LAUSR

Safety and operational efficiency (SOE) are of great significance for the development of a smart metro system (SMS). To improve the SOE of SMS, Internet-of-Things technology is applied first to collect metro environmental data (MED), and then these data are analyzed by intelligent algorithms to achieve safety risk prediction, defect detection or operational efficiency improvement. As MEDs are generated quickly in the practical engineering field and most algorithms to process these data also have high computational complexity, the SMS must have sufficient computing power to process these MEDs in time. However, modern train computing resources are insufficient to meet this requirement. To address this challenge, edge intelligence (EI) technology, which enables trains to offload computationally-intensive tasks to nearby EI systems, is proposed. This article aims to develop an energy-efficient and high-performance EI system to improve the SOE of SMSs. An EI architecture that considers the three driving forces, data, algorithms and computing power, is proposed first, and then the enabling techniques of this EI architecture, including intelligent algorithms, domain-specific architecture (DSA)-based hardware acceleration, end-edge-cloud collaborative computing, hardware platform management, and security issues, are investigated. This EI system can enable SMS to process the large-scale MEDs with not only high accuracy but also low latency and low energy cost. As a study case, a real-world EI system is built to run three kinds of SMS applications to assess the safety risks of SMSs. The evaluation results demonstrate the effectiveness of the proposed schemes in this article.