Implementation of LiDAR and SLAM on a Small-Scale Autonomous Platform

Abstract

Autonomous driving (AD) technology continues to evolve with goals of enhancing road safety, lowering emissions and increasing transportation efficiency. Modern solutions rely on sensors including GPS, radar, camera and LiDARs to perceive their surroundings and enabling safe and precise navigation. Continued research and development in this field require accessible and adaptable platforms for doing so on.

This thesis, conducted in collaboration with Infotiv AB, builds on two previous master theses and focuses on developing such a platform and expanding it, making the development more available. The autonomous platform 4 (AP4) is a small-scale autonomous platform based on the Ninebot GoKart, which has been implemented with a Raspberry Pi 4b and ROS2 environment.

The focus of this thesis was to add onto the existing platform by integrating a LiDAR and explore sensor fusion possibilities with an IMU, both in simulation and in real-life. By doing this, the work aimed to improve localization and path planning through simultaneous localization and mapping (SLAM) to further improve the AP4. This was done while simultaneously keeping the platform modular and scalable.

A LiDAR and IMU pipeline has been integrated both in simulation and on the physical platform. For evaluation, a local go-kart track was rendered and used as grounds for both cases. The simulations proved that the concept using SLAM with a LiDAR and sensor fusion of an IMU can work for self driving on a small-scale platform. However, the implementation to the physical environment highlights the necessities of well tuned sensors and position estimations.