Unifying F1TENTH Autonomous Racing: Survey, Methods and Benchmarks

TL;DR

This survey addresses fragmentation in F1TENTH autonomous racing by cataloging classical perception–planning–control pipelines and learning-based approaches, and by providing open benchmarks to enable direct method comparisons. It details particle-filter localisation, offline trajectory optimisation with tracking, model predictive contouring control, follow-the-gap, and end-to-end deep reinforcement learning, including training setups and reward designs such as TAL. Benchmark results indicate offline trajectory optimisation with tracking yields the fastest lap times, with MPCC as a close second, while learning-based methods vary in reliability and speed depending on reward signals and training maps. The work emphasizes rigorous full-stack evaluations, robustness, and future directions toward vision-based, 3D-aware, and multi-agent racing to advance practical applicability.

Abstract

The F1TENTH autonomous driving platform, consisting of 1:10-scale remote-controlled cars, has evolved into a well-established education and research platform. The many publications and real-world competitions span many domains, from classical path planning to novel learning-based algorithms. Consequently, the field is wide and disjointed, hindering direct comparison of developed methods and making it difficult to assess the state-of-the-art. Therefore, we aim to unify the field by surveying current approaches, describing common methods, and providing benchmark results to facilitate clear comparisons and establish a baseline for future work. This research aims to survey past and current work with F1TENTH vehicles in the classical and learning categories and explain the different solution approaches. We describe particle filter localisation, trajectory optimisation and tracking, model predictive contouring control, follow-the-gap, and end-to-end reinforcement learning. We provide an open-source evaluation of benchmark methods and investigate overlooked factors of control frequency and localisation accuracy for classical methods as well as reward signal and training map for learning methods. The evaluation shows that the optimisation and tracking method achieves the fastest lap times, followed by the online planning approach. Finally, our work identifies and outlines the relevant research aspects to help motivate future work in the F1TENTH domain.

Figures (18)

• Figure 1: The F1TENTH platform provides a development platform for autonomous algorithms with an accurate simulator allowing rapid deployment.
• Figure 2: Classical approaches use either offline planning with a separate control module, or online planning and control.
• Figure 3: End-to-end, planning, residual, and safe agent architectures for autonomous racing.
• Figure 4: Motion and measurement updates for the particle filter.
• Figure 5: Centre line, minimum curvature line, heading angle $\psi$, segment length $L$, and optimisation variable $\alpha$ and limits $[\alpha_{\text{min}}, \alpha_{\text{max}}]$ for a track segment.