Deep Reinforcement Learning of Autonomous Control Actions to Improve Bus-Service Regularity. Anonymous 2023. abstract bibtex Bus Bunching is caused by irregularities in demand across the bus route, together with other factors such as traffic. The effect of this problem is that buses operating on the same route start to catch up with each other, severely impacting the regularity and the quality of the service. Control actions such as Bus Holding and Stop Skipping can be introduced to regulate the service and adjust the headway, the temporal gap between the arrival times of two buses on the same bus stop. Traditionally, this phenomenon is mitigated either reactively online through simple rule-based control, or preemptively through analytical scheduling solutions such as mathematical opti- mization. Over time, both approaches degrade to an irregular service. In this work, we investigate the use of Deep Reinforcement Learn- ing algorithms to train a policy that determines which actions should take place at specific control points, in order to regularise the bus service. While prior studies are typically restricted to one control ac- tion, we consider both Bus Holding and Stop Skipping. We replicate benchmarks found in the latest literature, and also introduce traffic to increase the realism of the simulation. Furthermore we also con- sider scenarios where the service is already unstable and buses are already bunched together, a first of this kind of study. We compare the performance of the RL-based policies with a no-control policy and a rule-based policy, and the learnt policies not only keep a sig- nificantly lower headway variance and mean waiting time, but also recover from unstable scenarios and restore service regularity.
@misc{anonymous_deep_2023,
title = {Deep {Reinforcement} {Learning} of {Autonomous} {Control} {Actions} to {Improve} {Bus}-{Service} {Regularity}},
abstract = {Bus Bunching is caused by irregularities in demand
across the bus route, together with other factors such as traffic. The
effect of this problem is that buses operating on the same route start
to catch up with each other, severely impacting the regularity and
the quality of the service. Control actions such as Bus Holding and
Stop Skipping can be introduced to regulate the service and adjust the
headway, the temporal gap between the arrival times of two buses on
the same bus stop. Traditionally, this phenomenon is mitigated either
reactively online through simple rule-based control, or preemptively
through analytical scheduling solutions such as mathematical opti-
mization. Over time, both approaches degrade to an irregular service.
In this work, we investigate the use of Deep Reinforcement Learn-
ing algorithms to train a policy that determines which actions should
take place at specific control points, in order to regularise the bus
service. While prior studies are typically restricted to one control ac-
tion, we consider both Bus Holding and Stop Skipping. We replicate
benchmarks found in the latest literature, and also introduce traffic
to increase the realism of the simulation. Furthermore we also con-
sider scenarios where the service is already unstable and buses are
already bunched together, a first of this kind of study. We compare
the performance of the RL-based policies with a no-control policy
and a rule-based policy, and the learnt policies not only keep a sig-
nificantly lower headway variance and mean waiting time, but also
recover from unstable scenarios and restore service regularity.},
author = {Anonymous},
year = {2023},
}
Downloads: 0
{"_id":"MWqwLcdF3axetpc88","bibbaseid":"anonymous-deepreinforcementlearningofautonomouscontrolactionstoimprovebusserviceregularity-2023","author_short":["Anonymous"],"bibdata":{"bibtype":"misc","type":"misc","title":"Deep Reinforcement Learning of Autonomous Control Actions to Improve Bus-Service Regularity","abstract":"Bus Bunching is caused by irregularities in demand across the bus route, together with other factors such as traffic. The effect of this problem is that buses operating on the same route start to catch up with each other, severely impacting the regularity and the quality of the service. Control actions such as Bus Holding and Stop Skipping can be introduced to regulate the service and adjust the headway, the temporal gap between the arrival times of two buses on the same bus stop. Traditionally, this phenomenon is mitigated either reactively online through simple rule-based control, or preemptively through analytical scheduling solutions such as mathematical opti- mization. Over time, both approaches degrade to an irregular service. In this work, we investigate the use of Deep Reinforcement Learn- ing algorithms to train a policy that determines which actions should take place at specific control points, in order to regularise the bus service. While prior studies are typically restricted to one control ac- tion, we consider both Bus Holding and Stop Skipping. We replicate benchmarks found in the latest literature, and also introduce traffic to increase the realism of the simulation. Furthermore we also con- sider scenarios where the service is already unstable and buses are already bunched together, a first of this kind of study. We compare the performance of the RL-based policies with a no-control policy and a rule-based policy, and the learnt policies not only keep a sig- nificantly lower headway variance and mean waiting time, but also recover from unstable scenarios and restore service regularity.","author":[{"firstnames":[],"propositions":[],"lastnames":["Anonymous"],"suffixes":[]}],"year":"2023","bibtex":"@misc{anonymous_deep_2023,\n\ttitle = {Deep {Reinforcement} {Learning} of {Autonomous} {Control} {Actions} to {Improve} {Bus}-{Service} {Regularity}},\n\tabstract = {Bus Bunching is caused by irregularities in demand \nacross the bus route, together with other factors such as traffic. The \neffect of this problem is that buses operating on the same route start \nto catch up with each other, severely impacting the regularity and \nthe quality of the service. Control actions such as Bus Holding and \nStop Skipping can be introduced to regulate the service and adjust the \nheadway, the temporal gap between the arrival times of two buses on \nthe same bus stop. Traditionally, this phenomenon is mitigated either \nreactively online through simple rule-based control, or preemptively \nthrough analytical scheduling solutions such as mathematical opti- \nmization. Over time, both approaches degrade to an irregular service.\nIn this work, we investigate the use of Deep Reinforcement Learn- \ning algorithms to train a policy that determines which actions should \ntake place at specific control points, in order to regularise the bus \nservice. While prior studies are typically restricted to one control ac- \ntion, we consider both Bus Holding and Stop Skipping. We replicate \nbenchmarks found in the latest literature, and also introduce traffic \nto increase the realism of the simulation. Furthermore we also con- \nsider scenarios where the service is already unstable and buses are \nalready bunched together, a first of this kind of study. We compare \nthe performance of the RL-based policies with a no-control policy \nand a rule-based policy, and the learnt policies not only keep a sig- \nnificantly lower headway variance and mean waiting time, but also \nrecover from unstable scenarios and restore service regularity.},\n\tauthor = {Anonymous},\n\tyear = {2023},\n}\n\n\n\n\n\n\n\n","author_short":["Anonymous"],"key":"anonymous_deep_2023","id":"anonymous_deep_2023","bibbaseid":"anonymous-deepreinforcementlearningofautonomouscontrolactionstoimprovebusserviceregularity-2023","role":"author","urls":{},"metadata":{"authorlinks":{}},"downloads":0,"html":""},"bibtype":"misc","biburl":"https://bibbase.org/zotero/matteocarde","dataSources":["BLJPWpPi3hneTGaPF"],"keywords":[],"search_terms":["deep","reinforcement","learning","autonomous","control","actions","improve","bus","service","regularity","anonymous"],"title":"Deep Reinforcement Learning of Autonomous Control Actions to Improve Bus-Service Regularity","year":2023}