HF RFID--based Book Localization Via Mobile Scanning In Libraries

The book localization is one of the most important functions for inventory man-agement in large libraries,which is able to help library users find the misplaced books and pinpoint where they are.Traditional solutions suffer from intensive manual labor,long scanning delay,or high-cost infrastructure renewal,which is a great barrier to commercial adoption.Based on this,some advanced works perform book localization automatically by combining the robot vehicle and the RFID technology.The robot is used for replacing librarians to scan along the bookshelf.The RFID device is installed on the robot and used for locating books.However,most of them use UHF RFID technology for locating books,which fail to fit libraries that have deployed HF RFID systems.Few of them which use HF RFID for locating books require complex de-ployment or focus on key robotic enabling technology rather than high accurate book localization.Hence,we in this dissertation propose a new HF RFID-based book lo-calization system by integrating robot vehicle and HF RFID.The biggest challenge is that the commodity HF RFID device cannot capture RF phase,which is a fine-grained metric with high resolution for locating.Considering above limitations,we in this paper proposes a two-stage mobile lo-calization framework.The key point is the influence of the distance difference between the tag and the reader antenna on the energy acquisition of the tag.By analyzing and extracting features from the raw RFID data,we can accurately locate each book.The contributions of this paper are summarized as follows:(1)To address book localization problem,we break down the task of the localiza-tion into two stages,tier-level localization and order estimation.The tasks of tier-levellocalization and order estimation are determining in which bookshelf,in which tier,and in which order a book is placed,which helps users find out the books of interest quickly.(2)For the tier-level localization,we take the effect of variabilities in distances between RFID tags in different tiers and reader antenna on tags' energy acquisition as starting point.A fuzzy inference model based on two useful metrics,number of reads and mean RSSI value,is proposed to accurately predict in which bookshelf,in which tier a book is located.(3)For the order estimation,we propose a new one-dimensional Bayesian infer-ence model based on number of reads and mean RSSI of tag.We treat a tier of the bookshelf as X- Axis and the books(tags)in the tier as points on the X-axis.After that,we use the one-dimensional Bayesian inference model to predict the coordinate value of each point on the X-axis,and then each book is sorted according to it's coordinate value,which we can achieve high ordering accuracy.(4)We implement a book localization system via mobile scanning by integrating COTS HF RFID device and self-designed AGV.Long-term extensive experimental re-sults show that our system can achieve 97% tier-level localization accuracy and 90%order accuracy,which is able to achieve cm-level localization along X-axis.