@inproceedings{Yun2025,

title = {PrefDrive: Enhancing Autonomous Driving through Preference-Guided Large Language Models},

author = {Yun Li and Ehsan Javanmardi and Simon Thompson and Kai Katsumata and Alex Orsholits and Manabu Tsukada},

url = {https://github.com/LiYun0607/PrefDrive/},

year  = {2025},

date = {2025-06-22},

urldate = {2025-06-22},

booktitle = {36th IEEE Intelligent Vehicles Symposium (IV2025)},

address = {Cluj-Napoca, Romania},

abstract = {This paper presents PrefDrive, a novel framework that

integrates driving preferences into autonomous driving

models through large language models (LLMs). While recent

advances in LLMs have shown promise in autonomous driving,

existing approaches often struggle to align with specific

driving behaviors (e.g., maintaining safe distances, smooth

acceleration patterns) and operational requirements (e.g.,

traffic rule compliance, route adherence). We address this

challenge by developing a preference learning framework

that combines multimodal perception with natural language

understanding. Our approach leverages Direct Preference

Optimization (DPO) to fine-tune LLMs efficiently on

consumer-grade hardware, making advanced autonomous driving

research more accessible to the broader research community.

We introduce a comprehensive dataset of 74,040 sequences,

carefully annotated with driving preferences and driving

decisions, which, along with our trained model checkpoints,

will be made publicly available to facilitate future

research. Through extensive experiments in the CARLA

simulator, we demonstrate that our preference-guided

approach significantly improves driving performance across

multiple metrics, including distance maintenance and

trajectory smoothness. Results show up to 28.1% reduction

in traffic rule violations and 8.5% improvement in

navigation task completion while maintaining appropriate

distances from obstacles. The framework demonstrates robust

performance across different urban environments, showcasing

the effectiveness of preference learning in autonomous

driving applications. },

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{Trumpp2024,

title = {RaceMOP: Mapless Online Path Planning for Multi-Agent Autonomous Racing using Residual Policy Learning},

author = {Raphael Trumpp and Ehsan Javanmardi and Jin Nakazato and Manabu Tsukada and Marco Caccamo},

url = {http://github.com/raphajaner/racemop},

doi = {10.1109/IROS58592.2024.10801657},

year  = {2024},

date = {2024-09-14},

urldate = {2024-09-14},

booktitle = {The 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2024)},

address = {Abu Dhabi ,UAE},

abstract = {The interactive decision-making in multi-agent autonomous racing offers insights valuable beyond the domain of self-driving cars. Mapless online path planning is particularly of practical appeal but poses a challenge for safely overtaking opponents due to the limited planning horizon. Accordingly, this paper introduces RaceMOP, a novel method for mapless online path planning designed for multi-agent racing of F1TENTH cars. Unlike classical planners that depend on predefined racing lines, RaceMOP operates without a map, relying solely on local observations to overtake other race cars at high speed. Our approach combines an artificial potential field method as a base policy with residual policy learning to introduce long-horizon planning capabilities. We advance the field by introducing a novel approach for policy fusion with the residual policy directly in probability space. Our experiments for twelve simulated racetracks validate that RaceMOP is capable of long-horizon decision-making with robust collision avoidance during over- taking maneuvers. RaceMOP demonstrates superior handling over existing mapless planners while generalizing to unknown racetracks, paving the way for further use of our method in robotics. We make the open-source code for RaceMOP available at http://github.com/raphajaner/racemop.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{Tao2023c,

title = {Flowsim: A Modular Simulation Platform for Microscopic Behavior Analysis of City-Scale Connected Autonomous Vehicles},

author = {Ye Tao and Ehsan Javanmardi and Jin Nakazato and Manabu Tsukada and Hiroshi Esaki},

url = {https://github.com/tlab-wide/flowsim

https://arxiv.org/abs/2306.05738},

doi = {10.1109/ITSC57777.2023.10421900},

year  = {2023},

date = {2023-09-24},

urldate = {2023-09-24},

booktitle = {The 26th edition of the IEEE International Conference on Intelligent Transportation Systems (ITSC 2023)},

address = {Bilbao, Bizkaia, Spain},

abstract = {As connected autonomous vehicles (CAVs) become increasingly prevalent, there is a growing need for simulation platforms that can accurately evaluate CAV behavior in large-scale environments. In this paper, we propose Flowsim, a novel simulator specifically designed to meet these requirements. Flowsim offers a modular and extensible architecture that enables the analysis of CAV behaviors in large-scale scenarios. It provides researchers with a customizable platform for studying CAV interactions, evaluating communication and networking protocols, assessing cybersecurity vulnerabilities, optimizing traffic management strategies, and developing and evaluating policies for CAV deployment. Flowsim is implemented in pure Python in approximately 1,500 lines of code, making it highly readable, understandable, and easily modifiable. We verified the functionality and performance of Flowsim via a series of experiments based on realistic traffic scenarios. The results show the effectiveness of Flowsim in providing a flexible and powerful simulation environment for evaluating CAV behavior and data flow. Flowsim is a valuable tool for researchers, policymakers, and industry professionals who are involved in the development, evaluation, and deployment of CAVs. The code of Flowsim is publicly available on GitHub under the MIT license. },

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{Yamazaki2023,

title = {ToST: Tokyo SUMO traffic scenario },

author = {Yuji Yamazaki and Yasumasa Tamura and Xavier Defago and Ehsan Javanmardi and Manabu Tsukada},

url = {https://github.com/dfg-lab/ToSTScenario},

doi = {10.1109/ITSC57777.2023.10422517},

year  = {2023},

date = {2023-09-24},

urldate = {2023-09-24},

booktitle = {The 26th edition of the IEEE International Conference on Intelligent Transportation Systems (ITSC 2023)},

address = {Bilbao, Bizkaia, Spain},

abstract = {In recent years, research, development, and demonstrations aimed at the societal implementation of autonomous driving have attracted increasing attention. Localization, which involves obtaining information of the surrounding environment from sensor data and estimating the vehicle's position, is necessary for realizing autonomous driving. Localization is commonly performed with 3D LiDAR as a sensor owing to its high measurement accuracy and immunity to ambient light conditions, which allow for precise localization. However, when the surrounding area has distinctive features, localization accuracy may decrease. In this study, we proposed a method based on deep learning to predict the localization accuracy for autonomous driving. The overall localization accuracy can be improved by predicting the accuracy of localization using other sensors, such as GNSS and IMU, or pavement markings in areas with poor accuracy. We created a dataset for predicting the localization accuracy using an open-source autonomous driving simulator. In an experiment, we applied the proposed method to the created dataset. Thresholds were set for errors in the x-direction, y-direction, and distance for localization. Predictions with high accuracy and F-values were obtained. The results indicate that the proposed method can accurately predict the localization accuracy.  },

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{Asabe2023b,

title = {AutowareV2X: Reliable V2X Communication and Collective Perception for Autonomous Driving},

author = {Yu Asabe and Ehsan Javanmardi and Jin Nakazato and Manabu Tsukada and Hiroshi Esaki},

url = {https://github.com/tlab-wide/AutowareV2X

tlab-wide.github.io/AutowareV2X/main/

https://www.youtube.com/watch?v=57fx3-gUNxU},

doi = {10.1109/VTC2023-Spring57618.2023.10199425},

year  = {2023},

date = {2023-06-20},

urldate = {2023-06-20},

booktitle = {The 2023 IEEE 97th Vehicular Technology Conference (VTC2023-Spring)},

address = {Florence, Italy},

abstract = {For cooperative intelligent transport systems (C-ITS), vehicle-to-everything (V2X) communication is utilized to allow autonomous vehicles to share critical information with each other. We propose AutowareV2X, an implementation of a V2X communication module that is integrated into the autonomous driving (AD) software, Autoware. AutowareV2X provides external connectivity to the entire AD stack, enabling the end-to-end (E2E) experimentation and evaluation of connected autonomous vehicles (CAV). The Collective Perception Service was also implemented, allowing the transmission of Collective Perception Messages (CPMs). A dual-channel mechanism that enables wireless link redundancy on the critical object information shared by CPMs is also proposed. Performance evaluation in field experiments has indicated that the E2E latency of perception information is around 30 ms, and shared object data can be used by the AD software to conduct collision avoidance maneuvers. Dual-channel delivery of CPMs enabled the CAV to dynamically select the best CPM from CPMs received from different links, depending on the freshness of their information.},

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}

@inproceedings{Tsukada2020,

title = {AutoC2X: Open-source software to realize V2X cooperative perception among autonomous vehicles},

author = {Manabu Tsukada and Takaharu Oi and Akihide Ito and Mai Hirata and Hiroshi Esaki},

url = {https://github.com/esakilab/AutoC2X-AW

https://hal.archives-ouvertes.fr/hal-02942051/document?.pdf

https://youtu.be/kyv0sTyCIgU},

doi = {10.1109/VTC2020-Fall49728.2020.9348525},

year  = {2020},

date = {2020-11-18},

urldate = {2020-11-18},

booktitle = {The 2020 IEEE 92nd Vehicular Technology Conference (VTC2020-Fall)},

address = {Victoria, B.C., Canada},

abstract = {The realization of vehicle-to-everything (V2X) communication enhances the capabilities of autonomous vehicles in terms of safety efficiency and comfort. In particular, sensor data sharing, known as cooperative perception, is a crucial technique to accommodate vulnerable road users in a cooperative intelligent transport system (ITS). In this regard, open-source software plays a significant role in prototyping, validation, and deployment. Specifically, in the developer community, Autoware is a popular open-source software for self-driving vehicles, and OpenC2X is an open-source experimental and prototyping platform for cooperative ITS. This paper reports on a system named AutoC2X to enable cooperative perception by using OpenC2X for Autoware-based autonomous vehicles. The developed system is evaluated by conducting field experiments involving real hardware. The results demonstrate that AutoC2X can deliver the cooperative perception message within 100 ms in the worst case. },

keywords = {},

pubstate = {published},

tppubtype = {inproceedings}

}