From Virtual to Reality: Bridging the Simulation Gap for Autonomous Racing with Sim2Real (SA/MA/IDP)

Institute

Professur für autonome Fahrzeugsysteme

Type

Semester Thesis / Master's Thesis /

Content

experimental / theoretical / constructive /  

Description

F1TENTH Simulation

Are you ready for an exhilarating journey into the world of autonomous racing and cutting-edge research? Welcome to F1TENTH, where racing and academia merge seamlessly! Our 1/10th-scale self-driving cars are designed for thrilling autonomous racing while serving as a sophisticated research platform. These miniature vehicles have cutting-edge sensors, computing platforms, and control systems, allowing them to autonomously navigate dynamic racetracks without human intervention. Whether you're a racing enthusiast or a passionate researcher, F1TENTH offers an extraordinary opportunity to explore the frontiers of autonomous driving while embracing competition.

The current F1TENTH simulation for autonomous driving research has limitations in realism and may not accurately represent real-world driving conditions. This can lead to challenges in generalizing neural network models trained in the simulation to real on-road scenarios. End-to-end approaches, which integrate perception, decision-making, and control, may not be fully evaluated in the existing simulation environment, potentially yielding misleading results. Additionally, the simulation may lack diverse and extensive datasets for training and testing object detection algorithms, compromising their effectiveness in complex and dynamic environments. To advance research in autonomous driving, a more realistic simulation is therefore developed, offering a high-fidelity representation of real-world conditions and interactions with other vehicles. This improved simulation will empower us to develop and validate innovative algorithms more effectively, accelerating progress toward safer and more reliable autonomous driving systems.

Currently, we are looking for students to join us in the following areas:

• Development of a High Definition Model of our Racetrack
• Designing realistic sensor models for LiDAR- and camera sensors to provide accurate and diverse input data for other algorithms
• Evaluating the influence of different rendering resolutions on the performance of Deep Learning Algorithms
• Implementing an Interface with OpenAI

The functionality of the simulation and the respective work packages is verified via extensive tests with the real vehicle and the real track.

Requirements

Requirements:

• Coding experience (Python, C++)
• Framework experience (CARLA, Unreal Engine, ROS 1/2, Git)
• Knowledge in Automotive Engineering and Autonomous Driving
• Curious mindset
• Independent and organized work attitude
• Open-minded team player

Your Benefits:

• Fascinating, future-oriented field of research
• Young and dynamic team
• Monthly pitch of your thesis progress to Professor Betz
• Academic and professional support beyond the thesis
• Organized and structured thesis project
• If suitable, publication as a scientific paper
• Thesis project in German or English

The thesis can be started immediately. Please send your resume, recent grade report, and a brief description of your fields of interest and why you are the perfect fit for this thesis to felix.jahncketum.de.

I am always on the lookout for motivated and committed students and look forward to receiving your application so we can work together on the mobility of tomorrow.

Tags

AVS Jahncke

Possible start

sofort

Contact

Felix Jahncke