Enabling a Mobile Robot for Autonomous RFID-Based Inventory by Multilayer Mapping and ACO-Enhanced Path Planning

Abstract

Paper presents a novel application for an autonomous robot to perform RFID-based inventory in a retail environment. For this application, one challenge is to represent a complicated environment by a good quality map. LIDAR (light detection and ranging) sensors only generate a 2D plane map that loses a large amount of structural information. In contrast, stereo or RGB-D cameras provide abundant environmental information but in a limited field of view (FOV), which limits the robot’s ability to gain reliable poses. Another challenge is effectively counting inventory within a massive retail environment; the robot needs to navigate in an optimal route that covers the entire target area. To overcome the aforementioned challenges, we propose a multilayer mapping method combined with an Ant Colony enhanced path planning approach. Multilayer mapping utilizes a LIDAR and RGB-D camera (Microsoft Kinect camera) to obtain both accurate poses and abundant surrounding details to create a reliable map. To improve inventory efficiency, ACO-enhanced path planning is deployed to optimize the entire inventory route that minimizes total navigating distance without losing the inventory accuracy. Our experimental results show that multilayer mapping provides a precise and integrated map that enables the robot to navigate in a mock apparel store. Additionally, the efficiency of RFID-based inventory is greatly improved. Compared with the traditional method of manual inventory, ACO-enhanced path planning reduced total navigational distance by up to 28.2% while keeping inventory accuracy the same as before.