Predviđanje ponašanja pješaka u blizini pješačkih prijelaza namijenjena optimalnom upravljanju brzinom autonomnog vozila

master's thesis

• Title (english): Prediction of pedestrian crossing behaviour at vicinity of crosswalks aimed for optimal speed control of autonomous vehicle
• Author: Ana Bićanić
• Mentor: Branimir Škugor (mentor)
• Granter: University of Zagreb; Faculty of Mechanical Engineering and Naval Architecture; Zagreb
• Defense date and country: 2023-12-07, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Mechanical Engineering

Abstract

Brojni zadaci niže razine autonomne vožnje mogu se smatrati u dobroj mjeri riješenima. Međutim, kada se radi o autonomnoj vožnji koja iziskuje neki oblik interakcije s drugim sudionicima u prometu, kao što su biciklisti i pješaci, još uvijek postoje značajni izazovi koje treba riješiti. S obzirom na kompleksnost zadatka, predviđanje ponašanja ostalih sudionika u prometu u sklopu upravljačke strategije autonomnog vozila od neophodne je važnosti za postizanje sigurne i udobne vožnje. Na tom tragu u diplomskom radu se razmatra računalno učinkovit predikcijski model ponašanja pješaka u blizini pješačkih prijelaza bez semafora, koji je prikladan za adaptaciju u stvarnom vremenu. U radu je predložena metoda stohastičkog filtriranja podataka snimljenih tijekom svake interakcije vozila i pješaka, čija je glavna svrha propuštanje samo najinformativnijih podataka za potrebe adaptacije modela, te se na taj način ostvaruje minimiziranje prometa podataka ako bi se adaptacija modela provodila u oblaku. Demonstracija funkcionalnosti predloženog modela te njegova adaptacija provedena je u postojećem simulacijskom okruženju za slučaj jednog vozila i jednog pješaka, pri čemu je autonomno vozilo vođeno u otvorenoj petlji.

Abstract (english)

A number of tasks of lower level of autonomous driving can be considered to a good extent solved. However, when it comes to autonomous driving that requires some form of interaction with other road users, such as cyclists and pedestrians, there are still significant challenges to be addressed. Given the complexity of the task, predicting the behavior of other road users as a part of autonomous vehicle's management strategy is essential for achieving safe and comfortable driving. On this track, the thesis considers a computationally efficient predictive model of pedestrian behavior near pedestrian crossings without traffic lights, which is suitable for real-time adaptation. The paper proposes a method of stochastic filtering of data recorded during each vehicle-pedestrian interaction, the main purpose of which is to pass only the most informative data for the purpose of model adaptation, thus minimizing data traffic if the model adaptation is carried out in the cloud. The demonstration of the functionality of the proposed model and its adaptation was carried out in the existing simulation environment for the case of one vehicle and one pedestrian, where an autonomous vehicle was driven in an open loop.

• Keywords: autonomna vožnja; pješak; pješački prijelaz; predikcijski model; logistička regresija; simulacija
• Keywords (english): autonomous driving; pedestrian; crosswalk; prediction model; logistic regression; simulation
• Language: croatian
• URN:NBN: urn:nbn:hr:235:267578
• Study programme: Title: Mechanical Engineering; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka strojarstva (magistar/magistra inženjer/inženjerka strojarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2023-12-06 18:56:34