic management is on the cusp of transformation. While self-driving cars and autonomous drones capture headlines, an international research project has been exploring whether trains could manage themselves. The Sorted Mobility project, funded by JPI Urban Europe, brings together rail operators and researchers from France, Italy, Denmark, and the Netherlands to investigate whether self-organised rail traffic could match or even outperform traditional centralised management.

By Sahfayet Choudhury

Published: 5 December 2024

“Today, trains work according to a timetable, and when something happens, an operator decides how to deal with traffic, so delays don’t propagate through the network,” explains Paola Pellegrini, the project’s coordinator. She adds, “but with networks becoming increasingly complex, we wanted to explore if trains could have their own ‘brain’ and negotiate with each other.”

Shifting Away from Central Control

In this self-organised system, trains would communicate directly, making local decisions about priorities and platforms without central control. The project developed a sophisticated testing environment combining mathematical optimisation with simulations. “Every detail of railway infrastructure and train movements is represented,” Paola says, including train weight, length, and acceleration curves.

Within Sorted Mobility’s autonomous system, trains use a consensus system inspired by artificial intelligence models when delays occur. Each train maintains a list of preferred options but doesn’t need to reveal why it prefers certain choices—crucial for maintaining commercial privacy. If trains can’t reach an agreement, they default to their original schedule, ensuring system stability.

Testing Across Diverse Railways

The project tested this concept in distinctly different scenarios. One was a busy Italian line between Pioltello and Rovato, which is 54km long and carries both freight and passenger traffic. Here, researchers examined how self-organisation could work in a competitive environment where multiple operators share a track. Running simulations with over 75 trains in a five-hour period, they found the system could maintain overall performance while protecting operators’ commercial interests.

“Most trains could be a bit happier without worsening system performance and keeping their private information,” Paola notes. This proved particularly valuable for freight operators, who might not want to reveal the importance of specific shipments but still need to negotiate priority.

In France, the team studied a rural “capillary” line between Guingamp and Paimpol, which is 37km long with a movement of 150 passengers and sees 21 trains every five and a half hours. Here, they found that self-organisation could maintain service quality while potentially reducing operational overhead. “These types of lines are very critical in France and other similar countries,” Paola explains. “There must be a way to operate them with minimum costs while maintaining service.”

The single-operator environment allowed researchers to focus on passenger benefits. They discovered that trains that could use their own passenger counting data to make decisions achieved better outcomes than centralised systems that lack access to this information due to privacy barriers.

Tackling Urban Complexity

The most ambitious test came in Copenhagen’s suburban rail network, comprised of 170km of a star-shaped urban network with multiple travel operators and over 16000 passengers using the system in the three hours of the morning peak. Here, researchers integrated passenger prediction into the decision-making process. Using a sophisticated population model incorporating socioeconomic data, work patterns, and travel habits, they created the first-ever system of this kind that considers how railway performance affects passenger route choices.

“We predict demand flows considering the impact of railway traffic on passenger choices,” Paola explains. “If I’m a passenger, I’m going to check my travel app and decide based on what lines are working. This had never been done in the literature before.”

While time constraints limited the full exploration of this approach, initial results suggested significant potential. The team observed that looking 20 minutes ahead in passenger predictions could inform better decisions, though they also identified challenges with “border effects” when significant events occurred just outside their prediction window.

Across all three contexts, self-organised systems performed comparably to traditional centralised management while offering distinct advantages. Beyond maintaining privacy and improving passenger service, self-organisation could help solve the scaling challenges that plague current systems.

“The problem of scaling up decisions is very real today,” Paola notes. While centralised systems struggle with more extensive networks, self-organisation’s local decision-making approach could naturally handle network expansion.

However, implementation faces significant institutional and legal hurdles. As Paola explains, “We probably won’t have a fully self-organised system soon, but we could start thinking of hybrid systems where trains make some decisions.” The project’s final round table, which included representatives from Deutsche Bahn, Network Rail, and other European railway infrastructure providers, endorsed this gradual approach.

Rural lines might see the first implementations. “Changing the system in these isolated lines may be easier, operated by a single operator where problems are less critical – we might see change there within five to ten years,” Paola suggests. Success in these environments could pave the way for broader adoption.

Building Better Assessment Tools

Perhaps equally valuable to the self-organisation concept itself, the project developed sophisticated assessment tools combining optimisation, simulation, and passenger modelling. This framework allows for a realistic evaluation of any traffic management approach, whether centralised or self-organised.

“What we started doing was a very original contribution,” Paola notes. “We predicted demand flows considering the impact of railway traffic on passenger choices. This opens up crucial questions about how people would react to smarter systems. Would they use them more?”

A Model for Future Research

Beyond its technical findings, the project demonstrated the value of allowing truly exploratory research in railway systems. “This was amazing,” Paola reflects. “We had actual collaboration between academia and industry without needing to end up with a commercialised product. We were really doing pure research.”

This freedom allowed the team to investigate a concept that could reshape railway management thoroughly. While full implementation remains distant, the project proved self-organisation’s viability and identified promising paths forward through hybrid systems and targeted deployment.

素材來(lái)源：讀者投稿

·軌道交通，通向美好生活·