Localization and 2D Mapping Using Low-Cost Lidar

Abstract

Autonomous vehicles are expected to make a profound change in auto industry. An autonomous vehicle is a vehicle that is able to sense its surroundings and travel with little or no human intervention. The four key capabilities of autonomous vehicles are a comprehensive understanding of sensor data, knowledge of their positions in the world, building the map of unknown environment, as well as following the planed route and collision avoidance.

This thesis is aimed at building a low-cost autonomous vehicle prototype that is capable of localization and 2D mapping simultaneously. In addition, the prototype should be able to detect obstacles and avoid collision. In this thesis, a Redbot is utilized as a moving vehicle to evaluate collision avoidance functionality. A mechnical bumper in front of the Redbot is used to detect obstacles, and a remote user can send appropriate commands to control the Redbot via Zigbee network, then Redbot acts accordingly, including driving straightly, changing direction to right or left, and stop. Redbot are also used to carry the lidar scanner which consists of Lidar Lite V3 and a servo motor. Lidar data are sent back to a Laptop running ROS via Zigbee network. In ROS, Hector SLAM metapackage is adopted to process the lidar data, and realize the functionality of simultaneous localization and 2D mapping.

After implementing the autonomous vehicle prototype, a series of tests are con- ducted to evaluate the functionality of localization, 2D mapping, obstacle detection, and collision avoidance. The results demonstrated that the prototype is capable of building usable 2D maps of unknown environment, simultaneous localization, obstacle detection and collision avoidance in time. Due to the limited scan range of the low-cost lidar scanner, boundary missing problem can happen. This limitation can be solved through the use of a lidar scanner with larger scan range.