GTI 无源物联网——人工智能篇白皮书 GTI Passive Internet of Things - Artificial Intelligence White Paper

GTIAmbient IoT—Artificial Confidentiality:This document may contain information that is confidential and accessto this document is restricted to the persons listed in the Confidential Level. Thisdocument may not be used, disclosed or reproduced, in whole or in part, without theprior written authorization of GTI, and those so authorized may only use this document Document History Preface In recent years, facing development needs of industrial digitalization, smart home, andrefined social governance,Ambient IoTtechnology has entered a period of rapiddevelopment, becoming a key path to realize the networking and intelligence of massive"dumb" terminals. It is expected to form a trillion-level connection scale in warehousing, The copyright of this white paper belongs to China Mobile. Without authorization, anyunit or individual may not copy or reproduce any part or all of the contents of this Table of Contents 1. AI ×Ambient IoTTechnology Overview......................................................................72. AI Fusion Enhancing Basic Performance ofAmbient IoT.............................................92.1 AI ×Ambient IoTIntelligent Networking..............................................................92.2 AI ×Ambient IoTIntelligent Scheduling............................................................123. AI Fusion Expanding Additional Capabilities ofAmbient IoT......................................163.1 AI ×Ambient IoTSingle Device Fusion Sensing...............................................163.2 AI ×Ambient IoTLow-Cost Positioning............................................................193.3 AI ×Ambient IoTSpatial Sensing.....................................................................223.4 Logistics and Warehousing Scheduling Large Model Based on Passive Data..253.5Ambient IoTTag Security Intelligent Assurance...............................................284. Summary and Outlook...............................................................................................30Contributing Organizations and Personnel....................................................................31 1. AI ×Ambient IoTTechnology Overview In recent years, facing development needs of industrial digitalization, smart home, andrefined social governance,Ambient IoTtechnology has entered a period of rapiddevelopment.Ambient IoTterminals do not require built-in power sources and rely onenvironmental energy harvesting and backscatter communication, becoming a key path Facing the above technical problems, the industry has continued to explore paths fordeep integration of artificial intelligence andAmbient IoTin recent years. Through First, in terms of basic performance ofAmbient IoTsystems, the integration of AI hassignificantly improved dynamic network resource optimization capabilities.Ambient IoTintelligent networking and scheduling technologies utilize optimization algorithms and Second, in terms of additional capabilities ofAmbient IoTsystems, the integration of AIhas effectively expanded the sensing range and precision ofAmbient IoTand achieved"value leap" of passive data. At the level of sensing capability expansion, by integrating and-play" high-precision recognition and behavior judgment in scenarios such asclothing retail access control and smart equipment cabinets. Through passive positioningtechnology empowered by AI, positioning accuracy has achieved a leap to "centimeterlevel," providing low-cost, high-precision positioning solutions for scenarios such assmart warehousing and smart parking. AI spatial sensing technology transforms passive 2. AI Fusion Enhancing Basic Performance ofAmbient IoT The integration of AI andAmbient IoTis systematically improving its basiccommunication and networking capabilities, enablingAmbient IoTto achieve moreefficient and reliable end-to-end connections under the premise of low cost. AI ×Ambient 2.1 AI ×Ambient IoTIntelligent Networking Ambient IoTsystems are composed of reading and writing nodes, excitation nodesdeployed in different geographical locations, and tags attached to items. Theirperformance is easily affected by environmental and inter-device interference, andnetwork deployment in complex scenarios has been proven to be an NP-hard problem.Therefore, efficient intelligent networking technology becomes the key to whether the Ambient IoTnetwork system deployment needs to simultaneously optimize multipleobjectives such as tag coverage rate, signal interference, and cost. Its intelligent solutionmethods for multi-objective optimization are mainly divided into two categories. The first efficiently solving them using swarm intelligence optimization algorithms and buildingnonlinear integer programming models. Among them, the fuzzy k-coverage method caneffectively handle coverage uncertainty problems in network planning; the two-levelplanning model built based on hierarchical decoupling principles significantly reducescomputational complexity by