Design and development of full self-service smart shopping system for unmanned supermarket

Abstract Traditional supermarkets often encounter challenges such as inefficient shopping guidance, long checkout lines, and poor customer experience. To address these issues, this study proposes a fully self-service smart shopping system, comprising three key modules: a smart shopping cart, a client app, and a back-end management platform, with full-chain encryption using TLS (Transport Layer Security) and RSA (Rivest Shamir Adleman) to ensure secure data transmission. The smart shopping system utilizes a KNN (K Nearest Neighbor) collaborative filtering recommendation algorithm to improve the relevance and accuracy of product suggestions. By integrating Mask R-CNN and UWB (Ultra Wide Band) into SLAM (Simultaneous Localization and Mapping) framework, we construct MU-SLAM (Mask R-CNN UWB SLAM), which enables precise localization and robust obstacle avoidance in unknown supermarket environments. Additionally, integrating a distributed DIMP (Discriminative Model Prediction) visual-tracking algorithm and infrared range sensors ensures that the smart shopping cart always follows the customer at a safe distance. Comparative experiments show that the MU-SLAM algorithm’s positioning accuracy increased by 92%, while the distributed DIMP algorithm’s tracking persistence improved by 107%. Furthermore, validation of the real supermarket environment demonstrated that the smart shopping system respectively reduces average shopping time and average checkout time by 33.15% and 90.17%, significantly improving both operational efficiency and the overall customer experience.