AI Infrastructure and Systems

Scope

The track of AI Infrastructure and Systems aims to bring together researchers and practitioners working on the foundations, architectures, and system-level innovations that power modern artificial intelligence. As AI models continue to grow in scale and complexity, there is an urgent need for robust, scalable, and efficient infrastructure to support training, deployment, and lifecycle management of AI workloads. This track focuses on the design, implementation, and optimization of AI-centric systems spanning cloud, edge, and hybrid environments. We welcome research that addresses the essential challenges of large-scale model training, distributed inference, heterogeneous hardware acceleration, resource orchestration, system reliability, and observability. We are particularly interested in research work that bridges AI algorithms and system design, enabling co-optimization across hardware, software, and networking layers. Our motivation is to foster cross-disciplinary dialogue between systems researchers, AI practitioners, hardware designers, and industry engineers to advance next-generation AI infrastructure that is scalable, reliable, trustworthy, and efficient.

Topics of Interest

Our track seeks original contributions in the following topical areas, plus others that are not explicitly listed but are closely related:

• Distributed and large-scale training systems for foundation models
• AI-native cloud and data center architectures
• Reliability, availability, and serviceability of AI systems
• Observability for AI infrastructure and systems
• Heterogeneous computing for AI (over GPU, TPU, NPU, FPGA, ASIC accelerations)
• High-performance networking and communication for AI workloads
• Efficient inference systems and serving architectures
• Edge AI systems and collaborative cloud-edge intelligence
• Resource management and scheduling for AI clusters
• AI workload characterization and benchmarking
• Storage systems optimized for AI training and data pipelines
• MLOps, AIOps, and lifecycle management of AI models
• Security, privacy, and trust of AI models
• Federated and distributed AI systems
• System support for generative AI and large language models
• Co-design of AI algorithms and system architectures

Track Co-Chairs

• Dian Shen, Southeast University, China
• Haoyu Zhang, Meta, USA
• Binhang Yuan, The Hong Kong University of Science and Technology, Hong Kong SAR, China

RF Computing, Next-Gen AIoT, and Embodied Intelligence

Scope

We are on the cusp of a paradigm shift in the Internet of Things (IoT). The field is evolving from simple connectivity to a sophisticated ecosystem of pervasive sensing, distributed intelligence, and physical interaction. This track seeks to explore the synergy between three transformative technologies: RF Computing, which enables battery-free and non-intrusive sensing; AIoT, which brings intelligence to the edge; and Embodied AI, which translates digital insights into physical action. This track invites researchers, practitioners, and industry experts to submit original contributions that bridge the gap between wireless signals, algorithmic intelligence, and robotic systems. We are particularly interested in works that demonstrate how RF signals can inform embodied agents, how AIoT architectures can support mobile robots, and how these systems can operate efficiently in real-world environments.

Topics of Interest

We solicit papers covering a broad range of topics related to this convergence. Topics of interest include, but are not limited to:

1. RF Computing & Wireless Sensing
   • Joint Communication and Sensing (JCAS) / ISAC
   • WiFi, LoRa, and UWB Sensing (Gesture, Vital Signs, Occupancy)
   • Backscatter Communication and Computational RFID
   • RF-based Localization and Tracking for Autonomous Systems
   • Energy Harvesting and Battery-free Computing
   • Reconfigurable Intelligent Surfaces (RIS) for Sensing
2. AIoT & Edge Intelligence
   • TinyML and Efficient Neural Networks for Edge Devices
   • Distributed Inference and Learning across IoT Clusters
   • Cross-modal Learning (RF + Vision + Audio)
   • Resource-constrained Learning for Embedded Systems
   • Privacy-preserving AI in IoT (Federated Learning, Split Learning)
   • Security in IoT systems
3. Embodied AI & Robotics
   • RF-Guided Navigation and SLAM (Simultaneous Localization and Mapping)
   • Sim-to-Real Transfer for Wireless Robots
   • Human-Robot Interaction via Wearables/RF Sensing
   • Multi-agent Coordination in AIoT Environments
   • Sensor Fusion for Embodied Agents (Vision-RF, Lidar-RF)
4. Systems & Applications
   • Smart Home, Smart Health, and Smart Factory Applications
   • Testbeds, Datasets, and Evaluation Metrics for RF-AI Systems
   • Hardware-Software Co-design for Sensing and Actuation

Track Co-Chairs

• Wei Wang, Wuhan University, China
• Yuan He, Tsinghua University, China
• Lu Wang, Shenzhen University, China

Web3.0 Security and Privacy

Scope

Web3.0 represents the next phase of Internet evolution, characterized by emerging technologies such as blockchain, cryptographic verification, and programmable smart contracts. These technologies enable new economic models, enhance transparency and security, and support trustworthy digital interactions at scale. The integration of AI further improves usability and automation, accelerating the adoption of intelligent, data-driven services. Together, these advances position Web3.0 as a transformative paradigm for the digital future.

The Web3.0 Security and Privacy track aims to bring together innovative research that safeguards the next generation of decentralized digital infrastructures. We welcome contributions that advance the cryptographic foundations, protocol robustness, and application-level security of Web3.0, spanning decentralized systems, smart contracts, DeFi ecosystems, DAOs, and cross-chain environments. This track seeks to foster interdisciplinary dialogue among researchers and practitioners in cryptography, system security, economics, privacy engineering, and trusted hardware, highlighting solutions that enhance resilience against emerging threats while preserving usability and compliance. Particular interest lies in privacy-preserving blockchain architectures, Layer-2 scalability, secure integration in critical industries, AI-driven security analytics, and empirical measurement of real-world Web3.0 risks. Our motivation is to bridge theoretical breakthroughs with practical deployment challenges, enabling trustworthy, transparent, and resilient Web3.0 systems that can support sustainable innovation across digital society.

Topics of Interest

Our track seeks original contributions in the following topical areas, plus others that are not explicitly listed but are closely related:

• Cryptographic foundations for Web3.0
• Decentralized protocol security
• Decentralized application and smart contract security
• DeFi security and economic resilience
• Governance and DAO security
• Privacy-preserving blockchain systems
• Layer-2 and cross-chain security
• System security for decentralized infrastructure
• Trusted execution and hardware-assisted Web3.0
• Secure Web3.0 integration in critical industries
• Web3.0 data analytics and forensics
• Usability, compliance, and human-centric privacy in Web3.0
• AI and emerging technologies in Web3.0 security and privacy
• Web3.0 measurement and empirical studies

Track Co-Chairs

• Ting Chen, University of Electronic Science and Technology of China, China
• Chunhua Su, University of Aizu, Japan
• Zihao Li, University of Electronic Science and Technology of China, China

Agentic Design for System and Network

Scope

Intelligent agents are rapidly reshaping how operating systems and networks are designed, controlled, and secured. Driven by breakthroughs in large language models, reinforcement learning, and multi-agent coordination, agents are increasingly capable of perceiving complex system states, reasoning over operational objectives, and executing adaptive actions across distributed infrastructures. From self-healing networks and autonomous traffic engineering to AI-driven security orchestration and intelligent resource scheduling, agent-based paradigms offer potential for modern systems and network management.

The Agentic Design for System and Network track aims to bring together pioneering research at the intersection of intelligent agents and systems/network engineering. We welcome contributions that explore how autonomous and semi-autonomous agents can enhance the performance, reliability, security, and efficiency of computing systems and communication networks. This track seeks to foster interdisciplinary dialogue among researchers and practitioners in distributed systems, networking, artificial intelligence, and security, highlighting solutions that harness agent intelligence to address real-world operational challenges. Particular interest lies in LLM-driven network management, multi-agent collaboration for distributed systems, agent-based cyber defense, autonomous cloud orchestration, and empirical evaluation of agent deployments in production environments. Our motivation is to bridge the gap between agent intelligence and systems engineering, enabling robust, adaptive, and self-managing infrastructures that can meet the demands of an increasingly complex and dynamic digital landscape.

Topics of Interest

Our track seeks original contributions in the following topical areas, plus others that are not explicitly listed but are closely related:

1. Agent for System
   • LLM-based and foundation model agents for operating system management and automation
   • Multi-agent coordination for distributed computing and parallel workload scheduling
   • Autonomous resource provisioning and orchestration in cloud and edge environments
   • Agent-driven fault detection and root cause analysis
   • Reinforcement learning agents for adaptive performance optimization
   • Agentic frameworks for testing, debugging, and program repair
2. Agent for Network
   • Intelligent agents for autonomous network configuration and management
   • LLM-driven network traffic engineering, routing optimization, and congestion control
   • Multi-agent systems for software-defined networking and network function virtualization
   • Autonomous agents for next-generation network environments
   • Agent-assisted network protocol design, verification, and simulation
3. Agent for Cybersecurity
   • LLM-based and agentic approaches for vulnerability discovery
   • Autonomous agents for intrusion detection and threat hunting
   • Agent-driven malware analysis, reverse engineering, and forensic investigation
   • Adversarial robustness of system-oriented agents
   • Privacy and ethical challenges in deploying agents for cybersecurity

Track Co-Chairs

• Yuchao Zhang, Beijing University of Posts and Telecommunications, China
• Chen Tian, Nanjing University, China
• Chuanpu Fu, Nanyang Technological University, Singapore

Edge Intelligence

Scope

With the rapid evolution of artificial intelligence (AI) and edge computing, Edge Intelligence (EI) has emerged as a transformative paradigm that shifts AI capabilities from centralized cloud data centers to the network edge—closer to end devices, sensors, and real-world scenarios. As the core driving force of the current AI era, EI is no longer a supplementary technology but the "first scene" for AI implementation, addressing critical demands for low latency, reduced bandwidth consumption, enhanced privacy protection, and efficient energy usage in diverse application domains. From high-performance edge AI chips and on-device large model deployment to embodied intelligence and edge-agent systems, EI is reshaping industries ranging from industrial manufacturing and autonomous vehicles to smart homes and healthcare.

Despite remarkable advancements—such as the deployment of large language models (LLMs) on edge devices, the rise of high-efficiency edge SoCs, and the integration of EI with digital twin and robotic systems—significant challenges remain. These include optimizing resource-constrained edge inference, addressing hardware-software co-design complexities, ensuring security and privacy in distributed edge environments, resolving toolchain fragmentation, and bridging the gap between large foundation models and edge deployment feasibility. To accelerate the advancement and adoption of EI, we invite original, high-quality research papers that explore novel theories, methodologies, technologies, and applications in this dynamic field. This call for papers aims to provide a platform for researchers, engineers, and practitioners from academia and industry to share breakthroughs, exchange ideas, and shape the future of Edge Intelligence.

Topics of Interest

We welcome submissions on a wide range of topics related to Edge Intelligence, including but not limited to:

1. Edge AI Model Optimization
   • Quantization, pruning, distillation, and compression techniques for resource-constrained edge devices
   • On-device deployment of LLMs, vision-language models, and multimodal models.
2. Edge Hardware & Software Co-Design
   • High-efficiency edge SoCs and NPUs
   • Heterogeneous computing architectures
   • Non-von Neumann paradigms for edge scenarios
3. Edge-Cloud & Edge-Edge Collaboration
   • Task offloading, model partitioning, and collaborative inference strategies
   • Federated learning and decentralized training for edge networks
   • Edge intelligence agents and their orchestration frameworks
4. Security, Privacy & Trust in EI
   • Privacy-preserving edge AI
   • Secure aggregation, differential privacy, and zero-trust mechanisms for edge networks
   • Defense against model stealing, data poisoning, and inference-side attacks
5. EI System Innovations
   • Edge AI operating systems (Agent OS)
   • Real-time scheduling and resource management
   • Energy-efficient computing for battery-powered edge devices
   • Benchmarks for heterogeneous edge platforms
6. Real-World EI Applications
   • Industrial IoT and autonomous factories
   • Autonomous vehicles and robotic systems
   • Smart healthcare, smart cities, and consumer electronics
   • AR/VR with edge intelligence support
7. Emerging Trends in EI
   • Physical AI
   • Semantic communication integrated with edge intelligence
   • 6G-edge AI convergence
   • Edge AI for sustainability and green computing

Track Co-Chairs

• Deyu Zhang, Central South University, China
• Peng Yang, Huazhong University of Science and Technology, China
• Sheng Yue, SUN YAT-SEN University, China

Intelligent Computing

Scope

The Intelligent Computing track at IEEE ICPADS 2026 focuses on the co-design of theoretical foundations and algorithmic innovations with parallel and distributed systems to enable scalable, efficient, and trustworthy learning and decision-making. The track emphasizes distributed training and inference, communication-efficient optimization, heterogeneous computing, and AI-driven system optimization across cloud, edge, and high-performance computing platforms. We welcome contributions that bridge theory and practice, including novel models, architectures, and applications that advance intelligent computing in large-scale, real-time, and resource-constrained environments. Research integrating intelligent computing models with parallel computing theory, distributed systems, and high-performance computing is strongly encouraged.

Our goal is to collaborate intelligent computing technologies applied in the field of embodied robot, services robot, and other intelligent applications in which parallel and distributed systems communities to advance intelligent systems or devices.

Topics of Interest

Our track seeks original contributions in the following topical areas, plus others that are not explicitly listed but are closely related:

• Deep Learning models and applications
• Distributed and Parallel Machine Learning
• Scalable Distributed Training and Optimization
• Communication-efficient Distributed Optimization
• Foundation Models and Large-scale training in Distributed Environments
• Graph Neural Networks and Large-scale Graph Intelligence
• Spatio-temporal and Sequential Learning Models
• Federated and Privacy-preserving Distributed Learning Systems
• Edge-Cloud Collaborative Intelligence Systems
• AI for High-Performance Computing
• Intelligent Resource Scheduling and System Optimization
• Data-parallel and Model-parallel Computing Strategies
• Trustworthy and Robust AI in Distributed Systems
• Heterogeneous Computing for Intelligent Workloads (CPU/GPU/TPU)
• Efficient Distributed Inference and Serving Systems
• Energy-efficient AI in Distributed Environments
• Intelligent Computing Applications on Scalable Distributed Platforms
• Intelligent computing upgrading traditional industry systems

Track Co-Chairs

• Songwen Pei, Shanghai Institute of Technology, China
• Joel Rodrigues, Federal University of Piauí (UFPI), Teresina - PI, Brazil
• Chao Li, Shanghai Jiao Tong University, China
• Zhihui Lv, Fudan University, China
• Liang Shi, East China Normal University, China