Incorporating Driving Suitability When Generating Routes for Driving Navigation
This paper investigates the effects of incorporating driving suitability as an added factor when generating routes for driving navigation on a person's route choice behavior. Mobile navigation apps being used by millions of people around the world focus on optimizing routes in terms of travel t...
Saved in:
Main Author: | |
---|---|
Format: | text |
Language: | English |
Published: |
Animo Repository
2022
|
Online Access: | https://animorepository.dlsu.edu.ph/etdm_softtech/1 https://animorepository.dlsu.edu.ph/cgi/viewcontent.cgi?article=1001&context=etdm_softtech |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | De La Salle University |
Language: | English |
id |
oai:animorepository.dlsu.edu.ph:etdm_softtech-1001 |
---|---|
record_format |
eprints |
spelling |
oai:animorepository.dlsu.edu.ph:etdm_softtech-10012022-07-02T17:59:03Z Incorporating Driving Suitability When Generating Routes for Driving Navigation Obaldo, Marc Anthony This paper investigates the effects of incorporating driving suitability as an added factor when generating routes for driving navigation on a person's route choice behavior. Mobile navigation apps being used by millions of people around the world focus on optimizing routes in terms of travel time and cost. Past research have shown that there are other factors that contribute to a driver's route choice, such as comfort and safety, among others. Using remote sensing, we determined the driving suitability of roads by combining safety and comfort scores for road images in various areas of Metro Manila on a scale of 0-10. To do this, we first crowdsourced safety and comfort ratings for road images to be used as labels. Then, for comfort, we trained a CNN based on U-Net to detect various road surface irregularities. For safety, we calculated the perceived lighting of various regions of road images. We then trained regression models to determine the relationship of these detected features with the crowdsourced labels and predict scores for our road network. We were able to produce models with an RMSE of 1.16 and 1.27 for comfort and safety scores respectively which were within acceptable levels. We incorporated the predicted scores into as factors when generating routes for a proof-of-concept mobile app. Next, we conducted a field study where 17 drivers used the mobile app in their trips which presented both our route recommendations and the top results from Google Maps. Regardless of their selected route, we found that between 40\% to 58\% of drivers deviated from recommended routes, citing reasons such as real-time visual information as well as various comfort, safety, and road reliability factors. We also recruited drivers to drive through our entire route recommendation end-to-end to assess the driving suitability compared to their typical route. We found that our route recommendations and their driving suitability closely resemble what drivers experienced in their trips. However, they noted sections of their trips with both static and temporal factors that contribute negatively to comfort and safety, such as time-based presence of pedestrians or vehicles, or undetected surface irregularities. In the future, we recommend a more refined comfort and safety scoring mechanism: exploring state-of-the-art segmentation techniques, adding more class labels, possibly using a multi-modal approach with IMU data, as well as taking into account more factors such as differing vehicle types, real estate pricing, or crime index for route areas. 2022-01-01T08:00:00Z text application/pdf https://animorepository.dlsu.edu.ph/etdm_softtech/1 https://animorepository.dlsu.edu.ph/cgi/viewcontent.cgi?article=1001&context=etdm_softtech Software Technology Master's Theses English Animo Repository |
institution |
De La Salle University |
building |
De La Salle University Library |
continent |
Asia |
country |
Philippines Philippines |
content_provider |
De La Salle University Library |
collection |
DLSU Institutional Repository |
language |
English |
description |
This paper investigates the effects of incorporating driving suitability as an added factor when generating routes for driving navigation on a person's route choice behavior. Mobile navigation apps being used by millions of people around the world focus on optimizing routes in terms of travel time and cost. Past research have shown that there are other factors that contribute to a driver's route choice, such as comfort and safety, among others. Using remote sensing, we determined the driving suitability of roads by combining safety and comfort scores for road images in various areas of Metro Manila on a scale of 0-10. To do this, we first crowdsourced safety and comfort ratings for road images to be used as labels. Then, for comfort, we trained a CNN based on U-Net to detect various road surface irregularities. For safety, we calculated the perceived lighting of various regions of road images. We then trained regression models to determine the relationship of these detected features with the crowdsourced labels and predict scores for our road network. We were able to produce models with an RMSE of 1.16 and 1.27 for comfort and safety scores respectively which were within acceptable levels. We incorporated the predicted scores into as factors when generating routes for a proof-of-concept mobile app. Next, we conducted a field study where 17 drivers used the mobile app in their trips which presented both our route recommendations and the top results from Google Maps. Regardless of their selected route, we found that between 40\% to 58\% of drivers deviated from recommended routes, citing reasons such as real-time visual information as well as various comfort, safety, and road reliability factors. We also recruited drivers to drive through our entire route recommendation end-to-end to assess the driving suitability compared to their typical route. We found that our route recommendations and their driving suitability closely resemble what drivers experienced in their trips. However, they noted sections of their trips with both static and temporal factors that contribute negatively to comfort and safety, such as time-based presence of pedestrians or vehicles, or undetected surface irregularities. In the future, we recommend a more refined comfort and safety scoring mechanism: exploring state-of-the-art segmentation techniques, adding more class labels, possibly using a multi-modal approach with IMU data, as well as taking into account more factors such as differing vehicle types, real estate pricing, or crime index for route areas. |
format |
text |
author |
Obaldo, Marc Anthony |
spellingShingle |
Obaldo, Marc Anthony Incorporating Driving Suitability When Generating Routes for Driving Navigation |
author_facet |
Obaldo, Marc Anthony |
author_sort |
Obaldo, Marc Anthony |
title |
Incorporating Driving Suitability When Generating Routes for Driving Navigation |
title_short |
Incorporating Driving Suitability When Generating Routes for Driving Navigation |
title_full |
Incorporating Driving Suitability When Generating Routes for Driving Navigation |
title_fullStr |
Incorporating Driving Suitability When Generating Routes for Driving Navigation |
title_full_unstemmed |
Incorporating Driving Suitability When Generating Routes for Driving Navigation |
title_sort |
incorporating driving suitability when generating routes for driving navigation |
publisher |
Animo Repository |
publishDate |
2022 |
url |
https://animorepository.dlsu.edu.ph/etdm_softtech/1 https://animorepository.dlsu.edu.ph/cgi/viewcontent.cgi?article=1001&context=etdm_softtech |
_version_ |
1740844658714476544 |