Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Towards Zero Bottlenecks for Scaling Autonomous Driving

Tonderski, Adam LU orcid (2025) In Doctoral Theses in Mathematical Sciences 2025(1).
Abstract
In this dissertation I examine the main scaling challenges in autonomous driving development, discussing recent advances in the field while contributing specific solutions to key bottlenecks. The first challenge is the reliance on human labor, particularly for annotations. Here we make two key contributions: new techniques to extract additional value from existing annotations through future prediction (I), and an adaptation of vision-language learning to 3D automotive sensors that reduces dependence on explicit labels while maintaining interpretability (II). The second challenge concerns access to training data covering the full spectrum of driving scenarios. We address this data bottleneck through complementary approaches: releasing a... (More)
In this dissertation I examine the main scaling challenges in autonomous driving development, discussing recent advances in the field while contributing specific solutions to key bottlenecks. The first challenge is the reliance on human labor, particularly for annotations. Here we make two key contributions: new techniques to extract additional value from existing annotations through future prediction (I), and an adaptation of vision-language learning to 3D automotive sensors that reduces dependence on explicit labels while maintaining interpretability (II). The second challenge concerns access to training data covering the full spectrum of driving scenarios. We address this data bottleneck through complementary approaches: releasing a diverse European driving dataset collected across multiple years and conditions (III), and developing a neural rendering method that enables scalable generation of realistic synthetic data (IV). Finally, to enable scalable safety testing, we introduce a closed-loop neural simulator that transforms ordinary driving scenarios into challenging near-collision cases (v). Together with broader advances in the field, our contributions suggest a promising path toward scaling autonomous vehicle development. (Less)
Please use this url to cite or link to this publication:
author
supervisor
opponent
  • Prof. Heide, felix, Princeton University, USA.
organization
publishing date
type
Thesis
publication status
published
subject
keywords
autonomous driving, computer vision, simulation, perception, vision-language, neural rendering
in
Doctoral Theses in Mathematical Sciences
volume
2025
issue
1
pages
82 pages
publisher
Lund University / Centre for Mathematical Sciences /LTH
defense location
Lecture Hall MH:G, Centre of Mathematical Sciences, Sölvegatan 18 A, Faculty of Engineering LTH, Lund University, Lund.
defense date
2025-01-31 13:00:00
ISSN
1404-0034
1404-0034
ISBN
978-91-8104-298-6
978-91-8104-299-3
language
English
LU publication?
yes
id
6c5272bc-a288-4d10-b885-956bb8ddc9f8
date added to LUP
2025-01-31 10:16:37
date last changed
2025-04-04 14:41:42
@phdthesis{6c5272bc-a288-4d10-b885-956bb8ddc9f8,
  abstract     = {{In this dissertation I examine the main scaling challenges in autonomous driving development, discussing recent advances in the field while contributing specific solutions to key bottlenecks. The first challenge is the reliance on human labor, particularly for annotations. Here we make two key contributions: new techniques to extract additional value from existing annotations through future prediction (I), and an adaptation of vision-language learning to 3D automotive sensors that reduces dependence on explicit labels while maintaining interpretability (II). The second challenge concerns access to training data covering the full spectrum of driving scenarios. We address this data bottleneck through complementary approaches: releasing a diverse European driving dataset collected across multiple years and conditions (III), and developing a neural rendering method that enables scalable generation of realistic synthetic data (IV). Finally, to enable scalable safety testing, we introduce a closed-loop neural simulator that transforms ordinary driving scenarios into challenging near-collision cases (v). Together with broader advances in the field, our contributions suggest a promising path toward scaling autonomous vehicle development.}},
  author       = {{Tonderski, Adam}},
  isbn         = {{978-91-8104-298-6}},
  issn         = {{1404-0034}},
  keywords     = {{autonomous driving; computer vision; simulation; perception; vision-language; neural rendering}},
  language     = {{eng}},
  month        = {{02}},
  number       = {{1}},
  publisher    = {{Lund University / Centre for Mathematical Sciences /LTH}},
  school       = {{Lund University}},
  series       = {{Doctoral Theses in Mathematical Sciences}},
  title        = {{Towards Zero Bottlenecks for Scaling Autonomous Driving}},
  url          = {{https://lup.lub.lu.se/search/files/207482229/Avhandling_Adam_Tonderski_LUCRIS.pdf}},
  volume       = {{2025}},
  year         = {{2025}},
}