Accelerating Real-World Overtaking in F1TENTH Racing Employing Reinforcement Learning Methods

2025-10-29

AI Summary

Current autonomous racing systems exhibit insufficient reliability in wheel-to-wheel racing—particularly in dynamic overtaking—struggling to balance safety and success rate. This paper proposes a reinforcement learning framework tailored to realistic racing scenarios, featuring a novel race-car–overtaking cooperative agent trained jointly in simulation and on physical vehicles, augmented with adversarial training to enhance policy robustness. Crucially, it achieves active overtaking policy learning under Sim-to-Real transfer, departing from conventional track-following paradigms. Evaluated on the F1TENTH standardized platform, the method attains an 87% overtaking success rate against diverse competing algorithms—representing a 31-percentage-point improvement over baseline approaches—and marks the first demonstration of high-reliability, deployable dynamic overtaking capability on real-world miniature racing platforms.

Technology Category

• Multiagent Systems: Adversarial Agents
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Game Theory and Economic Paradigms: Adversarial Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

While autonomous racing performance in Time-Trial scenarios has seen significant progress and development, autonomous wheel-to-wheel racing and overtaking are still severely limited. These limitations are particularly apparent in real-life driving scenarios where state-of-the-art algorithms struggle to safely or reliably complete overtaking manoeuvres. This is important, as reliable navigation around other vehicles is vital for safe autonomous wheel-to-wheel racing. The F1Tenth Competition provides a useful opportunity for developing wheel-to-wheel racing algorithms on a standardised physical platform. The competition format makes it possible to evaluate overtaking and wheel-to-wheel racing algorithms against the state-of-the-art. This research presents a novel racing and overtaking agent capable of learning to reliably navigate a track and overtake opponents in both simulation and reality. The agent was deployed on an F1Tenth vehicle and competed against opponents running varying competitive algorithms in the real world. The results demonstrate that the agent's training against opponents enables deliberate overtaking behaviours with an overtaking rate of 87% compared 56% for an agent trained just to race.

Problem

• Developing reliable overtaking algorithms for autonomous racing
• Improving wheel-to-wheel racing performance in real-world scenarios
• Enhancing overtaking success rates using reinforcement learning methods

Innovation

• Reinforcement learning trains overtaking agent
• Agent deployed on F1Tenth racing vehicle
• Training against opponents enables deliberate overtaking