## Abstrakti

In spite of several envisioned advantages of Connected Vehicles (CV) for urban traﬃccontrol, considerable investment and time are needed before all vehicles are equipped with connection technology. Recently, some methods have been proposed to compensate for thelack of data in the case of low penetration rate of CVs, aiming, for example, at estimating thequeue length at a signalized intersection by combination of available data from CVs and ﬁxedsensors. However, queue length estimation methods might be useful in intersection controlbut they are not practical for under-saturated traﬃc condition. Moreover, the queue length innext signal cycles cannot be predicted using those methods. In this work, we propose machine-

learning-based models for estimating the number and characteristic of vehicles in the vicinityof intersections by utilising only information from CVs in a partially connected environment.

In contrast to previously proposed methods, our method is based on the idea of estimating thenumber of non-connected vehicles between each pair of CVs, for the cases of moving vehicles,stopping vehicles, and vehicles queuing. The models are built employing high-resolution dataproduced by a traﬃc simulation software. For evaluation, we test the models for diﬀerent CV penetration rates by implementing statistical tests on prediction results. The early ﬁndings show the developed models can estimate the number of unequipped vehicles with acceptable

accuracy.

learning-based models for estimating the number and characteristic of vehicles in the vicinityof intersections by utilising only information from CVs in a partially connected environment.

In contrast to previously proposed methods, our method is based on the idea of estimating thenumber of non-connected vehicles between each pair of CVs, for the cases of moving vehicles,stopping vehicles, and vehicles queuing. The models are built employing high-resolution dataproduced by a traﬃc simulation software. For evaluation, we test the models for diﬀerent CV penetration rates by implementing statistical tests on prediction results. The early ﬁndings show the developed models can estimate the number of unequipped vehicles with acceptable

accuracy.

Alkuperäiskieli | Englanti |
---|---|

Otsikko | 3rd Symposium on Management of Future motorway and urban Traffic Systems (MFTS 2020): Booklet of abstracts |

Sivumäärä | 4 |

Tila | Hyväksytty/In press - 2020 |

OKM-julkaisutyyppi | A4 Artikkeli konferenssijulkaisuussa |

Tapahtuma | Symposium on Management of Future Motorway and Urban Traffic Systems - Luxembourg, Luxemburg Kesto: 6 heinäk. 2020 → 8 heinäk. 2020 Konferenssinumero: 3 https://mfts20.gforge.uni.lu/#sub |

### Conference

Conference | Symposium on Management of Future Motorway and Urban Traffic Systems |
---|---|

Lyhennettä | MFTS |

Maa/Alue | Luxemburg |

Kaupunki | Luxembourg |

Ajanjakso | 06/07/2020 → 08/07/2020 |

www-osoite |