Place recognition and re-localization for last-mile delivery robot

• Main Author: Xu, Lihaotian
• Other Authors: Wang Dan Wei
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/168056
• Institution: Nanyang Technological University

The need for delivery services has increased due to the growth of e-commerce and online shopping, making it more difficult for delivery companies to compete. Among the procedure, the last-mile delivery issue is the trickiest and most expensive part. To solve it, one approach is the use of delivery robots, which have a number of advantages over conventional delivery techniques. However, in order to perform daily distribution activities, last-mile delivery robots need a precise and dependable self-positioning mechanism. This paper introduces the concept of SLAM (Simultaneous Localization and Mapping) as a technique used in robotics and computer vision to address the issue of mapping an unknown environment while also figuring out where the robot is in that environment. The paper implements a SLAM algorithm (SC-LIO-SAM) on a delivery robot and verifies its performance by running various datasets. Experimental results show that the proposed framework has good precision, recall and viewpoint robustness on both public benchmark and self-built campus datasets. Paper also concludes that LiDAR-based SLAM algorithm have the potential to significantly improve the performance of last-mile delivery robots and lead to more efficient, reliable delivery operations.