Časovno načrtovanje železniškega prometa z uporabo metode spodbujevanega učenja : doktorska disertacija

Šemrov, Darja (author), Žura, Marijan (mentor), Todorovski, Ljupčo (comentor)

Zanesljivost železniškega prometa najpogosteje povezujemo s točnostjo vlakov, torej primerjamo odstopanje dejanskih prihodov/odhodov vlakov s prihodi/odhodi, objavljenimi v voznem redu. Manjšo zamudo vlaka omilimo ali celo izničimo s časovnimi dodatki v voznem redu, večja zamuda pa povzroči tako imenovane sekundarne zamude ostalih vlakov na omrežju. Odseki prog, na katerih je visoka izkoriščenost kapacitete, so še posebej podvrženi nastanku zamud, saj večje število vlakov pomeni večje število možnih konfliktov in višjo stopnjo interakcije med vlaki, posledično pa je težje omejiti sekundarne zamude. Osebji upravljavca in prevoznika sta zadolženi, da železniški promet poteka varno, nemoteno in v skladu z voznim redom. Pa vendar lahko zaradi nepredvidenih dogodkov nastanejo zamude; v tem primeru je treba vlakom določiti nove čase prihodov in odhodov. Časovno načrtovanje voženj vlakov je kompleksen optimizacijski problem, ki ga dispečerji trenutno rešujejo na osnovi izkušenj, vendar z večanjem števila vlakov kompleksnost problema narašča, zato dispečerji vedno bolj potrebujejo sistem za pomoč pri odločanju, ki bi predlagal optimalno vodenje vlakov glede na zadani cilj, npr. minimalne zamude vseh vlakov. Časovno načrtovanje voženj vlakov sodi v skupino NP-polnih problemov, kjer odpovedo klasične matematično-računalniške metode optimiranja, nakazuje pa se uporabnost pristopov umetne inteligence. V okviru doktorske disertacije smo razvili algoritem časovnega načrtovanja voženj vlakov, ki temelji na metodi spodbujevanega učenja, natančneje učenja Q. Agent, ki se uči iz nagrad in kazni, ki jih pridobi iz okolja, išče optimalno strategijo vodenja vlakov glede na izbrano kriterijsko funkcijo.

• Language: Slovenian
• Keywords: grajeno okolje, gradbeništvo, disertacije, vozni red, časovno replaniranje vlakov, učenje Q
• Work type: Doctoral dissertation
• Typology: 2.08 - Doctoral Dissertation
• Organization: FGG - Faculty of Civil and Geodetic Engineering
• Place of publishing: Ljubljana
• Publisher: [D. Šemrov]
• Year: 2016
• Number of pages: 1 optični disk (CD-ROM)
• PID: 20.500.12556/RUL-80647
• COBISS.SI-ID: 7395425
• Publication date in RUL: 01.03.2016

ŠEMROV, Darja, 2016, Časovno načrtovanje železniškega prometa z uporabo metode spodbujevanega učenja : doktorska disertacija [online]. Doctoral dissertation. Ljubljana : D. Šemrov. [Accessed 24 April 2025]. Retrieved from: https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=eng&id=80647

Secondary language

• Language: English
• Title: Railway traffic scheduling with use of reinforcement learning : doctoral thesis

Abstract

The reliability of railway traffic is commonly evaluated with train punctuality, where the deviations of actual train arrivals/departures and train arrivals/departures published in the timetable are compared. Minor train delays can be mitigated or even eliminated with running time supplements, while major delays can lead to so-called secondary delays of other trains on the network. Railway lines with high capacity utilization are more likely subject to delays, since a greater number of trains means a larger number of potential conflicts and more interactions between trains. Consequently, the secondary delays are harder to limit. Railway manager and carrier personnel are responsible for safe, undisturbed and punctual railway traffic. But unforeseen events can lead to delays, which calls for train rescheduling, where new train arrivals and departures are calculated. Train rescheduling is a complex optimization problem, currently solved based on dispatcher’s expert knowledge. With the increasing number of trains the complexity of the problem grows, the need for a decision support system increases. Train rescheduling is considered an NP-complete problem, where conventional mathematical and computer optimization methods fail to find the optimal solution, but artificial intelligence approaches have some measure of success. In this dissertation an algorithm for train rescheduling based on reinforcement learning, more precisely Q-learning, was developed. The Q-learning agent learns from rewards and punishments received from the environment, and looks for the optimal train dispatching strategy depending on the objective function.

• Keywords: building environment, civil engineering, thesis, timetable, train rescheduling, Q learning