Aims: Develop software and algorithms for simulating an autonomous racing car.
Background: As autonomous vehicles (AV) technology is rapidly advancing, rigorous analysis and testing of such vehicles is particularly important to ensure their safety and performance. F1tenth is a hardware and software testbed for building an autonomous mini racing car based on a radio-controlled 1/10th-scale car and the Robot Operating System (ROS). More generally, the F1tenth platform enables AV research without the costs and risks of a real autonomous car. To facilitate development of the software/algorithmic side, F1tenth includes a fully simulated environment that replaces the physical car.
In this project, you will use the F1tenth ROS-based simulator to implement and evaluate the main algorithms underpinning an autonomous racing car. These include control, obstacle avoidance, mapping and localization, planning, and tracking.
- Familiarity with Robot Operating System (ROS) environment and the F1tenth simulator. Demonstrate manual control of the simulated car.
- Report on F1tenth car hardware, sensors and communication.
- Report on installation and configuration of ROS and the F1tenth simulator
- Implementation and evaluation of basic algorithms (PID control, wall-following, obstacle avoidance).
- Report(s) covering relevant theory for the implemented algorithms and experimental evaluation
- Implementation and evaluation of at least one localization algorithm of choice.
- Implementation and evaluation of at least one planning algorithm of choice.
- Final report, covering motivation and background around autonomous vehicles and F1tenth, description of implemented algorithms, and experimental results.
- Opponent pose estimation and prediction.
- Implement and compare additional planning lgorithms.
- Implement and evaluate raceline optimization on a seleced track.
- Visualize future trajectories with model predictive control.
- Learn end-to-end neural controller.
- F1TENTH – Course Documentation https://f1tenth. -rg/learn.html
- O’Kelly, Matthew, et al. – F1/10: An open-source autonomous cyber-physical platform. – arXiv- peprint arXiv:1901.08567 (2019)
- O’Kelly, Matthew, et al. – F1TENTH: An Open-source Evaluation Environment for Continuous Control and Reinforcement Learning. – NeurIPS 2019 Competition and emonstration Track. PMLR, 2020
Prerequisites: Good programming skills and a willingness to install and understand new programming environm