TinyLidarNet: 2D LiDAR-based End-to-End Deep Learning Model for F1TENTH Autonomous Racing

TL;DR

TinyLidarNet is a lightweight 2D LiDAR-based end-to-end deep learning model for autonomous racing, demonstrating competitive performance and real-time processing capabilities on low-end hardware.

Contribution

The paper introduces TinyLidarNet, a novel lightweight 2D LiDAR-based deep learning model that outperforms MLP architectures and is suitable for real-time autonomous racing on micro-controllers.

Findings

TinyLidarNet won 3rd place at F1TENTH Autonomous Grand Prix.

It outperforms MLP-based architectures in accuracy.

It can run in real-time on low-end micro-controllers.

Abstract

Prior research has demonstrated the effectiveness of end-to-end deep learning for robotic navigation, where the control signals are directly derived from raw sensory data. However, the majority of existing end-to-end navigation solutions are predominantly camera-based. In this paper, we introduce TinyLidarNet, a lightweight 2D LiDAR-based end-to-end deep learning model for autonomous racing. An F1TENTH vehicle using TinyLidarNet won 3rd place in the 12th F1TENTH Autonomous Grand Prix competition, demonstrating its competitive performance. We systematically analyze its performance on untrained tracks and computing requirements for real-time processing. We find that TinyLidarNet's 1D Convolutional Neural Network (CNN) based architecture significantly outperforms widely used Multi-Layer Perceptron (MLP) based architecture. In addition, we show that it can be processed in real-time on low-end micro-controller units (MCUs).